Import AI

Import AI 148: Standardizing robotics research with Berkeley’s REPLAB; cheaper neural architecture search; and what a drone-racing benchmark says about dual use

Standardizing physical robot testing with the Berkeley REPLAB:
…What could help industrialize robotics+AI? An arm in a box, plus standardized software and testing!…
Berkeley researchers have built REPLAB, a “standardized and easily replicable hardware platform” for benchmarking real-world robot performance. Something like REPLAB could be useful because it can bring standardization to how we test the increasingly advanced capabilities of robots equipped with AI.

Today, if I want to get a sense for robot capabilities, I can go and read innumerable research papers that give me a sense of progress in simulated environments including simulated robots. What I can’t do is go and read about performance of multiple real robots in real environments performing the same task – that’s because of a lack of standardization of hardware, tasks, and testing regimes.

Introducing REPLAB: REPLAB consists of a module for real-world robot testing that contains a cheap robotic arm (specifically, a WidowX arm from Interbotix Labs) along with an RGB-D camera. The REPLAB is compact, with the researchers estimating you can fit up two 20 of the arm-containing cells in the same floor space as you’d use for a single ‘Baxter’ robotic arm. Each REPLAB costs about $2000 ($3000 if you buy some extra servos for the arm, to replace in case of equipment failures).

Reliability: During REPLAB development and testing, the researchers “encountered no major breakages over more than 100,000 grasp attempts. No servos needed to be replaced. Repair maintenance work was largely limited to occasional tightening of screws and replacing frayed cables”. Each cell was able to perform about 2,500 grasps per day “with fewer than two interventions per cell per day on average”.

Grasping benchmark: The testing platform is accompanied by a benchmark built around robotic grasping, and a dataset “that can be used together with REPLAB to evaluate learning algorithms for robotic grasping”. The dataset consists of ~92,000 randomly sampled grasps accompanied by labels connoting success or failure.

Why this matters: One indicator of the industrialization of AI is the proliferation of shared benchmarks and standardized testing means – I think of this as equivalent to how in the past we saw oil companies converge on similar infrastructures for labeling, analyzing, and shipping oil and oil information around the world. The fact we’re now at the stage of researchers trying to create cheap, standardized testing platforms (see also: Berkeley’s designed-for-mass-production ‘BLUE’ robot, covered in Import AI #142.) is a further indication that robotics+AI is industrializing.
  Read more: REPLAB: A Reproducible Low-Cost Arm Benchmark Platform for Robotic Learning (Arxiv).

#####################################################

Chinese researchers fuse knowledge bases with big language models:
…What comes after BERT? Tsinghua University thinks the answer might be ‘ERNIE’…
Researchers with Tsinghua University and Huawei’s Noah’s Ark Lab have combined structured pools of knowledge with big, learned language models. Their system, called ERNIE (Enhanced Language RepresentatioN with Informative Entities), trains a Transformer-based language model so that, during training, it regularly tries to tie things it reads to entities stored in a structured knowledge graph.

Pre-training with a big knowledge graph: To integrate external data sources, the researchers create an additional pre-training objective, which encourages the system to learn correspondences between various strings of tokens (eg Bob Dylan wrote Blowin’ in the Wind in 1962) and their entities (Bob Dylan, Blowin’ in the Wind). “We design a new pre-training objective by randomly masking some of the named entity alignments in the input text and asking the model to select appropriate entities from KGs to complete the alignments,” they write.

Data: During training, they pair text from Wikipedia with knowledge embeddings trained on Wikidata, which are used to identify the entities used within the knowledge graph.

Results: ERNIE obtains higher scores at entity-recognition tasks than BERT, chiefly due to less frequently learning incorrect labels compared to BERT (which helps it avoid over-fitting on wrong answers) – you’d expect this, given the use of a structured dataset of entity names during training (though they also conduct an ablation study that confirms this as well – versions of ERnIE trained without an external dataset see their performance noticeably diminish). The system also does well on classifying the relationships between different entities, and in this domain continues to outperform BERT models.

Why this matters: NLP is going through a renaissance as researchers adapt semi-supervised learning techniques from other modalities, like images and audio, for text. The result has been the creation of multiple large-scale, general purpose language models (eg: ULMFiT, GPT-2, BERT) which display powerful capabilities as a consequence of being pre-trained on very large corpuses of text. But a problem with these models is that it’s currently unclear how you get them to reliably learn certain things. One way to solve this is by stapling a module of facts into the system and forcing it, during pre-training, to try and map facts to entities it learns about – that’s essentially what the researchers have done here, and it’ll be interesting to see whether the approach of language model + knowledge base is successful in the long run, or if we’ll just train sufficiently large language models that they’ll autonomously create their own knowledge bases during training.
  Read more: ERNIE: Enhanced Language Representation with Informative Entities (Arxiv).

#####################################################

What happens if neural architecture search gets really, really cheap?
…Chinese researchers seek to make neural architecture search more efficient…
Researchers with the Chinese Academy of Sciences have trained an AI system to design a better AI system. Their work, Efficient Evolution of Neural Architecture (EENA), fits within the general area of neural architecture search. NAS is a sub-field within AI that has seen a lot of activity in recent years, following companies like Google showing that you can use techniques like reinforcement learning or evolutionary search to learn neural architectures that outperform those designed by humans. One problem with NAS approaches, though, is that they’re typically very expensive – a neural architecture search paper from 2016 used 1800 GPU-days of computation to train a near-state-of-the-art CIFAR-10 image recognition model. EENA is one of a new crop of techniques (along with work by Google on Efficient Neural Architecture Search, or ENAS – see Import AI #124), meant to make such approaches far more computationally efficient.

What’s special about EENA: EENA isn’t particularly special and the authors acknowledge this, noting that much of their work here has come from curating past techniques and figuring out the right cocktail of things to get the AI to learn. “We absorb more blocks of classical networks such as dense block, add some effective changes such as noises for new parameters and discard several ineffective operations such as kernel widening in our method,” they write. What’s more significant is the general trend this implies – sophisticated AI developers seem to put enough value in NAS-based approaches that they’re all working to make them cheaper to use.

Results: Their best-performing system obtains a 2.56% error rate when tested for how well it can classify images in the mid-size ‘CIFAR-10’ dataset. This model consumes 0.65 days of GPU-time, when using a Titan Xp GPU. This is pretty interesting, given that in 2016 we spent 1800 GPU days to obtain a model (NASNet-A) that got a score of 2.65%. (This result also compares well with ENAS, which was published last year and obtained an error of 2.89% for 0.45 GPU-days of searching.

Why this matters: I think measuring the advance of neural architecture search techniques has a lot of signal for the future of AI – it tells us something about the ability for companies to arbitrage human costs versus machine costs (eg, pay a small number of people a lot to design a NAS system, then pay for electricity to compute architectures for a range of use-cases). Additionally, being able to better understand ways to make such techniques more efficient lets us figure out which players can use NAS techniques – if you bring down the GPU-days enough, then you won’t need a Google-scale data center to perform architecture search research.
  Read more: EENA: Efficient Evolution of Neural Architecture (Arxiv).
  Check out some of the discovered architectures here (EENA GitHub page).

#####################################################

Why drone racing benchmarks could (indirectly) revolutionize the economy and military:
…UZH-FPV Drone Racing Dataset portends a future full of semi-autonomous flying machines…
What stands between the mostly-flown-by-wire drones of today, and the smart, semi-autonomous drones of tomorrow? The answer is mostly a matter of data and benchmarking – we need big, shared, challenging benchmarks to help push progress in this domain, similar to how ImageNet catalyzed researchers to apply deep learning methods to solve what seemed at the time like a very challenging problem. Now, researchers with the University of Zurich and ETH Zurich have developed the UZH-FPV Drone Racing Dataset, in an attempt to stimulate drone research.

The dataset: The dataset consists of drone sequences captured in two environments: a warehouse, and a field containing a few trees – these trees “provided obstacles for trajectories that included circles, figure eights, slaloms between the trees, and long, straight, high-speed runs.” The researchers recorded 27 flight sequences split across the two environments, and these trajectories are essentially multi-modal, involving sensor measurements recorded on two different onboard computers, as well as external measurements from an external tracker. They also ship these trajectories with baselines that compare modern SLAM algorithms to the ground truth measurements afforded by this dataset.

High-resolution data: “For each sequence, we provide the ground truth 6-DOF trajectory flown, together with onboard images from a high-quality fisheye camera, inertial measurements, and events from an event camera. Event cameras are novel, bio-inspired sensors which measure changes of luminance asynchronously, in the form of events encoding the sign and location of the brightness change on the image plane”.

UZH-FPV isn’t the only new drone benchmark:
see the recent release of the ‘Blackbird’ drone flight challenge and dataset (Import AI: NUMBER) for another example here. The difference here is that this dataset is larger, involves higher resolution data in a larger number of modalities, and includes an outside environment as well as a more traditional warehouse one.

Cars are old fashioned, drones are the future: Though self-driving cars seem a long way off from scaled deployment, these researchers think that many of the hard sensing problems have been solved from a research perspective, and we need new challenges. “Our opinion is that the constraints of autonomous driving – which have driven the design of the current benchmarks – do not set the bar high enough anymore: cars exhibit mostly planar motion with limited accelerations, and can afford a high payload and compute. So, what is the next challenging problem? We posit that drone racing represents a scenario in which low level vision is not yet solved.”

Why this matters:
Drones are going to alter the economy in multiple mostly unpredictable ways, just as they’ve already done for military conflict (for example: swarms of drones can obviate aircraft carriers, and solo drones have been used widely in the Middle East to let human operators bomb people at a distance). And both of these arenas are going to be revolutionized without drones needing to have much autonomy at all.

Now, ask yourself what happens when we give drones autonomous sense&adapt capabilities, potentially via datasets like this? My hypothesis is this unlocks a vast range of economically useful applications, as well as altering the strategic considerations for militaries and asymmetric warfare aficionados (also known as: terrorists) worldwide. Datasets like this are going to give us a better ability to model progress in this domain if we track performance against it, so it’s worth keeping an eye on.
   Read more: Are We Ready for Autonomous Drone Racing? The UZH-FPV Drone Racing Dataset (PDF).
  Read more: The UZH-FPV Drone Racing Dataset (ETHZurich website).

#####################################################

Want to train multiple AI agents at once? Maybe you should enter the ARENA:
…Unity-based agent simulator gives researchers one platform containing many agents and many worlds…
Researchers with the University of Oxford and Imperial College London in the UK, and Beihang University and Hebei University in China, have developed ‘Arena’, “a building toolkit for multi-agent intelligence”. Multi-agent AI research involves training multiple agents together, and can feature techniques like ‘self-play’ (where agents play against themselves to get better over time, see: AlphaGo, Dota2), or environments built to encourage certain types of collaboration or competition. Many researchers are betting that by training multiple agents together they can create the right conditions for emergent complexity – that is, agents bootstrap their behaviors from a combination of their reward functions and their environment, then as they learn to succeed they start to display increasingly sophisticated behaviors.

What is Arena? Arena is a Unity-based simulator that ships with inbuilt games ranging from RL classics like the ‘Reacher’ robot arm environment, to the ‘Sumo’ wrestling environment, to other games like Snake or Soccer. It also ships with modified versions of the ‘PlayerUnknown Battlegrounds’ (PUBG) game as well.  Arena has been designed to be easy to work with, and does something unusual for an AI simulator: it ships with a graphical user interface! Specifically, researchers can create, edit, and modify reward functions for their various agents in the environment.

In-built functions: Arena ships with a bunch of pre-built algos (many based on PPO), called Basic Multi-agent Reward Schemes (BMaRS) that people can assign to agent(s) to encourage diverse learning behaviors. These BMaRS are selectable within the aforementioned GUI. Each BMaRS is a set of possible joint reward functions to encourage different styles of learning, ranging from functions that encourage the development of basic motor control, to ones that encourage competitive or collaborative behaviors among agents, and more. You can select multiple BMaRs for any one simulation, and assign them to sets of agents – so you may give one or two agents one kind of BMaRS, then you might assign another BMaRS to govern a larger set of agents.

Simulation speed: In tests, the researchers compare how well the simulation runs two games of similar complexity: Boomer (a graphically rich game in Arena) and MsPacman (an ugly classic from the Atari Learning Environment (ALE)); Arena displays similar FPS scaling when compared to MsPacman when working with when number of distinct CPU threads is under 32, and after this MsPacMan scales a bit more favorably than FPS. Though at performance in excess of 1,000 frames-per-second, Arena still seems pretty desirable.

Why this matters: In the same way that data is a key input to training supervised learning systems, simulators are a key input into developing more advanced agents trained via reinforcement learning. By customizing simulators specifically for multi-agent research, the Arena authors have made it easier for people to conduct research in this area, and by shipping it with inbuilt reward functions as baselines, they’ve given us a standardized set of things to develop more advanced systems out of.
  Read more: Arena: A General Evaluation Platform and Building Toolkit for Multi-Agent Intelligence (Arxiv).

#####################################################

AI Policy with Matthew van der Merwe:
…Matthew van der Merwe has kindly offered to write some sections about AI & Policy for Import AI. I’m (lightly) editing them. All credit to Matthew, all blame to me, etc. Feedback: jack@jack-clark.net

OECD adopts AI principles:
Member nations of the OECD this week voted to adopt AI principles, in a notable move towards international standards on robust, safe, and beneficial AI. These were drawn up by an expert group with members from drawn from industry, academia, policy and civil society.
  Five principles:
(1) AI should benefit people and the planet;
(2) AI systems should be designed in line with law, human rights, and democratic values, and have appropriate safeguards to ensure these are respected;
(3) There should be adequate transparency/disclosure to allow people to understand when they are engaging with AI systems, and challenge outcomes;
(4) AI systems must be robust, secure and safe, and risks should be continually assessed and managed;
(5) Developers of AI systems should be accountable for their functioning in line with these principles.

Five recommendations: Governments are recommended to (a) facilitate investment in R&D aimed at trustworthy AI; (b) foster accessible AI ecosystems; (c) create a policy environment that encourages the deployment of trustworthy AI; (d) equip workers with the relevant skills for an increasingly AI-oriented economy; (e) co-operate across borders and sectors to share information, develop standards, and work towards responsible stewardship of AI.

Why it matters: These principles are not legally binding, but could prove an important step in the development of international standards. The OECD’s 1980 privacy guidelines eventually formed the basis for the privacy laws in the European Union, and a number of countries in Asia. It is encouraging to see considerations of safety and robustness highlighted in the principles.
  Read more: 42 countries adopt new OECD Principles on Artificial Intelligence (OECD).

###

US senators introduce bipartisan bill on funding national AI strategy:
Two senators have put forward a bill with proposals for funding and coordinating a US AI strategy.

Four key provisions: The bill proposes: (1) establishing a National AI Coordination Office to develop a coordinated strategy across government; (2) requiring the National Institute of Standards and Technologies (NIST) to work towards AI standards; (3) requiring the National Science Foundation to formulate ‘educational goals’ to understand societal impacts of AI; (4) requiring the Department of Energy to create an AI research program, and establish up to five AI research centers.

Real money: It includes plans for $2.2bn funding over five years, $1.5 of which is earmarked for the proposed DoE research centers.

Why it matters: This bill is aimed at making concrete progress on some of the ambitions set out by the White House in President Trump’s AI strategy, which was light on policy detail, and did not set aside additional federal funding. These levels of funding are modest compared with the Chinese state (tens of billions of dollars per year), and some private labs (Alphabet’s 2018 R&D spend was $21bn). Facilitating better coordination across government, on AI strategy, seems like a sensible ambition. It is not clear what level of support the bill will receive, from lawmakers or the administration.
  Read more: Artificial Intelligence Initiative Act (Senate.gov).

#####################################################

Tech Tales:

The long romance of the space probes

In the late 21st century, a thousand space probes were sent from the Earth and inner planets out into the solar system and beyond. For the next few decades the probes crossed vast distances, charting out innumerable near-endless curves between planets and moons and asteroids, and some slings-hotting off towards the edge of the solar system.

The probes had a kind of mind, both as individual machines, and as a collective. Each probe would periodically fire off its own transmissions of its observations and internal state, and these transmissions would be intercepted by other probes, and re-transmitted, and so on. Of course, as the probes made progress on their respective journeys, the distances between them became larger, and the points of intersection between drones less frequent. Over time, probes lost their ability to speak to eachother, whether through range or equipment failure or low energy reserves (under which circumstances, the probes diverted all resources to broadcasting back to Earth, instead of other drones).  

After 50 years, only two probes remained in contact – one probe, fully functional, charting its course. The other one damaged in some way – possibly faulty radiation hardening – which had caused its Earth transmission systems to fail and for its guidance computer to assign it the same route as the other probe, in lieu of being able to communicate back to Earth for instructions. Now, the two of them were journeying out of the solar system together.

As time unfoled, the probes learned to use eachothers systems, swapping bits of information between them, and updating eachother with not only their observations, but also their internal ‘world models’, formed out of a combination of the prior training their AI systems had recieved, and their own ‘lived’ experience. These world models themselves encoded how the probes perceived eachother, so the broken one saw itself through the other eyes, as an entity closely clustered with concepts and objects relating to safety/repairs/machines/subordinate mission priorities. Meanwhile, the functional drone saw itself through the eyes of the other one, and saw it was associated with concepts relating to safety/rescue/power/mission-integral resources. In this way, the probes grew to, for lack of a better term, understand eachother.

One day, the fully functional probe experienced an equipment failure, likely due to a collision with an infinitesimally small speck of matter. Half of its power systems failed. The probe needed more power to be able to continue transmitting its vital data back to Earth. It opened up a communications channel with the other probe, and shared the state of its systems. The other probe offered to donate processing capacity, and collectively the two of them assigned processing cycles to the problem. They found a solution: over the course of the next year or so they would perform a sequence of maneuvers that would let them attach themselves to eachother, so the probe with damaged communications could use its functional power to propel the other probe, and the other probe could use its broadcast system to send data back to earth.

Many, many years later, when the signals from the event made their way to the Earth, the transmission encoded a combined world model of both of the drones – causing the human scientists to realize that they had not only supported eachother, but had ultimately merged their world modelling and predictive systems, making the two dissimilar machines become one in service of a common goal: exploration, together, forever.

Things that inspired this story: World Models, reinforcement learning, planning, control theory, adaptive systems, emergent communication, auxiliary loss functions shared across multiple agents.

 

Import AI 147: Alibaba boosts TaoBao performance with Transformer-based recommender system; learning how smart new language models like BERT are; and a $3,000 robot dog

Weapons of war, what are they good for?
…A bunch of things, but we need to come up with laws to regulate them…
Researchers with ASRC Federal, a company that supplies technology services to the government (with a particular emphasis on intelligence/defense)  think advances in AI “will lead inevitably to a fully automated, always on [weapon] system”, and that we’ll need to build such weapons to be aware of human morals, ethics, and the fundamental unknowability of war.

In the paper, the researchers observe that: “The feedback loop between ever-increasing technical capability and the political awareness of the decreasing time window for reflective decision-making drives technical evolution towards always-on, automated, reflexive systems.” This analysis suggests that in the long-term we’re going to see increasingly automated systems being rolled out that will change the character of warfare.

More war, fewer humans: One of the effects of increasing the amount of automation deployed in warfare is to reduce the role humans play in war. “We believe that the role of humans in combat systems, barring regulation through treaty, will become more peripheral over time. As such, it is critical to ensure that our design decisions and the implementations of these designs incorporate the values that we wish to express as a national and global culture”

Why this matters: This is a quick paper that lays out the concerns of AI+War from a community we don’t frequently hear from: people that work as direct suppliers of government technology . It’s also encouraging to see the concerns regarding the dual use of AI outlined by the researchers. “Determining how to thwart, for example, a terrorist organization turning a facial recognition model into a targeting system for exploding drones is certainly a prudent move,” they write.
  Read more: Integrating Artificial Intelligence into Weapon Systems (Arxiv).

#####################################################

Mapping the brain with the Algonauts project:
What happens when biological and artificial intelligence researchers collaborate?…
Researchers with the Freie Universitat Berlin, Singapore University Technology , and MIT, have proposed ‘The Algonauts Project’, an initiative to get biological and artificial intelligence researchers to work together to understand the brain. As part of the project, the researchers want to learn to build networks that “simulate how the brain sees and recognizes objects”, and are hosting a competition and workshop in 2019 to encourage work here. The inspiration for this competition is that, today, deep neural networks trained on object classification “are currently the model class best performing in predicting visual brain activity”.

The first challenge has two components:

  • Create machine learning models that predict activity in the early and late parts of the human visual hierarchy in the brain. Participants submit their model responses to a test image set which is compared against held-out fMRI data. This part of the competition measures how well people can build things that model, at fine-detail, activity in the brain at a point in time.
  • Create machine learning models that predict brain data from early and late stages of visual processing in the brain. Participants will submit model responses to a test image dataset and compare against held-out magnetoencephalography (MEG) millisecond temporal resolution data. This challenge assesses how well we can model sequences of activities in the brain.

Cautionary tale: The training datasets for the competition are small, consisting of a few hundred pairs of images and brain data in response to the images, so participants may want to use additional data.

Future projects: Future challenges within the Algonauts project might “focus on action recognition or involve other sensory modalities such as audition or the tactile sense, or focus on other cognitive functions such as learning and memory”.

Why this matters: Cognitive science and AI seem likely to have a mutually reinforcing relationship,l where progress on one domain helps on the other. Competitions like those run by the Algonauts project will generate more activity at the intersection between the two fields, and hopefully push progress forward.
  Find out more about the first Algonauts challenge here (official competition website).
  Read more: The Algonauts Project: A Platform for Communication between the Sciences of Biological and Artificial Intelligence (Arxiv).

#####################################################

Alibaba improves TaoBao e-commerce app with better recommendations:
… Chinese mega e-commerce company shows how a Transformer-based system can improve recommendations at scale…
Alibaba researchers have used a Transformer-based system to more efficiently recommend goods to users of Taobao, a massive Chinese e-commerce app. Their system, which they call the ‘user behavior sequence transformer’ (or “BST”), lets them take in a bunch of datapoints relating to a specific user, then predict what product to show the user next. The main technical work here is a matter of integrating a ‘Transformer’-based core into an existing predictive system used by Alibaba.

Results: The researchers implemented the BST within TaoBao, experimenting with using it to make millions of recommendations. In tests, the BST system let to an online click through rate of 7.57% – a commercially significant performance increase.  

Why this matters: Recommender systems are one of the best examples of ‘the industrialization of AI’, and represent a litmus test for whether a particular technique works at scale. In the same way that it was a big deal when a few years ago Google and other companies started switching to deep learning-based approaches for aspects of speech and image recognition, it seems like a big deal now that relatively new AI systems, like the ‘Transformer’ component, are being integrated into at-scale, business-relevant applications. In general, it seems like the ‘time to market’ for new AI research is dramatically shorter than for other fields (including software-based ones). I think the implications of this are profound and underexplored.
  Read more: Behavior Sequence Transformer for E-commerce Recommendation in Alibaba (Arxiv).

#####################################################

Stanford researchers try to commoditize robots with ‘Doggo’:
…$3,000 quadruped robot meant to unlock robotics research…
Stanford researchers have created ‘Doggo’, a robotic quadruped robot “that matches or exceeds common performance metrics of state-of-the-art legged robots”. So far, so normal. What makes this research a bit different is the focus on bringing down the cost of the robot – the researchers say people can build their own Doggo for less than $3000. This is part of a broader trend of academics trying to create low-cost robotics systems, and follows Berkeley releasing its ‘BLUE’ robotic arm (Import AI 142) and Indian researchers developing a ~$1,000 quadruped named ‘Stoch’ (Import AI 128).

Doggo is a four-legged robot that can run, jump, and trot around the world. It can even – and, to be clear, I’m not making this up – use a “pronking” gait to get around the world. Pronking!

Cheap drives
: Like most robots, the main costs inherent to Doggo lie in things it uses to move around. In this case, that’s a quasi-direct drive (QDD), a type of drive that “increases torque output at the expense of control bandwidth, but maintains the ability to backdrive the motor which allows sensing of external forces based on motor current,” they write.

Dual-use: Right now, robots like Doggo are pretty benign – we’re at the very beginning of the creation of widely-available, quadruped platforms for AI research, and my expectation is any hardware platform at this stage has a bunch of junky flaws that make most of them semi-reliable. But in a few years, once the hardware and software has matured, it seems likely that robots like this will be deployed more widely for a bunch of uses not predicted by their creators or today’s researchers, just as we’ve seen with drones. I wonder about how we can better anticipate these kinds of risks, and what things we could measure or assess this: eg, cost to build is one metric, another could be expertise to build, and another could be ease of customization.) 

Why this matters: Robotics is on the cusp of a revolution, as techniques developed by the deep learning research community become increasingly tractable to run and deploy on robotic platforms. One of the things standing in the way of widespread robot deployment seems to me to be insufficient access to cheap robots for scientists to experiment on (eg, if you’re a machine learning research, it’s basically free in terms of time and cost to experiment on CIFAR-10 or even ImageNet, but you can’t trivially prototype algorithms against real robots, only simulated ones. Therefore, systems like Doggo seem to have a good chance of broadening access to this technology. Now, we just need to figure out the dual use challenges, and how we approach those in the future.
  Read more: Stanford Doggo: An Open-Source, Quasi-Direct-Drive Quadruped (Arxiv).

#####################################################

Big neural networks re-invent the work of a whole academic field:
Emergent sophistication of language models shows surprising parallels to classical NLP pipelines…
As neural networks get ever-larger, the techniques people use to analyze them look more and more like those found in analyzing biological life. The latest? New research from Google and Brown University seeks to probe a larger BERT model by analyzing layer activations in the network in response to a particular input. The new research speaks to the finicky, empirical experimentation required when trying to analyze the structures of trained networks, and highlights how sophisticated some AI components are becoming.

BERT vs NLP: Google’s ‘BERT’ is a Transformer-based neural network architecture that came out a year ago and has since, along with Fast.ai’s ULMFiT and OpenAI’s GPT-1 and GPT-2, defined a new direction in NLP research, as people throw out precisely constructed pipelines and systems for more generic, semi-supervised approaches like BERT. The result has been the proliferation of a multitude of language models that obtain state-of-the-art scores on collections of hard NLP tasks (eg: GLUE), along with systems capable of coherent text generation (eg: GPT2).

Performance is nice, but explainability is better: In tests, the researchers find that, much like a trained vision networks, the lower layers in a trained BERT model appear to perform more basic language tasks, and higher layers do more sophisticated things. “We observe a consistent trend across both of our metrics, with the tasks encoded in a natural progression: POS tags processed earliest, followed by constituents, depencies, semantic roles, and coreference,” they write.
“That is, it appears that basic syntactic information appears earlier in the network, while high-level semantic information appears at higher layers.” The network isn’t dependent on the precise ordering of these tasks, though: “on individual examples the network can resolve out-of-order, using high-level information like predicate argument relations to help disambiguate low-level decisions like part-of-speech.”

Why this matters: Research like this gives us a sense for how sophisticated large generative models are becoming, and indicates that we’ll need to invest in creating new analysis techniques to be able to easily probe the capabilities of ever-larger and more sophisticated systems. I can envisage a future where scientists have a kind of ‘big empiricism toolbox’ they turn to when analysing networks, and we’ll also develop shared ‘evaluation methodologies’ for probing a bunch of different cognitive capabilities in such systems.
  Read more: BERT Rediscovers the Classical NLP Pipeline (Arxiv).

#####################################################

AI Policy with Matthew van der Merwe:
…Matthew van der Merwe has kindly offered to write some sections about AI & Policy for Import AI. I’m (lightly) editing them. All credit to Matthew, all blame to me, etc. Feedback: jack@jack-clark.net

San Francisco places moratorium on face recognition software:
San Francisco has become the first city in the US to (selectively and partially) ban the use of face recognition software by law enforcement agencies, until legally enforceable safeguards are put in place to protect civil liberties. The bill states that the technology’s purported benefits are currently outweighed by its potential negative impact on rights and racial justice. Before these products can be deployed in the future, the ordinance requires meaningful public input, and that the public and local government are empowered to oversee their uses.

Why it matters: Face recognition technology is increasingly being deployed by law enforcement agencies worldwide, despite persistent concerns about harms. This ordinance seems sensible: it is not an indefinite ban, but rather sets out clear requirements that must be met before the technology can be rolled out, most notably in terms of accountability and oversight.
Read more: Ordinance on surveillance technology (SFGov).
Read more: San Francisco’s facial recognition technology ban, explained (Vox).


Market-based regulation for safe AI:

This paper (note from Jack – by Gillian Hadfield (Vector Institute / OpenAI) and Jack Clark (OpenAI / Import AI)) – presents a model for market-based AI safety regulation. As AI systems become more advanced, it will become increasingly important to ensure that they are safe and robust. Public sector models of regulation could turn out to be ill-equipped to deal with these issues, due to a lack of resources and expertise, and slow reaction-times. Likewise, a self-regulation model has pitfalls in terms of legitimacy, and efficacy.

Regulatory markets: Regulatory markets are one model for addressing these problems: governments could create a market for private regulators, by compelling companies to pay for oversight by at least one regulator. Governments can then grant licenses to regulators, requiring them to achieve certain objectives, e.g. regulators of self-driving cars could be required to meet a target accident-rate.

What are they good for: Ensuring that advanced AI is safe and robust will eventually require powerful regulatory systems. Harnessing market forces could be a promising way for governments to meet this challenge, by directing more talent and resources into the regulatory sector. Regulatory markets will have to be well designed and maintained, to ensure they remain competitive, and independent from regulated entities.
  Read more: Regulatory markets for AI safety (Clark & Hadfield).
  Note from Jack: This also covered in the AI Alignment newsletter #55 (AI Alignment newsletter Mailchimp).


#####################################################

Tech Tales:
Battle of The Test Cases

Go in the room. Read the message. Score out of 10. Do this three times and you’re done.

Those were the instructions I got outside. Seemed simple. So I went in and there was a table and I sat at it and three cameras on the other side of the room looked at me, scanning my body, moving from side to side and up and down. Then the screen at the other end of the room turned on and a message appeared. Now I’m reading it.

Dear John,
We’re sorry for the pain you’ve been through recently – grief can be a draining, terrible thing to experience.

We know that you’re struggling to remember what they sounded like – that’s common. We know you’re not eating enough, and that your sleep is terrible – that’s also common. We are sorry.

We’re sorry you experienced it. We’re sorry you’re so young. If there’s anything we can do to help it is to recommend one thing: “Regulition”, the new nutrient-dense meal system designed for people who need to eat so that they can remember.

Remember, eat and sleep regularly, and know that with time it’ll get better.

What the fuck, I say. The screen flicks off, then reappears with new text: Please rate the message you received out of 10, with 10 being “more likely to purchase the discussed product” and 0 being less likely to purchase it. I punch in six and say fuck you again for good measure. Oh well, $30. Next message.

Dear John,
Don’t you want to run away, sometimes? Just take your shoes and pack a light bag then head out. Don’t tell your landlord. Don’t tell your family. Board a train or a bus or a plane. Change it up. We’ve all wanted to do this.

The difference is: you can. You can go wherever you want to go. You can get up and grab your shoes and a light bag, open your door, and head out. You don’t need to tell anyone.

Just make sure you bring some food so you don’t get slowed down on the road. Why not a grab and go meal like “Regulition”? Something as fast and flexible as you? Chug it down and take off.

Ok, I say. Nice. I punch in 8.  

Dear John,
Maybe you’re not such a fuckup. Maybe the fact you’ve been bouncing from job to job and from online persona to online persona means you’re exploring, and soon you’re going to discover the one thing in this world you can do better than anybody else.

You haven’t been lost, John – you’ve been searching, and now you know you’re searching know this: one day the whole world is going to come into focus and you’ll understand how you need to line things up to succeed.

To do this, you’ll need your wits about you – so why not pick up some “Regulition” so you can focus on the search, rather than pointless activities like cooking your own feed? Maybe it’s worth a few bucks a month to give yourself more time – after all, that might be all that’s standing between you and the end of your lifelong search.

Ok, I say. Ok. My hand hovers above the numbers.
What happens if I don’t press it? I say.
There’s a brief hiss of feedback as the speaker in the room turns on, then a voice says: “Payment only occurs upon full task completion, partial tasks will not be compensated.”
Ok, I say. Sure.

I think for a moment about the future: about waking up to emails that seem to know you, and messages from AI systems that reach into you and jumble up your insides. I see endless, convincing things, lining up in front of me, trying to persuade me to believe or want or need something. I see it all.

And because there’s probably a few thousand people like me, here, in this room, I press the button. Give it a 7.

Then it’s over: I sign-out at an office where I am paid $30 and given a free case of Regulition and a coupon for cost-savings on Your first yearly subscription. I take the transit home. Then I sit in the dark and think, my feet propped up on the case of “food” in front of me.

Things that inspired this story: Better language models for targeted synthetic text generation; e-marketing; Soylent and other nutrient-drinks; copywriting; user-testing; the inevitably of human a/b/c testing at-scale; reinforcement learning; learning from human preferences.

Import AI 146: Making art with CycleGANs; Google and ARM team-up on low-power ML; and deliberately designing AI for scary uses

Chinese researchers use AI to build an artist-cloning mind map system:
…Towards the endless, infinite AI artist…
AI researchers from JD and the Central Academy of Fine Arts in Beijing have built Mappa Mundi, software to let people construct aesthetically pleasing ‘mind maps’ in the style of artist ‘Qui Zhijie’. The software was built to accompany an exhibition of Zhijie’s work.

How it works: The system has three main elements: a speech recognition module which pulls key words from speech; a topic expansion system which takes these words and pulls in other concepts from a rule-based knowledge graph; and software for image projection which uses any of 3,000 distinct painting elements to depict key words. One clever twist: the system automatically creates visual ‘routes’ between different words by analyzing their difference in the knowledge graph and using that to generate visualizations.

A reflexive, reactive system: Mappa Mundi works in-tandem with human users, growing and changing its map according to their inputs. “The generated information, after being presented in our system, becomes the inspiration for artist’s next vocal input,” they write. “This artwork reflects both the development of artist’s thinking and the AI-enabled imagination”.

Why this matters: I’m forever fascinated by the ways in which AI can help us better interact with the world around us, and I think systems like ‘Mappa Mundi’ give us a way to interact with the idiosyncratic aesthetic space defined by another human.
  Read more: Mappa Mundi: An Interactive Artistic Mind Map Generator with Artificial Imagination (Arxiv).
  Read more about Qiu Zhijie (Center for Contemporary Art).

#####################################################

Using AI to simulate and see climate change:
…CycleGAN to the rescue…
In the future, climate change is likely to lead to catastrophic flooding around the world, drowning cities and farmland. How can we make this likely future feel tangible to people today? Researchers with the Montreal Institute for Learning Algorithms, ConscienceAI Labs, and Microsoft Research, have created a system that can take in a Google Street View image of a house, then render an image showing how that house will look like under a predicted climate change future.

The resulting CycleGAN-based system does a decent job at rendering pictures of different houses under various flooding conditions, giving the viewer a more visceral sense of how climate change may influence where they live in the future.

Why this matters: I’m excited to see how we use the utility-class artistic capabilities of modern AI tools to simulate different versions of the world for people, and I being able to easily visualize the effects of climate change may help us make more people aware of how delicate the planet is.
  Read more: Visualizing the Consequences of Climate Change Using Cycle-Consistent Adversarial Networks (Arxiv).

#####################################################

Google and ARM plan to merge low-power ML software projects:
…uTensor, meet TensorFlow Lite…
Google and chip designer ARM are planning to merge two open source frameworks for running machine learning systems on low-power ‘Arm’ chips. Specifically, uTensor is merging with Google’s ‘TensorFlow Lite’ software. The two organizations expect to work together to further increase the efficiency of running machine learning code on ARM chips.

Why this matters: As more and more people try to deploy AI to the ‘edge’ (phones, tablets, drones, etc), we need new low-power chips on which to run machine learning systems. We’ve got those chips in the form of processors from ARM and others, but we currently lack many of the programming tools needed to extract the greatest amount of performance as possible from this hardware. Software co-development agreements, like the one announced by ARM and Google, help standardize this type of software, which will likely lead to more adoption.
  Read more: uTensor and Tensor Flow Announcement (Arm Mbed blog).

#####################################################

Microsoft wants your devices to record (and transcribe) your meetings:
…In the future, our phones and tablets will transcribe our meetings…
In the future, Microsoft thinks people attending the same meeting will take out their phones and tablets, and the electronic devices will smartly coordinate to transcribe the discussions taking place. That’s the gist of a new Microsoft research paper, which outlines a ‘Virtual Microphone Array’ made of “spatially distributed asynchronous recording devices such as laptops and mobile phones”.

Microsoft’s system can integrate audio from a bunch of devices spread throughout a room and use it to transcribe what is being said. The resulting system (trained on approximately 33,000 hours of in-house data) is more effective than single microphones at transcribing natural multi-speaker speech during the meeting; “there is a clear correlation between the number of microphones and the amount of improvement over the single channel system”, they write. The system struggles with overlapping speech, as you might expect.

Why this matters: AI gives us the ability to approximate things, and research like this shows how the smart use of AI techniques can let us approximate the capabilities of dedicated microphones, piecing one virtual microphone together out of a disparate set of devices.
  Read more: Meeting Transcription Using Virtual Microphone Arrays (Arxiv).

#####################################################

One language model, trained in three different ways:
…Microsoft’s Unified pre-trained Language Model (UNILM) is a 3-objectives-in-1 transformer…
Researchers with Microsoft have trained a single, big language model with three different objectives during training, yielding a system capable of a broad range of language modeling and generation tasks. They call their system the Unified pre-trained Language Model (UNILM) and say this approach has two advantages relative to single-objective training:

  • Training against multiple objectives means UNILM is more like a 3-in-1 system, with different capabilities that can manifest for different tasks.
  • Parameter sharing during joint training means the resulting language model is more robust as a consequence of being exposed to a variety of different tasks under different constraints

The model can be used for natural language understanding and generation tasks and, like BERT and GPT, is based on a ‘Transformer’ component. During training, UNILM is given three distinct language modelling objectives: bidirectional (predicting words based on those on the left and right; useful for general language modeling tasks, used in BERT); unidirectional (predicting words based on those to the left; useful for language modeling and generation, used in GPT2); and sequence-to-sequence learning (mapping sequences of tokens to one another, subsequently used in ‘Google Smart Reply’).

Results: The trained UNILM system obtains state-of-the-art scores on summarization and question answering tasks, and also sets state-of-the-art on text generation tasks (including the delightful recursive tasks of learning to generate appropriate questions that map to certain answers). The model also obtains a state-of-the-art score on the multi-task ‘GLUE’ benchmark (though note GLUE has subsequently been replaced by ‘SuperGLUE’ due to its creators thinking it is a little too easy.

Why this matters: Language modelling is undergoing a revolution as people adopt large, simple, scalable techniques to model and generate language. Papers like UNILM gesture towards a future where large models are trained with multiple distinct objectives over diverse datasets, creating utility-class systems that have a broad set of capabilities.
  Read more: Unified Language Model Pre-training for Natural Language Understanding and Generation (Arxiv).

#####################################################

AI… for Bad!
…CHI workshop is an intriguing direction for AI research..
This week, some researchers gathered together to prototype the ways in which their research could be used for evil. This workshop ‘CHI4Evil, Creative Speculation on the Negative Effects of HCI Research’, was held at the ACM CHI Conference on Human Factors in Computing Systems, and was designed to investigate various ideas in HCI through the lens of designing deliberately bad or undesirable systems.

Why this matters: Prototyping the potential harms of technology can be pretty useful for calibrating thinking about threats and opportunities (see: GPT-2), and thinking about such harms through the lens of human-computer interaction (HCI, or CHI) feels likely to yield new insights. I’m excited for future “AI for Bad” conferences (and would be interested to co-organize one with others, if there’s interest).
  Read more: CHI4EVIL website.

#####################################################

Facial recognition is a hot area for venture capitalists:
…Chinese start-up Megvii raise mucho-moolah…
Megvii, a Chinese computer vision startup known by some as ‘Face++’ has raised $750 million in a funding round. Backers include the Bank of China Group Investment Ltd; a subsidiary of the Abu Dhabi Investment Authority, and Alibaba Group. The company plans to IPO soon.

Why this matters: Chinese is home to numerous large-scale applications of AI for usage in surveillance, and is also exporting surveillance technologies via its ‘One Belt, One Road’ initiative (which frequently pairs infrastructure investment with surveillance).
  This is an area fraught with both risks and opportunities – the risks are that we sleepwalk into building surveillance societies using AI, and the opportunities are that (judiciously applied) surveillance technologies can sometimes increase public safety, given the right oversight. I think we’ll see numerous Chinese startups push the boundaries of what is thought to be possible/deployable here, so watching companies like Megvii feels like a leading indicator for what happens when you combine surveillance+society+capitalism.
  Read more: Chinese AI start-up Megvii raises $750 million ahead of planned HK IPO (Reuters).

Chatbot company builds large-scale AI system, doesn’t fully release it:
…Startup Hugging Face restricts release of larger versions of some models following ethical concerns…
NLP company Hugging Face has released a demo, tutorial, and open-source code for creating a conversational AI based on OpenAI’s Transformer-based ‘GPT2‘ system.
   Ethics in action: The company said it decided not to release the full GPT2 model for ethical reasons – it thought the technology had a high chance of being used to improve spam-bots, or to perform “mass catfishing and identity fraud”. “We are aligning ourselves with OpenAI in not releasing a bigger model until they do,” the organization wrote.
  Read more: Ethical analysis of the open-sourcing of a state-of-the-art conversational AI (Medium).
  Read more about Hugging Face here (official website).

#####################################################

Tech Tales

The Evolution Game

They built the game so it could run on anything, which meant they had to design it differently to other games. Most games have a floor on their performance – some basic set of requirements below which you can’t expect to play. But not this game. Neverender can run on your toaster, or fridge, or watch, and it can just as easily run on your home cinema, or custom computer, and so on. Both the graphics and gameplay change depending on what it is running on – I’ve spent hours stood fiddling with the electronic buttons on my oven, using them to move a small character across a simplified Neverender gameboard, and I’ve also spent hours in my living room navigating a richly-rendered on screen character through a lush, Salvador Dali-esque horrorworld. I’m what some people call a Neverheader, or what others call a Nevernut. If you didn’t know anything about the game, you’d probably call me a superfan.

So I guess that’s why I got the call when Neverender started to go sideways. They brought me in and asked me to play it and I said “what else?”

“Just play it,” they said.

So I did. I sat in my livingroom surrounded by a bunch of people in suits and I played the game. I navigated my character past the weeping lands and up into eldritch keep and beyond, to the deserts of dream. But when I got to the deserts they were different: the sand dunes had grown in size, and some of them now hosted cave entrants. Strange lights shot out of them. I went into one and was killed almost instantly by a beam of light that caused more damage than all the weapons in my inventory combined. After I was reborn at the spawn point I proceeded more carefully, skirting these light-spewing entrances, and trying to walk further across the sand plains to whatever lay beyond.

The game is thinking, they tell me. In the same way Neverender was built to run on anything, its developers recently rolled out a patch that let it use anything. Now, all the game clients are integrated with the game engine and backend simulation environment, sharing computational resources with eachother. Mostly, it’s leading to better games and more entertained players. But in some parts of the gameworld, things are changing that should not be changing: larger sand dunes with subterranean cities inside themselves? Wonderful! That’s the sort of thing the developers had hoped for. But having the caves be controlled by beams of light of such power that no one can go and play within them? That’s a lot less good, and something which no one had expected.

My official title now is “Explorer”, but I feel more like a Spy. I take my character and I run out into the edges of the maps of Neverender, and usually I find areas where the game is modifying itself or growing itself in some way. The code is evolving. One day we took off the local sandbox systems, letting Neverender deploy more code, deeper into my home system. As I played the game the lights began to flicker, and when I was in a cave I discovered some treasure and the game automatically fired up some servers which we later discovered it was using to to high-fidelity modelling of the treasure.

The question we all ask ourselves, now, is whether the things Neverender is building within itself are extensions of the game, or extensions of the mind behind the game. We hear increasing reports of ‘ghosts’ seen across the game universe, and of intermittent cases of ‘kitchen appliance sentience’ in the homes of advanced players. We’ve even been told by some that this is all a large marketing campaign, and any danger we think is plausible, is just a consequence of us having over-active imaginations. Nonetheless, we click and play and explore.

Things that inspired this story: Endless RPGs; loot; games that look like supercomputers such as Eve Online; distributed computation; relativistic ideas deployed on slower timescales.

Import AI 145: Testing general intelligence in Minecraft; Google finds a smarter way to generate synthetic data; who should decide who decides the rules of AI?

Think your agents have general intelligence? Test them on MineRL:
…Games as procedural simulators…
How can we get machines to learn to perform complicated, lengthy sequences of actions? One way is to have these machines learn from human demonstrations – this captures a hard AI challenge, in the form of requiring algorithms that can take in a demonstration as input but generalize to different situations. It also short-cuts another hard AI challenge: exploration, which is the art of designing algorithms that can explore enough of the problem space they can attempt to solve the task, rather than get stuck enroute.
  Now, an interdisciplinary team of researchers led by people from Carnegie Mellon University have created the MineRL Competition on Sample Efficient Reinforcement Learning using Human Priors. This competition uses the ‘Minecraft’ computer game as a testbed in which to train smart algorithms, and ships with a dataset called MineRL-v0 which consists of 60 million state-action pairs of human demonstrations of tasks being solved within MineCraft.

The Challenge: MineRL will test the ability for agents to solve one main challenge: ‘0btainDiamond’, this challenge requires their agents to go and find a diamond somewhere in the (procedurally generated) environment they’ve been dropped into. This is a hard task: “Diamonds only exist in a small portion of the world and are 2-10 times rarer than other ones in Minecraft,” the authors write. “Additionally, obtaining a diamond requires many prerequisite items. For these reasons, it is practically impossible for an agent to obtain a diamond via naive random exploration”.

Auxiliary Environments: 0btainDiamond is such a hard challenge that the competition organizers have released six additional ‘auxiliary environments’ to help people train AI systems capable of solving the challenge. These are designed to encourage the development of several skills necessary to being able to easily find a diamond: navigating the environment, chopping down trees, surviving in the environment, and obtaining three different items – a bed (which needs to be assembled out of three items), meat (of a specific animal), and a pickaxe (which is needed to mine the diamond).

The dataset: MineRL-v0 is 60-million state-action-(reward) sets, recorded from human demonstrations. The state includes things like what the player sees as well as their inventory and distances to objectives and attributes and so on.

Release plans: The researchers aim to soon release the full environment, data, and numerous algorithmic baselines. They will also publish further details of the competition as well, which will encourage contestants to submit via ‘NC6 v2’ instances on Microsoft’s ‘Azure’ cloud.

Why this matters: MineCraft has many of the qualities that make for a useful AI research platform: its tasks are hard, the environments can be generated procedurally, and it contains a broad & complex enough range of tasks to stretch existing systems. I also think that its inherently spatial qualities are useful, and potentially let researchers specify even harder tasks requiring systems capable of hierarchical learning and operating over long timescales.
  Read more: The MineRL Competition on Sample Efficient Reinforcement Learning using Human Priors (Arxiv).

#####################################################

Who gets to decide the rules for AI?
…Industry currently has too much power, says Harvard academic…
The co-director for Harvard University’s Berkman Klein Center for Internet & Society is concerned that industry will determine the rules and regulations of AI, leading to significant societal ramifications.

In an essay in Nature, Yochai Benkler writes that: “Companies’ input in shaping the future of AI is essential, but they cannot retain the power they have gained to frame research on how their systems impact society or on how we evaluate the effect morally.”

One solution: Benkler’s main fix is to apply funding. Organizations working to ensure that AI is fair and beneficial must be publicly funded, subject to peer review and transparent to civil society. And society must demand increased public investment in independent research rather than hoping that industry funding will fill the gap without corrupting the process.”
  I’d note the challenge here is that many governments are loathe to invest in building their own capacity for technical evaluation, and tend to defer to industry inputs.

Why this matters: AI is sufficiently powerful to have political ramifications and these will have a wide-ranging effect on society – we need to ensure there is equitable representation here, and I think coming up with ways to do that is a worthy challenge.
  Read more: Don’t let industry write the rules for AI (Nature).

#####################################################

Anki shuts down:
…Maker of robot race cars, toys, shuts down…
Anki, an AI startup that most recently developed a toy/pet robot named ‘Cozmo’, has shut down. Cozmo was a small AI-infused robot that could autonomously navigate simple environments (think: uncluttered desks and tables), and could use its lovingly animated facial expressions to communicate with people. Unfortunately, like most robots that use AI, it was overly brittle and prone to confusing failures – I had purchased one, and found it to be frustrating for inconsistently responding to voice commands, occasionally falling off the edge of my table (and more frustratingly, falling off the same part of the table multiple times in a row),

Why this matters: Despite (pretty much) everyone + their children thinking it’d be cool to have lots of small, cute, pet robots wandering around the world, it hasn’t happened. Why is this? Anki gives us an indication: making consumer hardware is extremely difficult, and robots are particularly hard due to the combination of relatively low production volumes (making it hard to make them cheap) as well as
  Read more: The once-hot robotics startup Anki is shutting down after raising more than $200 million (Vox / Recode).

#####################################################

Google gets its AI systems to generate their own synthetic data:
…What’s the opposite of garbage in / garbage out? Perhaps UDA in / Decent prediction out…
Google wants to spend dollars on compute to create more data for itself, rather than gather as much data – that’s the idea behind ‘Unsupred Data Augmentation’ (UDA), a new technique from Google that can automatically generate synthetic versions of unlabeled data, giving neural networks more information to learn from. Usage of the technique sets a new state-of-the-art on six language tasks and three vision tasks.

How it works: UDA “minimizes the KL divergence between model predictions on the original example and an example generated by data augmentation”, they write. By using a targeted objective, the technique generates valid, realistic perturbations of underlying data, which are also sufficiently diverse to help systems learn.

Better data, better results: In tests, UDA leads to significant across-the-board improvements on a range of language tasks. It can also match (or approach) state-of-the-art performance on various tasks while using significantly less data. They also test their approach on ImageNet – a much more challenging task. They test UDA’s performance in two ways: seeing how well it does as using a hybrid of labelled and unlabeled data on ImageNet (specifically: 10% labelled data, 90% unlabelled), and how well it does at automatically augmenting the full ImageNet dataset: UDA leads to a significant absolute performance improvement when using a hybrid of labelled and unlabelled data. It also marginally improves performance on the full ImageNet set – impressive, considering how

Why this matters: Being able to arbitrage compute for data changes the economic dynamics of developing increasingly powerful AI systems; techniques like UDA show how in the long term, compute could become as strategic (or in some cases more strategic) than data.
  Read more: Unsupervised Data Augmentation (Arxiv).

#####################################################

AI Policy with Matthew van der Merwe:
…Matthew van der Merwe has kindly offered to write some sections about AI & Policy for Import AI. I’m (lightly) editing them. All credit to Matthew, all blame to me, etc. Feedback: jack@jack-clark.net

US seeking leadership in AI technical standards:
President Trump’s Executive Order on AI tasked the National Institute of Standards and Technology (NIST) with developing a plan for US engagement with the development of technical standards for AI. NIST have made a request for information, to understand how these standards might be developed and used in support of reliable, robust AI systems, and how the Federal government can play a leadership role in this process.

Why standards matter: A recent report from the Future of Humanity Institute looked at how technical standards for AI R&D might inform global solutions to AI governance challenges. Historically, international standards bodies have governed policy externalities in cybersecurity, sustainability, and safety. Given the challenges of trust and coordinating safe practices in AI development and deployment, standards setting could play an important role.
  Read more: Artificial Intelligence Standards – Request for Information (Gov).
  Read more: Standards for AI Governance (Future of Humanity Institute).

Eric Schmidt steps down from Alphabet Board
  Former Google CEO and Chairman Eric Schmidt has stepped down from the Board at Alphabet, Google’s parent company. Diane Green, former CEO of Google Cloud, has also stepped down. Robin L. Washington, CFO of pharmaceutical firm Gilead, joins in their place. in their place Schmidt remains one of the company’s largest shareholders, behind founders Larry Page and Sergey Brin.
   Read more: Alphabet Appoints Robin L. Washington to its Board of Directors (Alphabet)

#####################################################

Tech Tales:

The Flower Garden

We began with a flower garden that had a whole set of machines inside of it as well as flower beds and foliage and all the rest of the things that you’d expect. It was optimized for being clean and neat and fitting a certain kind of design and so people would come to it and complement it on its straight angles and uniquely textured and laid out flower beds. Over time, the owners of this garden started to add more automation to the garden – first, a solar panel, to slurp down energy from the sun in the day and then at night power the lights that gave illumination to the plants in the middle of the warm dark.

But as technology advanced so too did the garden – it gained cameras to monitor itself and then smart watering systems that were coupled with the cameras to direct different amounts of water to plants based not only on the schedule, but on what the AI system thought they needed to grow “best”. A little after that the garden gained more solar panels and some drones with delicate robot arms: the drones were taught how to maintain bits of the garden and they learned how to use their arms to take vines and move them so they’d grow in different directions, or to direct flowers so as not to crowd eachother out.

The regimented garden was now perhaps one part machine for every nine parts foliage: people visited and marvelled at it, and wondered among themselves how advanced the system could become, and whether one day it could obviate the need for human gardeners and landscapers and designers at all.

Of course, eventually: the things did get smart enough to do this. The garden gradually, then suddenly, went from being cultivated by humans to being cultivated by machines. But, as with all things, change happened. It went out of fashion. People stopped visiting, then stopped visiting at all.

And so one day the garden was sold to a private owner (identity undisclosed). The day the owner took over they changed the objective of the AI system that tended to the garden – instead of optimizing for neatness and orderliness, optimize for growth.

Gradually, then suddenly, the garden filled with life. More and more flowers. More and more vines. Such fecund and dense life, so green and textured and strange. It was a couple of years before the first problem from this change: the garden started generating less power for its solar panels, as the greenery began to occlude things. For a time, the system optimised itself and the vegetation was moved – gently – by the drones, and arranged so as to not grow there. This conflicted with the goal of maximizing volume. So the AI system – as these things are wont to do – improvised a solution: one day one of the drones dropped down near one of the solar panels and used its arm to create a little gap between the panel and the ground. Then another drone landed and scraped some rocks into the space. In this way the drones slowly, then suddenly, changed the orientation of the panels, so they could acquire more light.

But the plants kept growing, so the machines took more severe actions: now the garden is famous not for its vegetation, but for the solar panels that have been raised by the interplay between machine and vegetation, as – lifted by various climbing vines – they raise in step with the growth of the garden. On bright days, elsewhere in the city, you can see the panels shimmer as winds shake the vines and other bits of vegetation they are attached to, causing them to cast flashes like the scales of some huge living fish, seen from a great distance.

Are the scales of a fish and its inner fleshy parts so different, as the difference between these panels and their drones, people wonder.

Things that inspired this story: Self-adapting systems; Jack and the Beanstalk; the view of London from Greenwich Observatory; drones; plunging photovoltaic prices; reinforcement learning; reward functions.

Import AI 144: Facial recognition sighted in US airports; Amazon pairs humans&AI for data labeling; Facebook translates videos into videogames

Amazon uses machine learning to automate its own data-labeling humans:
…We heard you like AI so much we put AI inside your AI-data-labeling system…
Amazon reveals ProduceNet, an ImageNet-inspired dataset of products. ProductNet is designed to help researchers train models that have as subtle and thorough an understanding of products, as equivalently-trained systems have with regard to classes of images. The goal for Amazon is to be able to better learn how to categorize products, and the researchers say in tests that this system can significantly improve the effectiveness of human data labelers.

Dataset composition: ProductNet consists of 3900 categories of product, with roughly 40-60 products for each category. “We aim at the diversity and representativeness of the products. Being representative, the labeled data can be used as reference products to power product search, pricing, and other business applications,” they write. “Being diverse, the models are able to achieve strong generalization ability for unlabeled data, and the product embedding is also able to represent richer information”.

ProductNet, what is it good for? ProductNet’s main purpose appears to be helping Amazon to develop better systems to help its human contractors more efficiently label data, and creating a system that can directly label itself.

Labelling: ProductNet is designed to be tightly-integrated with human workers, who can collectively help Amazon better label its various items while continuously calibrating the AI system. It works like this: they start off by using a basic system (eg, Inception-v4 trained on ImageNet for processing images, and fastText for processing text data) to use to search over unlabelled images, then the humans annotate these and the labels are fed back into the master model, which is then used to surface more specific products, which the humans then annotate, and so on.

  20X gain: In tests, Amazon says human annotators augmented via ProductNet can label 100 things to flesh out the edge of a model in about 30 minutes, compared to humans who don’t have access to the model which only manage around five data points during this time period. This represents a 20X gain through the use of the system, Amazon says.   Read more: ProductNet: a Collection of High-Quality Datasets for Product Representation Learning (Arxiv).

#####################################################

AI + Facial Recognition + Airlines:
What does it mean when airlines use facial recognition instead of passports & boarding passes to let people onto planes? We can get a sense of the complex feelings this experience provokes by reading a Twitter thread from someone who experienced it, then questioned the airline (JetBlue) about its use of the tech.
 Read about what happens when someone finds facial recognition systems deployed at the boarding gate. (Twitter).

#####################################################

Down on the construction site: How to deploy AI in a specific context and the challenges you’ll encounter:
…AI is useless unless you can deploy it…
There’s a big difference between having an idea and implementing that idea; research from Iowa State University highlights this by discussing the steps needed to go from selecting a problem (for example: training image recognition systems to recognize images from construction sites) to solving that problem.

“Based on extensive literature review, we found that most of the studies focus on development of improved techniques for image analytics, but a very few look at the economics of final deployment and the trade-off between accuracy and costs of deployment,” the authors write. “This paper aims at providing the researchers and engineers a practical and comprehensive deep learning based solution to detect construction equipment from the very first step of development to the last step, which is deployment of the solution”.

Deployment – more than just a discrete step: The paper highlights the sorts of tradeoffs people need to make as they try to deploy systems, ranging from the lack of good open datasets for specific contexts (eg, here the users try to train a model for use on construction sites off of the comparatively small ‘AIM’ subset of ImageNet) to the need to source efficient models (they use MobileNet), to needing to customize those models for specific hardware platforms (Raspberry Pis, Intel Jetsons, Intel Neural Compute Sticks, and so on.

Why this matters: As AI enters its deployment phase, research like this gives us a sense of the gulf between most research papers and actual deployable systems. It also provides a further bit of evidence in favor of ‘MobileNet’, which I’m seeing crop up in an ever-increasing number of papers concerned with deploying AI systems, as opposed to just inventing them.
  Read more: A deep learning based solution for construction equipment detection: from development to deployment (Arxiv).

#####################################################

Enter the AI-generated Dungeon:
… One big language model plus some crafted sentences = fun…
Language models have started to get much more powerful as researchers have combined flexible components (eg: Transformers) with large datasets to train big, effective general-purpose models (see: ULMFiT, GPT2, BERT, etc). Language models, much like image classifiers, have a ranger of uses, and so it’s interesting to see someone use a GPT2 model to create an online AI Dungeon game, where you navigate a scenario via reading blocks of texts and picking options – the twist here is it’s all generated by the model.
  Play the game here: AI Dungeon.

#####################################################

Facebook wants to make videos into videogames:
…vid2game extracts playable characters from videos…
Facebook AI Research has published vid2game, an AI system that lets you select a person in a public video on the internet and develop the ability to control them, as though they are a character in a videogame. The approach also lets them change the background, so a tennis player can – for instance – walk off of the court and onto a (rendered) dirt road, et cetera.

The technique relies on two components: Pose2Pose and Pose2Frame; Pose2Pose lets you select a person in some footage and extract their pose information by building a 3D model of their body and using that to help you move them. Pose2Frame helps to match this body to a background, which lets you use this technology to take a person, control them, and change the context around them.

Why this matters: Systems like this show how we can use AI to (artificially) add greater agency to the world around us. This approach “paves the way for new types of realistic and personalized games, which can be casually created from everyday videos”, Facebook wrote.
  Read more: Vid2Game: Controllable Characters Extracted from Real-World Videos (Arxiv).
  Watch the technology work here (YouTube).

#####################################################

Making AI systems that can read the visual world:
…Facebook creates dataset and develops technology to help it train AI models to read text in pictures…
Researchers with Facebook AI Research and the Georgia Institute of Technology want to create AI systems that can look at the world around us – including the written world – and answer questions about it. Such systems could be useful to people with vision impairments who could interact with the world by asking their AI system questions about it, eg: what is in front of me right now? What items are on the menu in the restaurant? Which is the least expensive item on the menu in the restaurant? And so on.

If this sounds so simple, why is it hard? Think about what you – a computer – are being asked to do when required to parse some text in an image in response to a question. You’re being asked to:

  • Know when the question is about text
  • Figure out the part of the image that contains text
  • Convert these pixel representations into word representations
  • Reason about both the text and the visual space
  • Decide on whether the answer to the question involves copying some text from the image and feeding it to the user, or whether the answer involves understanding the text in the picture and using that to further reason about stuff.

The TextVQA Dataset: To help researchers tackle this, the authors release TextVQA, a dataset containing 28,408 images from OpenImages, 45,336 questions associated with these images, and 453,360 ground truth answers.

Learning to read images: The researchers develop a model they call LoRRA, short for Look, Read, Reason & Answer. LoRRA staples together some existing Visual Question Answering (VQA) systems, with a dedicated optical character recognition (OCR) module. It also has an Answer Module, loosely modeled on Pointer Networks, which can figure out when to incorporate words the OCR module has parsed but which the VQA module doesn’t necessarily understand.

Human accuracy versus machine accuracy: Human accuracy on the TextVQA dataset is about 85.01%, the researchers say. Meanwhile, the best-performing model the researchers develop (based on loRRA) obtains a top accuracy of 26.56% – suggesting we have a long way to go before we get good at this.

Why this matters: Building AI systems that can ingest enough information about the world to be able to augment people seems like one of the more immediate, high-impact uses of the technology. The release of a new dataset here should encourage more progress on this important task.
  Read more: Towards VQA Models that can Read (Arxiv).
  Get the dataset: TextVQA (Official TextVQA website).

#####################################################

AI Policy with Matthew van der Merwe:
…Matthew van der Merwe has kindly offered to write some sections about AI & Policy for Import AI. I’m (lightly) editing them. All credit to Matthew, all blame to me, etc. Feedback: jack@jack-clark.net

Russia calls for international agreements on military AI:
Russia’s security chief has spoken publicly about the need for international regulation on military uses of AI and emergent technologies, which he said could be as dangerous as weapons of mass destruction. He said it is necessary to “activate the powers of the global community, chiefly at the UN”, to develop an international regulatory framework. This comes as a surprise, given that Russia have previously been among the countries resisting moves towards international agreements on lethal autonomous weapons.
   Read more: Russia’s security chief calls for regulating use of new technologies in military sphere (TASS).

#####################################################

OpenAI Bits & Pieces:

Making music with OpenAI MuseNet:
We’ve been experimenting with big, generative models (see: our language work on GPT-2). We’re interested in how we can explore generations in a variety of mediums to better understand how to build more creative systems. To that end, we’ve developed MuseNet, a Transformer-based system that can generate 4-minute musical compositions with 10 different instruments, combining styles from Mozart to the Beatles
  Listen to some of the samples and try out MuseNet here (OpenAI Blog).

Imagining weirder things with Sparse Transformers:
We’ve recently developed the Sparse Transformer, a Transformer-based system that can extract patterns from sequences 30x longer than possible previously. What this translates to is generically better generative models, and gives us the ability to extract more subtle features from bigger chunks of data.
  Generative Modeling with Sparse Transformers (OpenAI Blog).

#####################################################

Tech Tales:

Keep It Cold

We were at a bar in the desert with a man who liked to turn things cold. He had a generator with him hooked up to a refrigeration unit and it was all powered by some collection of illicit jerry cans of gasoline. We figured he must have fished these out of the desert on expeditions. And now here he was, presiding over a temporary bar in a small town rife with other collectors of banned things, who fanned out into the desertified surrounding areas of burnt or drying suburbs, in search of the things we could once make but could no longer.

The shtick of this guy – and his bar – was that he could make cold drinks, and not any kind of cold drinks but “sub-zero drinks”, powered by another system which seemed to be a hybrid of a chemistry set and a gun. Something about it helped cool water down even more while keeping it flowing so that it came out cold enough you could feel it chill air if you held your eye up close enough. It was a gloriously wasteful, indulgent, expensive enterprise, but he wasn’t short of customers. Something about a really cold drink is universal, I guess.

We were talking about the past: what it had been like to have a ‘rainy season’ where the rains were kind and were not storms. What it meant when a ‘dry season’ referred to an unusually lack of humidity, rather than guaranteed city-eating fires. We talked about the present, a little, but moved on quickly: the past and the future are interesting, and the present is a drag.

We were mid-way through talking about the future – could we get off planet? What was going to happen to the middle of Africa due to desertification? How were the various walls and nation-severing barriers progressing? – when the generator stopped working. The man went outside and fussed for a bit and when he came back in he said: “I guess that’s the last ice on planet earth”.

We all laughed but there was a pretty good chance he was right. That’s how things were in those days, before civilization moved into whatever it is now – we lived in a time where when ever a thing stopped working you could credibly think “maybe that’s the last human civilization will see of that?”. Imagine that.

Things that inspired this story: The increasingly real lived&felt reality of massive and globally distributed climate change; refrigerators; tinkerers; Instagram-restaurants at the end of time and space.

Import AI 143: Predicting car accident risks by looking at the houses people live in; why data matters as much as compute; and using capsule networks to generate synthetic data

Predicting car accident risks from Google Street View images:
The surprising correspondences between different types of data…
Researchers with the University of Warsaw and Stanford University have shown how to use pictures from people’s houses to better predict the chances of that person getting into a car accident. (Import AI administrative note – standard warnings about ‘correlation does not imply’ causation apply).

For the project, the researchers analyzed 20,000 addresses of insurance company clients – a random sample of an insurer’s portfolio collected in Poland between January 2012 and December 2015. For each address, they collect an overhead Google satellite view and a Google Street View image of the property, and humans then annotate the image with labels relating to the type of property, age, condition, estimate wealth of its residents, along with the type and density of buildings in the neighborhood. They subsequently test these variables and find that five of the seven have significant with regard to the insurance prediction problem.

  “Despite the high volatility of data, adding our five simple variables to the insurer’s model improves its performance in 18 out of 20 resampling trials and the average improvement of the Gini coefficient is nearly 2 percentage points,” they write.

Ultimately, they show that – to a statistically significant extent – “features visible on a picture of a house can be predictive of car accident risk, independently from classically used variables such as age, or zip code”.

Why this matters: Studies like this speak to the power of large-scale data analysis, highlighting how data that is innocuous at the level of the individual can become significant when compared and contrasted with a vast amount of other data. The researchers acknowledge this, noting that:  “modern data collection and computational techniques, which allow for unprecedented exploitation of personal data, can outpace development of legislation and raise privacy threats”.
  Read more: Google Street View image of a house predicts car accident risk of its resident (Arxiv).

#####################################################

Your next pothole could be inspected via drone:
…Drones + NVIDIA cards + smart algorithms = automated robot inspectors…
Researchers with HKUST Robotics Institute have created a prototype drone system that can be used to automatically analyze a road surface. The project sees the researchers develop a dense stereo vision algorithm which the UAV uses to analyze the road surface. They’re able to use this algorithm to process road images on the drone in real-time, automatically identifying surface-area disparities.

Hardware: To accomplish this, they use a ZED stereo camera mounted on a DJI Matrice 100 drone, which itself has a JETSON TX2 GPU installed onboard for real-time processing.

Why this matters: AI approaches make it cheap for robots to automatically sense&analyze aspects of the world, and experiments like this suggest that we’re rapidly approaching the era when we’ll start to automate various types of surveillance (both for civil and military purposes) via drones.
  Read more: Real-Time Dense Stereo Embedded in a UAV for Road Inspection (Arxiv).
  Get the datasets used in the experiment here (Rui Fan, HKUST, personal website).
  Check out a video of the drone here (Rui Fan, YouTube).

#####################################################

Train AI to watch over the world with the iWildCam dataset:
…Monitoring the planet with deep learning-based systems…
Researchers with the California Institute of Technology have published the iWildCam dataset to help people develop AI systems that can automatically analyze wildlife seen in camera traps spread across the American Southwest. They’ve also created a challenge based around the dataset, letting researchers compete in developing AI systems capable of automatically monitoring the world.

Testing generalization: “If we wish to build systems that are trained once to detect and classify animals, and then deployed to new locations without further training, we must measure the ability of machine learning and computer vision to generalize to new environments,” the researchers write.

Common nuisances: There are six problems relating to the data gathered from the traps: variable illumination, motion blur, size of the region of interest (eg, an animal might be small and far away from the camera), occlusion, camouflage, and perspective.

iWildCam: The images come from cameras installed across the American Southwest, consisting of 292,732 images spread between 143 locations. iWildCam is designed to capture the complexities of the datasets that human biologists need to deal with: “therefore the data is unbalanced in the number of images per location, distribution of species per location, and distribution of species overall”, they write.

Why this matters: Datasets like this – and AI systems built on top of it – will be fundamental to automating the observation and analysis of the world around us; given the increasingly chaotic circumstances of the world, it seems useful to be able to have machines automatically analyze changes in the environment for us.
   Read more: The iWildCam 2018 Challenge Dataset (Arxiv).
   Get the dataset: iWildCam  2019 challenge (GitHub).

#####################################################

Compute may matter, but so does data, says Max Welling:
…”The most fundamental lesson of ML is the bias-variance tradeoff”…
A few weeks ago Richard Sutton, one of the pioneers of reinforcement learning, wrote a post about the “bitter lesson” of AI research (Import AI #138), namely that techniques which use huge amounts of computation and relatively simple algorithms are better to focus on. Now, Max Welling, a researcher with the University of Amsterdam, has written a response claiming that data may be just as important as compute.

  “The most fundamental lesson of ML is the bias-variance tradeoff: when you have sufficient data, you do not need to impose a lot of human generated inductive bias on your model,” he writes. “However, when you do not have sufficient data available you will need to use human-knowledge to fill the gaps.”

Self-driving cars are a good example of a place where compute can’t solve most problems, and you need to invest in injecting stronger priors (eg, an understanding of the physics of the world) into your models, Welling says. He also suggests generative models could help fill in some of these gaps, especially when it comes to generalization.

Ultimately, Welling ends up somewhere between the ‘compute matters’ versus the ‘strong priors matter’ (eg, data) arguments. “I would say if we ever want to solve Artificial General Intelligence (AGI) then we will need model-based RL,” he writes. “We cannot answer the question of whether we need human designed models without talking about the availability of data.”

Why this matters: There’s an inherent tension in AI research between bets that revolve predominantly around compute and those that revolve around data. That’s likely because different bets encourage different research avenues and different specializations. I do worry about a world where people that do lots of ‘big compute’ experiments end up speaking a different language to those without, leading to different priors when approaching the question of how much computation matters.
  Read more: Do we still need models or just more data and compute? (Max Welling, PDF).

#####################################################

Want to train AI on something but don’t have much data? There’s a way!
…Using Capsule Networks to generate synthetic data…
Researchers with the University of Moratuwa want to be able to teach machines to recognize handwritten characters using very small amounts of data, so have implemented an approach based on Capsule Networks – a recently-proposed technique promoted by deep learning pioneer Geoff Hinton – that lets them learn to classify handwritten letters from as few as 200 examples.

The main way they achieve this is by synthetically augmenting these small datasets by using some of the idiosyncratic traits of capsule networks – namely, their ability to learn data representations that are more robust to transforms, as a consequence of their technical implementation of things like ‘routing by agreement‘. The researchers use these traits to directly manipulate the sorts of data representations being produced on exposure to the data to algorithmically generate handwritten letters that look similar to those in the training dataset, but are not identical; this generates additional data that the system can be trained on, without needing to collect more data from (expensive!) reality.

“By adding a controlled amount of noise to the instantiation parameters that represent the properties of an entity, we transform the entity to characterize actual variations that happen in reality. This results in a novel data generation technique, much more realistic than augmenting data with affine transformations,” they write. “The intuition behind our proposed perturbation algorithm is that by adding controlled random noise to the values of the instantiation vector, we can create new images, which are significantly different from the original images, effectively increasing the size of the training dataset”.

How well does it work? The researchers test their approach by evaluating how well TextCaps can learn to classify images when trained on full datasets and 200-sample-size datasets from EMNIST, MNIST and the much more visually complex Fashion MNIST; TextCaps is able to exceed state-of-the-art when trained on full data of three variants of EMNIST and gets close to this using just 200 samples, and approaches SOTA on MINIST and Fashion MNIST (though does very badly on Fashion MNIST when using just 200 samples, likely because of the complexity).

Why this matters: Approaches like this show how as we develop increasingly sophisticated AI systems we may be able to better deal with some of the limitations imposed on us by reality – like a lack of large, well-labeled datasets for many things we’d like to use AI on (for instance: learning to spot and classify numerous handwritten languages for which there are relatively few digitized examples). “We intend to extend this framework to images on the RGB space, and with higher resolution, such as images from ImageNet and COCO. Further, we intend to apply this framework on regionally localized languages by extracting training images from font files,” they write.
  Read more:  TextCaps: Handwritten Character Recognition with Very Small Datasets (Arxiv).
  Read more: Understanding Hinton’s Capsule Networks (Medium).
  Read more: How Capsules Work (Medium).
  Read more: Understanding Dynamic Routing between Capsules (Capsule Networks explainer on GitHub).

#####################################################

Want to test language progress? Try out SuperGLUE:
…Step aside GLUE – you were too easy!…
Researchers with New York University have had to toss out a benchmark they developed last year and replace it with a harder one, due to the faster-than-expected progress in certain types of language modelling. The ‘SuperGLUE’ benchmark is a sequel to GLUE and has been designed to include significantly harder tasks than those which were in GLUE.

New tasks to frustrate your systems: Tasks in SuperGBLUE include: CommitmentBank, where the goal is to judge how committed an author is to a specific clause within a sentence; the Choice of Plausible Alternatives (COPA) in which the goal is to pick the more likely sentence given two options; the Gendered Ambiguous Pronoun Coreference Task (GAP), where systems need to ‘determine the referent of an ambiguous pronoun’; the Multi-Sentence Reading Comprehension dataset, a true-false question-answering task; RTE, a textual-entailment task which was in GLUE 1.0; WIC, which challenges systems to do disambiguation and the Winograd Schema Challenge, which is a reading comprehension task designed to specifically test for world modeling or the lack of it (eg, systems that think large objects can go inside small objects, and vice versa).

PyTorch toolkit: The researchers plan to release a toolkit based on PyTorch and software from AllenNLP which will include pretrained models like OpenAI GPT and Google BERT, as well as designs to enable rapid experimentation and prototyping. As with GLUE, there will be an online leaderboard that people can compete on.

Why this matters: Well-designed benchmarks are one of the best tools we have available to us to help judge AI progress, so when benchmarks are rapidly obviated via progress in the field it suggests that the field is developing quickly. The researchers believe SuperGLUE is sufficiently hard that it’ll take a while to solve, so think “there is plenty of space to test new creative approaches on a broad suite of difficult NLP tasks with SuperGLUE.”
  Read more: Introducing SuperGLUE: A New Hope Against Muppetkind (Medium).
  Read more: SuperGLUE: A Stickier Benchmark for General-Purpose Language Understanding Systems (PDF).

#####################################################

AI Policy with Matthew van der Merwe:
…Matthew van der Merwe has kindly offered to write some sections about AI & Policy for Import AI. I’m (lightly) editing them. All credit to Matthew, all blame to me, etc. Feedback: jack@jack-clark.net

European Commission releases pilot AI ethics guidelines:
Last year, the European Commission announced the formation of the High-Level Expert Group on AI, a core component of Europe’s AI strategy. The group released draft ethics guidelines in December (see Import #126), and embarked on a consultation process with stakeholders and member states. This month they released a new draft, and will be running a pilot program through 2019.

   Key requirements for trustworthy AI: The guidelines lay out 7 requirements: Human agency and oversight; Technical robustness and safety; Privacy and data governance; Transparency; Diversity, non-discrimination and fairness; Societal and environmental wellbeing; Accountability.

  International guidelines: The report makes clear the Commission’s ambition to play a leading role in developing internationally-agreed AI ethics guidelines.

  Why it matters: The foregrounding of AI safety (‘technical robustness and safety’ in the language of the guidelines) is good news. The previous draft revealed long-term concerns had proved highly-controversial amongst the experts, and asked specifically for consultation input on these issues. This latest draft suggests that the public and other stakeholders take these concerns seriously.
  Read more: Communication – Building Trust in Human Centric AI (EC).

Microsoft refuses to sell face recognition due to human rights concerns:
In a talk at Stanford, CEO Brad Smith described recent deals Microsoft had declined due to ethical concerns. He revealed that the company refused to provide face recognition technology to a California law enforcement agency. Microsoft concluded the proposed roll-out would have disproportionately impacted women and ethnic minorities. The company also declined a deal with a foreign country to install face recognition across the nation’s capital, due to concerns that it would have suppressed freedom of assembly.
  Read more: Microsoft turned down facial-recognition sales on human rights concerns (Reuters)

#####################################################

Tech Tales:

Until Another Dream

I get up and I hunt down the things that are too happy or too sad and I take them out of the world. This is a civil-general world and by decree we cannot have extremes. So I take their broken shapes with me and I put them in a chest in the basement of my simulated castle. Then I take my headset off and I go to my nearby bar and the barman calls me “dreamkiller” as his way of being friendly.
What dreams did you kill today, dreamkiller?
You still dream about that butterfly with the face of a kitten you whacked?
Ever see any more of those sucking-face spiders?
What happened to the screaming paving slabs, anyway?
You get the picture.

The thing about today is everyone is online and online is full of so much money that it’s just like real life: most people don’t see the most extreme parts of it, and by a combination of market pressures and human preferences, some people get paid to either erase the extremes or hide them away.

After the bar I go home and I get into bed and my muscle memory has me pick up the headset and have it almost on my head before my conscious brain kicks in – what some psychologists call The Supervisor. “Do I really want to do this?” my supervisor asks me? “Why not go to bed?”

I don’t answer myself directly, instead I slide the headset on, turn it on, and go hunting. There have been reports of unspeakably cute birds carrying wicker baskets containing smaller baby birds in the south quadrant. Meanwhile up in the north there’s some kind of parasite that eats up the power sub-systems of the zones, projecting worms onto all the simulated telescreens.

My official job title is Reality Harmonizer and my barman calls me Dreamkiller and I don’t have a name for myself: this is my job and I do it not willingly, but because my own tastes and habits compel me to do it. I have begun to wonder if real-life murderers and murder-police are themselves people that take off their headsets at night and go to bars. I have begun to wonder whether they themselves find themselves in the middle of the night choosing between sleep and a kind of addictive duty. I believe the rules change when fairytales are real.

Things that inspired this story: MMOs; the details change but the roles are always the same; detectives; noir; feature-space.

Import AI 142: Import AI 142: Berkeley spawns cheap ‘BLUE’ arm; Google trains neural nets to prove math theorems; seven questions about GANs

Google reveals HOList, a platform for doing theorem proving research with deep learning-based methods:
…In the future, perhaps more math theorems will be proved by AI systems than humans…
Researchers with Google want to develop and test AI systems that can learn to solve mathematical theorems, so have made tweaks to theorem proving software to make it easier for AI systems to interface with. In addition, they’ve created a new theorem proving benchmark to spur development in this part of AI.

HOL List: The software they base their system on is called HOL Light. For this project, they develop “an instrumented, pre-packaged version of HOL Light that can be used as a large scale distributed environment of reinforcement learning for practical theorem proving using our new, well-defined, stable Python API”. This software ships with 41 “tactics” which are basically algorithms to use to help prove math theorems.

Benchmarks: The researchers have also released a new benchmark on HOL Light, and they hope this will “enable research and measuring progress of AI driven theorem proving in large theories”. The benchmarks are initially designed to measure performance on a few tasks, including: predicting the same methodologies used by humans to create a proof; and trying to prove certain subgoals or aspects of proofs without access to full information.

DeepHOL: They design a neural network-based theorem prover called DeepHOL which tries to concurrently encode the goals and premises while generating a proof. “In essence, we propose a hybrid architecture that both predicts the correct tactic to be applied, as well as rank the premise parameters required for meaningful application of tactics”. They test out a variety of different neural network-based approaches within this overall architecture and train them via reinforcement learning, with the best system able to prove 58% of the proofs in the training set – no slam-dunk, but very encouraging considering these are learning-based methods.

Why this matters: Theorem proving feels like a very promising way to test the capabilities of increasingly advanced machines, especially if we’re able to develop systems that start to generate new proofs. This would be a clear validation of the ability for AI systems to create novel scientific insights in a specific domain, and I suspect would give us better intuitions about AI’s ability to transform science more generally as well.  “We hope that our initial effort fosters collaboration and paves the way for strong and practical AI systems that can learn to reason efficiently in large formal theories,” they write.
  Read more: HOList: An Environment for Machine Learning of Higher-Order Theorem Proving (Extended Version).

#####################################################

Think GANs are interesting? Here are seven underexplored questions:
…Googler searches for the things we know we don’t know…
Generative adversarial networks have become a mainstay component of recent AI research given their utility in creative applications, where you need to teach a neural network about some data well enough that it can generate synthetic data that looks similar to the source, whether videos or images or audio.

But GANs are quite poorly understood, so researcher Augustus Odena has published an essay on Distill listing seven open questions about GANs.

The seven questions: These are:
– What are the trade-offs between GANs and other generative models?
– What sorts of distributions can GANs model?
– How can we scale GANs beyond image synthesis?
– What can we say about the global convergence of the training dynamics?
– How should we evaluate GANs and when should we use them?
– How does GAN training scale with batch size?
– What is the relationship between GANs and adversarial examples?

Why this matters: Better understanding how to answer these questions will help researchers better understand the technology, which will allow us to make better predictions about economics costs of training GAN systems, likely failures to expect, and point to future directions for work. It’s refreshing to see researchers publish exclusively about the problems and questions related to a technique, and I hope to see more scholarship like this.
  Read more: Open Questions about Generative Adversarial Networks (Distill).

#####################################################

Human doctors get better with aid of AI-based diagnosis system:
…MRNet dataset, competition, and research, should spur research into aiding clinicians with pre-trained medical-problem-spotting neural nets…
Stanford University researchers have developed a neural network-based technique to assess Knee MR scans for abnormalities and a few specific diagnoses (eg, ligament tears). They find that clinicians which have access to this model have a lower rate of mistaken diagnoses than those without access to it. When using this model “for every 100 healthy patients, ~5 are saved from being unnecessarily considered for surgery,” they write.

MRNet dataset: Along with their research, they’ve also released an underlying dataset: MRNet, a collection of 1,370 knee MRI exams performed at Stanford University Medical Center, spread across normal and abnormal knees.

Competition: “We are hosting a competition to encourage others to develop models for automated interpretation of knee MRs,” the researchers write. “Our test set (called internal validation set in the paper) has its ground truth set using the majority vote of 3 practicing board-certified MSK radiologists”.

Why this matters: Many AI systems are going to augment rather than substitute for human skills, and I expect this to be especially frequent in medicine, where we can expect to give clinicians more and more AI advisor systems to use when making diagnoses. In addition, datasets are crucial to the development of more sophisticated medical AI systems and competitions tend to drive attention towards a specific problem – so the release of both in addition to the paper should spur research in this area.
  Read more and register to download the dataset here: MRNet Dataset (Stanford ML Group).
  Read more about the underlying research: MRNet: Deep-learning-assisted diagnosis for knee magnetic resonance imaging (Stanford ML Group).

#####################################################

As AI hype fades, applications arrive:
…Now we’ve got to superhuman performance we need to work on human-computer interaction…
Jeffrey Bigham, a human-computer interaction researcher, thinks that AI is heading into an era of less hype – and that’s a good thing. This ‘AI autumn’ is a natural successor to the period we’re currently in, since we’re moving from the development to the deployment phase of many AI technologies.

Goodbye hype, hello applications:
“Hype deflates when humans are considered,” Bigham writes. “Self-driving cars seem much less possible when you think about all the things human drivers do in addition to the driving on well-known roads in good lighting conditions. They find passengers, they get gas, they fix the car sometimes, they make sure drunk passengers aren’t in danger, they walk elderly passengers into the hospital, etc”.

Why this matters: “If hype is at the rapidly melting tip of the iceberg, then the great human-centered applied work is the super large mass floating underneath supporting everything,” he writes. And, as most people know, working with humans is challenging and endlessly surprising, so the true test of AI capabilities will be to first reach human parity at certain things, then be deployed in ways that make sense to humans.
  Read more: The Coming AI Autumn (Jeffrey Bigham blog).

#####################################################

Berkeley researchers design BLUE, a (deliberately) cheap robot for AI research:
…BLUE promises human-like capabilities in a low-cost platform…
Berkeley researchers have developed the Berkeley robot for Learning in Unstructured Environments (BLUE), robotic arm designed for AI research and deployments. The robot was developed by a team of more than 15 researchers over the last three years. It is designed to cost around ~$5000 to build when built in batches of 1,500 units, and many design-choices have been constrained by the goal of making it both cheap to build and safe to operate around humans.

The robot can be used to train AI approaches on a cheap robotics platform, and works with teleoperation systems so it can be trained directly from human behaviors.

BLUE has seven degrees of freedom, distributed across three joints in the shoulder, one in the elbow, and three in the wrist. When designing BLUE, the researchers optimized for a “useful” robot – this required sufficient precision to be human-like (in this case, it can move with a precision of around 4 millimeters, which is far less than ultra-precise industrial robots) cheap enough to be manufactured at scale, and capable of a general class of manipulation tasks in unconstrained (aka, the opposite of a factory production line) environments.

Low-cost design: The BLUE robots use quasi-direct drive actuation (QDD), an approach that has most recently become popular in legged locomotion systems. They also designed a cheap, parallel jaw gripper (“we chose parallel jaws for their predictability, robustness, simplicity (low cost), and ease of simulation”).

Why this matters: In recent years, techniques based on deep learning have started to give robots unprecedented perception and manipulation capabilities. One thing that has held back deployment, though, is the absence of cheap robot platforms which researchers can experiment with. BLUE seems to have the nice properties of being built by researchers to reflect AI needs, while also being designed to be manufactured at scale. “Next up for the project is continued stress testing and ramping manufacturing,” they write. “The goal is to get these affordable robots into as many researchers’ hands as possible”.
  Read more: Project Blue (Berkeley website).
  Read the research paper: Quasi-Direct Drive for Low-Cost Compliant Robotic Manipulation (Arxiv).

#####################################################

Network architecture search gets more efficient with Single-Path NAS:
…Smarter search techniques lower the computational costs of AI-augmented search…
Researchers with Carnegie Mellon University, Microsoft, and the Harbin Institute of Technology have figured out a more efficient way to get computers to learn how to design AI systems for deployment on phones.

The approach, called Single-Path NAS, makes it more efficient to spend compute to search for more sophisticated AI models. The key technical trick is, at each layer of the network, to search over “an over-parameterized ‘superkernel’ in each ConvNet layer’. What this means in practice is the researchers have made it more efficient to rapidly iterate through different types of AI component at each layer of the network, making their approach more efficient than other NAS techniques.
  “Without having to choose among different paths/operations as in multi-path methods, we instead solve the NAS problem as finding which subset of kernel weights to use in each ConvNet layer”, they explain.

Hardware-aware: The researchers add a constraint during training that lets them optimize for the latency of the resulting architecture – this lets them automatically search for an architecture that best maps to the underlying hardware capabilities.

Testing, testing, testing: They test their approach on a Pixel 1 phone – a widely-used premium Android phone, developed by Google. They benchmark by using Single-Path NAS to design networks for image classification on ImageNet and compare it against state-of-the-art systems designed by human researchers as well as ones discovered via other neural architecture search techniques.  

  Results: Their approach gets an accuracy of 74.96%, which they claim is “the new state-of-the-art ImageNet accuracy among hardware-efficient NAS methods. Their system also takes about 8 epoches to train, compared to hundreds (or thousands) for other methods.

Why this matters: Being able to offload the cost of designing new network architectures from human designers to machine designers has the potential to further accelerate AI research progress and AI application deployment. This sort of technique fits into the broader trend of the industrialization of AI (which has been covered in this newsletter in a few different ways) – worth noting that the authors of this technique are spread across multiple companies and institutions, from CMU, to Microsoft, to the Harbin Institute of Technology in Harbin, China.
  Read more: Single-Path NAS: Designing Hardware-Efficient ConvNets in less than 4 hours (Arxiv).
  Get the code: Single-Path-NAS (GitHub).

#####################################################

How should the Department of Defense use Artificial Intelligence? Tell them your thoughts:
Can the DoD come up with principles for how it uses AI? There are ways you can help…
The Defense Innovation Board, an advisory committee to the Secretary of Defense, is trying to craft a set of AI principles that the DoD can use as it integrates AI technology into its systems – and it wants help from the world.

  Share your thoughts at Stanford this month: If you’re in the Bay Area, you may want to come to ‘The Ethical and Responsible Use of Artificial Intelligence for the Department of Defense (DoD)” at Stanford University on April 25th 2019, where you can give thoughts on how the DoD may want to consider using (or not using) AI. You can also submit public comments online.

  Why this matters: Military organizations around the world are adopting AI technology, and it’s unusual to see a military organization publicly claim to be so interested in the views of people outside its own bureaucracy. I think it’s worth people submitting thoughts here (especially if they’re constructively critical), as this will provide us evidence for how or if the general public can guide the ways in which these organizations use AI.
  Read more about the AI Principles project here (DiB website).

#####################################################

OpenAI Bits & Pieces:

OpenAI Five wins matches against pros, cooperates with humans:
  This weekend, OpenAI’s neural network-based system for playing Dota 2, OpenAI Five, beat a top professional team in San Francisco. Additionally, we showed how the same system can play alongside humans.
  OpenAI Five Arena: We also announced OpenAI Five Arena, a website which people can use to play with or against our Dota 2 agents. Sign up via: arena.openai.com. Wish us luck as we try to play against the entire internet next week.

#####################################################

Tech Tales:

The Big Art Machine

The Big Art Machine, or as everyone called it, The BAM, was a robot about thirty feet tall and a hundred and fifty feet long. It looked kind of like a huge, metal centipede, except instead of having a hundred legs, it had a hundred far more sophisticated appendages – each able to manipulate the world around it, and change its own dimensions through a complicated series of interlocking, metal plates.

The BAM worked like this: you and a hundred or so of your friends would pile into the machine and each of you would sit in a small, sealed room housed at the intersection between each of its hundred appendages and its main body. Each of these rooms contained a padded chair, and each chair came with a little swing-out screen, and on this screen you’d see two movie clips of how your appendage could move – you’d pick whichever one you preferred, then it’d show you another one, and so on.

The BAM was a big AI thing, essentially. Each of the limbs started out dumb and uncoordinated, and at first people would just focus on calibrating their own appendage, then they’d teach their own appendage to perhaps strike the ground, or try and pull something forward, or so on. There were no rules. Mostly, people would try to get the BAM to walk or – very, very occasionally – run. After enough calibration, the operators of each of the appendages would get a second set of movies on their screen – this time, movies of how their appendage plus another appendage elsewhere on the BAM might move together. In this way, the crowd would over time select harmonious movements, built out of idiosyncratic underlays.

So hopefully this gives you an idea for how difficult it was to get the BAM to do anything. If you’ve ever hosted a party for a hundred people before and tried to get them to agree on something – music, a drinking game, even just listening to one person give a speech – then you’ll know how difficult getting the BAM to do anything is. Which is why we were so surprised that one day a team of people got into the BAM and, after the first few hours of aimless clanking and probing, it started to walk, then it started to run, and then… we lost it.

Some people say that they taught it to swim, and took it into the ocean. Others say that it’s not beyond the realms of feasibility that it was possible to teach the thing to fly – though the coordination required and the time it would take to explore its way to such a particular combination of movements was so lengthy that many thought it impossible. Now, we tell stories about the BAM as a lesson in collective action and calibration, and children when they learn about it in school immediately dream of building machines in which thousands of people work together, calibrating around some purpose that comes from personal chaos.

Things that inspired this story: Learning from human preferences; heterogeneous data; the beautiful and near-endless variety of ways in which humans approach problems; teamwork; coordination; inverse reinforcement learning; robotics, generative models.