Last Week in AI #139: DeepMind acquires MuJoCo physics simulator and releases it for free, UK schools warned over facial recognition, Google lawyers slow down internal AI research

DeepMind acquired and made MuJoCo free w/ plans to open source, UK government warns schools to avoid facial-recognition payment systems, Google lawyers reportedly slow down research, and more!

Oct 25, 2021

Opening up a physics simulator for robotics

Last week, Google-backed AI research lab Deepmind announced that it has acquired the physics simulator MuJoCo (which stands for MuJoCo, which stands for Multi-Joint Dynamics with Contact). This simulator has long been relied upon by many robotics researchers to develop and test deep reinforcement learning algorithms, along with PyBullet and several other Physics simulators. Deepmind also announced that MuJoCo will now be freely available, with it having required a paid license in the past unlike alternatives to it, and also committed to making it an open-source and community-driven project. AI researchers generally regarded this as good news, with many welcoming MuJoCo becoming free and gaining such a powerful backer in Deepmind.

More on this - On the use of simulation in robotics: Opportunities, challenges, and suggestions for moving forward

Schools warned over facial recognition systems

Facial recognition is finding its way into more and more parts of our lives. In the UK, a few schools have started allowing students to make contactless payments with the technology. While the North Ayrshire council says it helped reduce Covid risks, a few actors pushed back. The Information Commissioner's Office said the schools should "carefully consider the necessity and proportionality of collecting biometric data before they do so" and consider less intrusive methods of letting students pay for meals. While a spokesman for the council said most people involved had consented to the use of facial recognition, it stands to reason that using such an obviously intrusive technology on children should be carefully thought through before being implemented. Early adopters of any technology for a new purpose will set a precedent in how they decide to use that technology. Schools should think through their role carefully.

Google's AI researchers say their output is being slowed by lawyers after a string of high-level exits: 'Getting published really is a nightmare right now.'

Some Google AI researchers have reported that internal research reviews by lawyers have slowed down the pace of paper publications. This follows the controversies at Google's Ethical AI team, where the exit of multiple high-profile researchers last year created a "chilling" effect in internal research teams. As a result, there appears to be increased scrutiny at the company, where even "uncontentious studies" are being brought to a standstill. However, a Google spokesperson states that the rate of internal paper approvals is on track with last year's numbers, so perhaps internal review effort only increased in specific research areas, not and not across the entire organization.

📚 Last Week in AI is growing!

With 2500+ subscribers, we want to bring you more, and ask for your help by becoming a paid subscriber.

While the newsletter will remain free and unchanged, paid subscribers will have access to the following perks:

Weekly op-eds on the latest developments in AI
Early access to news briefs
Submit AI topic requests for in-depth research and coverage

Other News

Research

Google AI Introduces ‘Uncertainty Baselines Library’ For Uncertainty and Robustness in Deep Learning - "Machine Learning has been a trending word in today’s technology. It is growingly used in a diverse range of real-world applications like Image and Speech recognition, Self-driving cars, Medical Diagnosis, to name a few."

DeepMind’s Fictitious Co-Play Trains RL Agents to Collaborate with Novel Humans Without Using Human Data - "Generating AI agents that can collaborate effectively with humans is a long-standing challenge in the broad field of human-machine interaction."

Breakthrough proof clears path for quantum AI - "Novel theorem demonstrates convolutional neural networks can always be trained on quantum computers, overcoming threat of ‘barren plateaus’ in optimization problems"

Yann LeCun Team Challenges Current Beliefs on Interpolation and Extrapolation Regarding DL Model Generalization Performance - "In the new paper Learning in High Dimension Always Amounts to Extrapolation, a team from Facebook AI Research and New York University that includes Yann LeCun theoretically and empirically challenges the machine learning (ML) community’s conventional wisdom regarding interpolation, demonstrating t"

Applications

The Age of Machine Learning As Code Has Arrived - "The 2021 edition of the State of AI Report came out last week. So did the Kaggle State of Machine Learning and Data Science Survey. There's much to be learned and discussed in these reports, and a couple of takeaways caught my attention."

Business

Inside Mailchimp’s push to bring AI to content marketing - "Earlier this month, Mailchimp released Content Optimizer, a new product that uses artificial intelligence to help improve the performance of email marketing campaigns. Thanks to its vast trove of data, Mailchimp is in a unique position to discover common patterns of successful marketing campaigns."

AI can see through you: CEOs' language under machine microscope - "Executives, beware! You could become your own worst enemy."

How Etsy Is Giving AI To Its Army Of 5 Million Artisan-Entrepreneurs To Build The Anti-Amazon - "Josh Silverman is using cutting-edge technology to transform Etsy from a hippie flea market into a Wall Street hero—without losing its soul."

SoftBank in talks to sell French robotics business to Germany’s United Robotics - "It is not clear whether SoftBank will retain a stake in the business, nor how much the deal would be worth. It is not clear whether SoftBank will retain a stake in the business, nor how much the deal would be worth."

This startup is creating personalized deepfakes for corporations - "Major companies in India are using Rephrase.ai to create avatars of celebrities and executives, and commercial use is coming soon."

Concerns

Q&A: Ghost Robotics CEO on Armed Robots for the U.S. Military - "Applications for the CAD software extend far beyond medicine and throughout the burgeoning field of synthetic biology, which involves redesigning organisms to give them new abilities."

Man Arrested for Unblurring Japanese Porn With AI in First Deepfake Case - "Japanese police on Monday arrested a 43-year-old man for using artificial intelligence to effectively unblur pixelated porn videos, in the first criminal case in the country involving the exploitative use of the powerful technology."

The AI oracle of Delphi uses the problems of Reddit to offer dubious moral advice - "Got a moral quandary you don’t know how to solve? Fancy making it worse? Why not turn to the wisdom of artificial intelligence, aka Ask Delphi: an intriguing research project from the Allen Institute for AI that offers answers to ethical dilemmas while demonstrating in wonderfully clear terms why"

Egypt detains artist robot Ai-Da before historic pyramid show - "Sculpture and its futuristic creator held for 10 days, possibly in fear she is part of spying plot"

Palantir's Peter Thiel thinks people should be concerned about surveillance AI - "Tech billionaire Peter Thiel believes that people should be more worried about "surveillance AI" rather than artificial general intelligences, which are hypothetical AI systems with superhuman abilities."

A Harvard freshman made a social networking app called 'The FaceTag.' It's sparked a debate about the ethics of facial recognition. - "Harvard freshman Yuen Ler Chow created an app in his dorm room that lets students sign up, scan the face of another user, and exchange contact information like phone numbers and Instagram handles. Right now, it's only available at Harvard. Chow calls it, "The FaceTag.""

Analysis

State of AI Report tracks transformers in critical infrastructure - "Artificial intelligence and machine learning pioneers are rapidly expanding on techniques that were originally designed for natural language processing and translation to other domains, including critical infrastructure and the genetic language of life."

Policy

Tesla fans cry foul as Biden administration moves toward Autopilot regulations - "The question of who will regulate cutting-edge driving technologies appears to be answered."

Vatican meeting explores challenge of artificial intelligence - "The symposium on Artificial Intelligence – or AI – organized by the Pontifical Council for Culture, in cooperation with the German Embassy to the Holy See, will open in Rome on Thursday."

Podcast

Check out our weekly podcast covering these stories!
Website | RSS | iTunes | Spotify | YouTube

Last Week in AI