Last Week in AI #141: AI understands whale speech, Facebook shuts down face recognition system, AI in US national security

Training language models to predict whale speech, Facebook will deactivate facial recognition features, and private lobbying for AI in US national security, and more!

Nov 8, 2021

Are We on the Verge of Chatting with Whales?

Project CETI (Cetacean Translation Initiative), conducted by an interdisciplinary team of biologists and computer scientists, aims to understand the language of whales, assuming there is one. Recent studies suggest that other mammals may also possess language in the same sense that humans do, namely, a mode of communication that has semantics (specific words meaning specific things), grammar (widely accepted rules to compose words into sentences), and is learned, not innate (like smiling or yawning). Sperm whales are known to communicate acoustically over long distances under water with sequences of "click" noises, which makes them especially amenable to analysis.

The key to Project CETI is the statistical language hypothesis, which supposes that the structure and meaning of a language can be extracted from the statistical relationships of its words. We've seen large language models like GPT-3 seemingly leverage this property to generate realistic texts after training on large amounts of language data. Project CETI aims to follow a similar approach by collecting a large dataset of 4 billion sperm whale "words" and trying to train such a language model, which may be able to "generate grammatically correct whale utterances." If this works, the next step is to translate whale codas into human words, something the researchers hope to do by analyzing the contextual information during speech.

Our take: the high-level direction of the project is worth a shot - even if the method does not find anything interesting, we will have collected a large dataset that can be used for future studies. However, expectations of human-whale communication should be low even if we can generate "sensible" whale sentences, as there is no reason to believe that statistical methods alone can ground the meaning of a totally non-human language.

Facebook, Citing Societal Concerns, Plans to Shut Down Facial Recognition System

In a surprise announcement, Facebook has stated it will no longer use facial recognition on its service, and will even delete billions of records that were used as part of it. With the feature -- meant to speed up tagging of people in images with facial recognition -- having been in place since 2010, this represents the shuttering of a technology Facebook has heavily invested in and was an industry leader in. Although the feature was only used to power detection of people in images uploaded to Facebook, the amount of data that was behind it still worried privacy advocates, and this move comes after Facebook has long been under scrutiny over how much data it collects and the ways in which it invades peoples’ privacy. Privacy advocates generally lauded this development.

Our take: this is a big deal, and a complete surprise. With Amazon, Microsoft and IBM limiting their facial recognition offerings over the past several years as well, there is hope that this technology can progress more slowly until regulation on the topic can be passed.

This Group Pushed More AI in US Security—and Boosted Big Tech

The US's tech giants may fight for market share in the cloud computing industry, but their executives have managed to team up for one thing: to advise the US government on using AI to bolster national security. As part of the National Security Commission on AI (NSCAI), the executives' recommendations--such as steering the Pentagon to work more closely with the tech industry--have already been written into law. Concerns about the group's financial incentives exist, and the companies did not directly respond to WIRED's questioning about whether industry participation in the commission created conflicts of interest.

While the group existed for three years and many of its recommendations were written into law, the results lean heavily towards deploying AI with "American values." Ben Winters, a lawyer who works on AI issues at the Electronic Privacy Information Center, said the suggestions "failed to recommend comprehensive privacy legislation or any concrete rights by people impacted by harmful AI." Following NSCAI's expiry, Eric Schmidt announced a private organization called the Special Competitive Studies Project that will research the impact of AI and emerging technologies in a range of areas.

Other News

Research

Recent Studies Find Ways To Demystify AI Black Boxes - "Deep learning neural networks, which are at the heart of modern artificial intelligence, are frequently characterized as “black boxes” with mysterious inner workings. However, recent research casts doubt on that notion, with profound privacy implications."

Why Facebook (Or Meta) Is Making Tactile Sensors for Robots - "Durable and affordable fingers and skin could help virtual agents understand their world"

Solving Math Word Problems - "OpenAI - We’ve trained a system that solves grade school math problems with nearly twice the accuracy of a fine-tuned GPT-3 model."

NIH awards nearly $75M to catalyze data science research in Africa - "New program will establish data science research and training network across the continent. The National Institutes of Health is investing about $74.5 million over five years to advance data science, catalyze innovation and spur health discoveries across Africa."

Scientists Claim to Improve Human Brain Function With Implants and AI - "In a study involving 12 patients who were undergoing brain surgery for epilepsy, the scientists say they were able to improve the patients’ mental function when stimulating areas of the brain with small amounts of electrical energy."

Google AI Introduces ‘GoEmotions’: An NLP Dataset for Fine-Grained Emotion Classification - "The emotions one experiences daily can motivate them to act and influence the significant and minor decisions they make in their lives. Therefore, they greatly influence how people socialize and form connections."

Researchers Present Global Effort to Develop Machine Learning Tools for Automated Assessment of Radiographic Damage in Rheumatoid Arthritis - "If a machine-learning approach could provide a quick, accurate quantitative score estimating the degree of joint damage in hands and feet, it would greatly help clinical research"

New Machine-Learning System Gives Robots Social Skills - "Robots can deliver food on a college campus and hit a hole-in-one on the golf course, but even the most sophisticated robot can’t perform basic social interactions that are critical to everyday human life."

Applications

Security AI is the next big thing - "In the world of cybersecurity, speed kills."

AI Is Keeping Watch Over Government Spending - "As the world turns increasingly more digital and data-driven, there is increasing desire for greater visibility and transparency of data. Governments around the world have turned to digital means to submit and pay taxes as well as collect a variety of revenue from different sources."

Autonomous Boats Seem More Solvable Than Autonomous Cars - "Technology was on duty miles downwind, where, as the battle against these fires went on for days, even weeks, many residents of the Bay Area and Sacramento River Delta region turned to air-quality sensor networks, particularly AirNow, maintained by the U.S."

DISA Moves to Combat Intensifying Cyber Threats with Artificial Intelligence - "In the near term, Defense Information Systems Agency officials plan to strategically employ artificial intelligence capabilities for defensive cyber operations. “First of all, the threat has never been higher."

NASA’s new rovers will be a fleet of mobile robots that work together - "NASA is exploring a concept for a new fleet of mini-rovers that can work together to solve problems and make decisions as a unit. If one fails or gets stuck somewhere, the others could carry on without it."

As the Arctic Warms, AI Forecasts Scope Out Shifting Sea Ice - "Global warming is making it harder to predict the movement and location of the ice cover, crucial information for fishing and global shipping."

The synthetic data that will help build AI and the metaverse - "Synthetic data — the generation of artificial images to train AI and computer vision — will be key to building out a future metaverse. Why it matters: AI has long been trained on images — including human faces — captured from the real world, but doing so can create serious privacy concerns."

AIBA uses AI technology to vet judges and refs at Belgrade worlds - "Boxing's governing AIBA claimed a win against corruption and bout manipulation on Friday after using military-approved artificial intelligence to screen judges and referees at the world championships in Belgrade"

Business

How AI is shaping Adobe’s product strategy - "Smarter image and video editing apps are the future"

Stanford engineers team up with Michelin-star chef to build modular restaurants - "Mezli’s prototype robot restaurant is making a minority of the bowls served to customers and is supplemented by a human-powered kitchen."

Zillow Quits Home-Flipping Business, Cites Inability to Forecast Prices - "Termination of ‘iBuying’ comes after company said it was halting new home purchases for rest of 2021"

Waymo is bringing its autonomous vehicles to New York City for (manually driven) mapping - "The latest autonomous vehicle company to tackle the mean streets of New York City will be Waymo. The Google spinoff is bringing a handful of cars to the Big Apple, where they will start mapping the city streets using manually driven vehicles with two people riding inside."

Microsoft is adding more AI smarts and other new tools to Office - "Microsoft's new Loop collaboration app and components isn't the only new thing coming to Office. At its Ignite IT pro conference this week, Microsoft officials also highlighted some other new tools and capabilities coming to Office."

Miso Introduces Second Generation Restaurant Kitchen Robot, the Flippy 2 - "Today Miso Robotics, a maker of restaurant robots, unveiled the second generation of its flagship robot Flippy, the Flippy 2."

Even as China Cracks Down on Tech, AI Companies Plan IPOs - "SenseTime and Megvii both include facial recognition technology among their offerings and do a lot of business with government agencies. "

Alphabet launches Isomorphic Labs, an AI-driven drug discovery startup - "Google parent company Alphabet announced a new venture Thursday that plans to accelerate the drug discovery process using advances in artificial intelligence."

New Azure OpenAI Service combines access to powerful GPT-3 language models with Azure’s enterprise capabilities - "Since OpenAI, an AI research and deployment company, introduced its groundbreaking GPT-3 natural language model platform last year, users have discovered countless things that these AI models can do with their powerful and comprehensive understanding of language."

Concerns

Australia Ordered Clearview AI to Destroy its Database, As Its Violating Privacy Laws - "After a couple of investigations, the Australian Information Commissioner’s Office declared that Clearview AI had violated the Australian privacy laws. Clearview AI is a US company aiding police personnel to identify people by scanning their faces in public or searching them in their repository."

Putin Wants Russia to Lead the Way in AI Weapons - "Even as numerous leaders in the tech world have repeatedly warned of the dangers that artificial intelligence- (AI) enabled weapons could present for the future of mankind, Russian Federation president Vladimir Putin has long been a firm supporter of AI weapons."

Google Allegedly After Military Deal That Puts AI Ethics Into Question - "Google is reportedly drafting a proposal for a defense contract that aims to employ artificial intelligence and cloud for military applications."

Analysis

GPT: Magic or Illusion? - "Hi and welcome to the fourth and final step of the Language Modeling Path a series of articles from Machine Learning Reply about the birth of the huge language models able to imitate and (let’s say) understand the human language, like BERT and GPT-3."

The future of AI is a conversation with a computer - "AI writing tools can tell us a lot about artificial intelligence"

The slippery slope of using AI and deepfakes to bring history to life - "Deepfakes could strengthen our love for history, but not without risk"

Study: China Far Outpacing U.S. in Worldwide Robotics Patents - "China has become the clear global leader in the total number of robotics patents issued, with nearly triple the number of robotics patents granted per year as the United States, according to a lengthy study published Tuesday by the Center for Security and Emerging Technology."

Policy

Sen. Peters' GOOD AI bill advances in Senate - "The Government Ownership and Oversight of Data in Artificial Intelligence (GOOD AI) Act has advanced from the committee markup stage in the Senate."

Henry Kissinger's Last Crusade: Stopping Dangerous AI - "At the age of 98, former Secretary of State Henry Kissinger has a whole new area of interest: artificial intelligence."

Expert Opinions

A.I. Is Not A-OK - "An interview with Eric Schmidt on the future of AI"

Sex Bots, Religion and the Wild World of A.I. - "The writer Jeanette Winterson explores the wild world unfolding alongside the rise of artificial intelligence."

How AI is hijacking art history - "But my lingering suspicion emerges from an awareness of how public support for the sciences and disparagement of the humanities means that, in the endeavor to gain funding and acceptance, the humanities will lose what makes them vital. "

AI is everywhere, but does it impact art? - "AI and art seem to represent two different worlds: technology and humanity. A closer look reveals they are intimately intertwined. Humans create technology, after all, so the two go hand in hand. AI follows a long tradition of technology reaching back to the cradle of civilization."

Humans bigger threat than AI, says chess legend Kasparov - "Russian chess legend Garry Kasparov was beaten by a supercomputer—but when it comes to artificial intelligence, he is firmly convinced that it's the humans who pose the real threat."

Explainers

How AI Creates Synthetic Speech - "Having machines turn text into speech is nothing new. Professor Stephen Hawking communicated with a computerized voice for many years, and by now, we're used to our GPS devices or smart speakers asking questions and responding to our queries."

Podcast

Check out our audio discussion of last week’s AI news stories!

Subscribe to the podcast here: Website | RSS | iTunes | Spotify | YouTube

Last Week in AI