Last Week in AI #185: PyTorch to be governed by Linux Foundation, FTC to regulate unfair AI use, Censorships in Chinese image-making AI, and more!

Meta moved PyTorch to independent governance body, FTC to regulate AI surveillance among other unfair uses, Chinese DALL-E censors Tiananmen Square and other politically sensitive content

Top News

Announcing the PyTorch Foundation: A new era for the cutting-edge AI framework

Meta has moved PyTorch to a new, independent PyTorch Foundation, under the Linux Foundation, whose governing board consists of representatives from a broad range of tech companies. The PyTorch Foundation will focus on “business and product marketing” and will not affect the technical development of the popular deep learning framework, used by more than 80% of researchers at top AI conferences. The new governance body will ensure the openness of the framework as well as its neutrality (e.g. not biasing development toward Meta’s internal needs), and it will “maintain a clear separation between the business and technical governance of PyTorch.”

Our take: This news is certainly welcomed by both AI researchers and practitioners, as a tool as important as PyTorch should not have its fate decided by a single tech giant. Lately, Meta has built an impressive track record of engaging with the broader AI community, with releases of large language models (OPT, BlenderBot3) and now this move of PyTorch to the Linux Foundation. Maybe these efforts will encourage other big tech companies to be more open and neutral with their AI advancements as well.

The FTC Is Closing In on Runaway AI

The Federal Trade Commission (FTC) voted 3-2 along party lines last month in favor of adopting an Advanced Notice of Proposed Rulemaking (ANPR) to address unfair and deceptive forms of data collection and the use of AI. Establishing ANPR is the first step toward regulating AI by the FTC. Some of these rules are directed toward ending the sales of data to third parties, regulating commercial surveillance and AI, extending the definition of discrimination to include teenagers, rural communities, homeless people, or people who speak English as a second language, restricting the time duration companies can retain consumer data, and audting automated decision-making systems to verify accuracy, reliability, and error rates. The FTC is inviting public comment about these issues and will continue to do so till the rules are finalized.

Our Take: The bias and potential misuse of deployed AI algorithms are well-known, and regulatory bodies across the world are slowly catching up. It is good that FTC is allowing public comments and soliciting inputs from many parties– academia, regulatory agencies, and companies. While several policies in the past have been implemented at the state level, federal policies will help standardize and streamline regulations and expand fair AI use across all states.

There’s no Tiananmen Square in the new Chinese image-making AI

Just as with GPT-3, China has followed the United States in releasing its own image generation models. ERNIE-ViLG, part of Baidu's large-scale NLP project Wenxin, is a 10 billion parameter model trained on 145 million image-text pairs that can generate images capturing China's cultural specificity. But there's a catch: ask it for "Tiananmen Square" or "Xi Jinping" and you'll be met with the words “The content entered doesn’t meet relevant rules. Please try again after adjusting it." This follows a January regulation of China banning AI-generated content that “endangers national security and social stability."

Our take: ERNIE-ViLG is likely going to be an important player in the image generation space--Technology Review observes that it has been embraced by people in the Japanese anime community and that its portrayals of Chinese historical figures, celebrities, and food outclass those of DALL-E 2 and Stable Diffusion. It may be, as Technology Review states, too useful to give up despite its explicit censorship. I do wonder, however, if ERNIE-ViLG will meet its own "Stable Diffusion"--who, if anyone, will train the open-source version of ERNIE-ViLG without censorship in place? What would the Chinese government do about it, if they are able to do anything?

Other News

Research

How Transformers Seem to Mimic Parts of the Brain - "Neural networks originally designed for language processing turn out to be great models of how our brains understand places."

NIH Injects $130M to Provide Quality Tools, Data for AI in Medicine - "As part of the National Institutes of Health’s launch of the Bridge to Artificial Intelligence program—or Bridge2AI—the agency is investing $130 million over four years in an effort to increase the use of artificial intelligence in biomedical and behavioral research."

Tiny, caterpillar-like soft robot folds, rolls, grabs and degrades - "When you hear the term "robot," you might think of complicated machinery working in factories or roving on other planets. But "millirobots" might change that. They're robots about as wide as a finger that someday could deliver drugs or perform minimally invasive surgery."

University of Washington graduates use artificial intelligence to create new proteins - "The findings show how machine learning can create protein molecules that are more accurate and made quicker than before. "

Results suggest dogs are more attuned to actions rather than to who or what is doing the action - "Scientists have decoded visual images from a dog's brain, offering a first look at how the canine mind reconstructs what it sees. The Journal of Visualized Experiments published the research done at Emory University."

Applications

AI could revamp the once-doomed Smell-O-Vision - "Computer scientists and chemical engineers at the Tokyo Institute of Technology in Japan have developed a machine-learning algorithm that can reverse-engineer a smell based on its chemical makeup."

Machine-learning prediction of hosts of novel coronaviruses requires caution as it may affect wildlife conservation - "A study published in Nature Communications used a similarity-based machine-learning pipeline approach to predict associations between coronaviruses and their potential mammalian hosts."

Steps to avoid overuse and misuse of machine learning in clinical research - "Machine learning algorithms are a powerful tool in healthcare, but sometimes perform no better than traditional statistical techniques. Steps should be taken to ensure that algorithms are not overused or misused, in order to provide genuine benefit for patients."

World’s First Bot-Generated Graphic Novel Hits the Market - "AI art hit the news this week for beating humans in a fine art competition. This has caused a storm of clashing opinions in the art world. And now, there is a new evolution to the skirmish. A graphic novel that features AI-generated art will go on sale next week."

Artist uses AI to extract color palettes from text descriptions - "A London-based artist named Matt DesLauriers has developed a tool to generate color palettes from any text prompt, allowing someone to type in "beautiful sunset" and get a series of colors that matches a typical sunset scene, for example."

How A.I. technologies could help resolve food insecurity - "When it comes to food aid, timing is vital. Availability is often thwarted when crises occur, like climate and economic shocks, or conflict and migration. Accurately forecasting helps relief arrive to the right areas on time."

YouTube Shorts is Making Ads Vertical, Uses Machine Learning to Adopt it from Landscape Videos - "YouTube is making machine learning work its technology to adopt vertical ads from landscape videos for it to apply to its latest video format with YouTube Shorts."

It’s not just floods and fires: This AI forecasts how climate change will impact your city - "Climate Alpha is a new platform that uses algorithms to determine how cities will transform in the face of climate change."

‘Predictive-Maintenance’ Tech Is Taking Off as Manufacturers Seek More Efficiency - "Pepsi, Colgate and other firms are populating their plants with sensors from AI startup Augury to ‘listen’ for machinery problems."

Midjourney Founder David Holz On The Impact Of AI On Art, Imagination And The Creative Economy - "Midjourney is one of the leading drivers of the emerging technology of using artificial intelligence (AI) to create visual imagery from text prompts."

Business

Cerebras Proposes AI Megacluster with Billions of AI Compute Cores - "Chipmaker Cerebras is patching its chips – already considered the world’s largest – to create what could be the largest-ever computing cluster for AI computing."

Northwell Health-backed JV invests $100M in AI companies tackling health inequities - "The joint venture will launch AI companies in 2022 with an initial focus on maternal health and chronic disease detection."

Artificial Intelligence Spending Grew 20.7% Worldwide in 2021, According to IDC - "Worldwide revenues for the artificial intelligence (AI) market, including software, hardware, and services for both AI centric and AI non-centric applications*, totaled $383.3 billion in 2021"

Inventor Behind Amazon’s Alexa Raises Funds for AI Startup - "The inventor of several key technologies used by Amazon.com Inc.’s Alexa service raised $20 million to fund a new startup in the UK."

GM's Cruise to Launch Robotaxis in Austin and Phoenix by End of 2022 - "Cruise will launch its autonomous taxi service in Phoenix and Austin before the end of 2022, according to Kyle Vogt, the CEO of driverless car company."

Upset by high prices, GM's Cruise develops its own chips for self-driving cars - "General Motors' autonomous driving unit Cruise has developed its own chips for self-driving cars to be deployed by 2025, as they aim to bring down costs and scale up volume, executives said Tuesday"

Google cancels half the projects at its internal R&D group Area 120 - "According to Area 120 lead Elias Roman, the division aims to sharpen its focus to only AI-first projects, as opposed to its earlier mandate to fuel product incubation across all of Google."

Intel, AMD and Nvidia propose new standard to make AI processing more efficient - "In pursuit of faster and more efficient AI system development, Intel, AMD and Nvidia today published a draft specification for what they refer to as a common interchange format for AI."

How WhiteLab Genomics is using AI to aid gene and cell therapy development - "French biotech company WhiteLab Genomics has raised $10 million in funding for an AI platform designed to aid the discovery and development of genomic therapies."

Concerns

Online Art Communities Begin Banning AI-Generated Images - "As AI-generated art platforms like DALL-E 2, Midjourney, and Stable Diffusion explode in popularity, online communities devoted to sharing human-generated art are forced to make a decision: should AI art be allowed? "

Artificial intelligence is playing a bigger role in cybersecurity, but the bad guys may benefit the most - "Artificial intelligence is playing an increasingly important role in cybersecurity — for both good and bad. Organizations can leverage the latest AI-based tools to better detect threats and protect their systems and data resources."

Tesla is being sued over Autopilot and Elon Musk’s Full Self-Driving predictions - "The plaintiff, Briggs Matsko, says he spent $5,000 for the package in 2018, like many Tesla drivers who’ve paid thousands of dollars for Enhanced Autopilot. That was sold as a precursor to “Full Self-Driving” tech, a now $15,000 software add-on package that still isn’t ready to ship."

The Meta Oversight Board says Facebook’s automated image takedowns are broken - "Meta’s Oversight Board says the company should be more careful with automated moderation tools, criticizing it for removing a cartoon depicting police violence in Colombia. The decision arrived as the board took on a set of new cases, including a question about a video of sexual assault in India."

This artist is dominating AI-generated art. And he’s not happy about it. - "Greg Rutkowski is a more popular prompt than Picasso."

How deep-network models take potentially dangerous 'shortcuts' in solving complex recognition tasks - "Deep convolutional neural networks (DCNNs) don't see objects the way humans do -- using configural shape perception -- and that could be dangerous in real-world AI applications, says Professor James Elder, co-author of a York University study published today."

Food Delivery Robot Casually Drives Under Police Tape, Through Active Crime Scene - "Holl A delivery robot in Hollywood said fuck the police earlier this week, when it decided that no police tape or crime scene would prevent it from delivering food to a customer."

Twitter Slams AI Scientists Who "Fixed" Ariel In 'The Little Mermaid' - "Twitter recently suspended two accounts after artificial intelligence scientists shared a modified and 'racist' clip from Disney's latest movie ‘The Little Mermaid'. The short clip whitewashed the skin of actor Halle Bailey, who is playing the role of Ariel."

Policy

Biden to hit China with broader curbs on U.S. chip and tool exports - "The Biden administration plans next month to broaden curbs on U.S shipments to China of semiconductors used for artificial intelligence and chipmaking tools, several people familiar with the matter said."

US will lose China tech war if 5G, AI, microchip ‘core’ fronts are not fortified, report warns - "The US could lose the new technology competition to China if it does not take dramatic action on three core fronts soon, a new report has warned."

Expert Opinions

10 years later, deep learning ‘revolution’ rages on, say AI pioneers Hinton, LeCun and Li - "Artificial intelligence (AI) pioneer Geoffrey Hinton, one of the trailblazers of the deep learning “revolution” that began a decade ago, says that the rapid progress in AI will continue to accelerate."

Explainers

What is AI super resolution? How it improves video images - "AI super resolution uses machine learning (ML) algorithms to infer from an original image ways that more pixels may be added to improve that image in some way. "

---

Copyright © 2022 Skynet Today, All rights reserved.