Last Week in AI #94
AlphaFold, Timnit Gebru, and more!
The problem of predicting protein structures from amino acid sequences is an important one in biology. Accurate prediction of protein structure would have benefits such as enabling quicker and more advanced drug discovery. Google's AlphaFold was recently recognized as a solution to this problem by the organisers of the biennial Critical Assessment of protein Structure Prediction (CASP). DeepMind's latest version of AlphaFold significantly outperformed other methods at a protein-folding contest. DeepMind's own announcement was accompanied by a great deal of praise and reporting, but not everyone believes that DeepMind has truly "solved" the problem, as a great deal of reporting seems to indicate. While DeepMind's AlphaFold is an impressive step forward, we will likely see much more analysis and discussion of its true capabilities in the coming weeks.
On December 2, the well-known AI ethics researcher Timnit Gebru tweeted that she had been fired from her position at Google as a co-lead of their Ethical AI research group. Gebru is known for, among other research, her work on the Gender Shades project that showed a number of industrial facial recognition systems worked worse for some races and genders than for others. According to Gebru, there was conflict over a paper that was critical of environmental costs and biases in large language models that include those used by Google. Gebru argued that her work was being censored, while Google claims Gebru did not allow enough time for internal review, having already submitted her paper to an external conference before sending it for internal review. An email Gebru sent to a Google Brain listserv also appears to have been involved in Google's decision to let her go. Google's Jeff Dean has responded publicly to the news of Gebru's firing and explained Google's internal review process. Given Gebru's standing in the AI ethics community, Google's decision to fire her has drawn ire, while the story has been covered by The New York Times, The Washington Post, MIT Technology Review, and a number of other news organizations.
Advances & Business
2021 will see the spread of AI and automation in the workplace - "Enterprises are already interested in how to automate invoicing and expensing procedures, says SAP Concur."
Robots And AI Can 10X Productivity. Here's Why That Might Be Good For Workers - "Lior Elazary believes that robots are a net positive for humans. That's a good thing, because he is the CEO of InVia Robotics and offers a robots-as-a-service platform that can 5-10X productivity in shipping and receiving warehouses. And presumably, he wants to sleep at night."
WaveOne aims to make video AI-native and turn streaming upside down - "Video has worked the same way for a long, long time. And because of its unique qualities, video has been largely immune to the machine learning explosion upending industry after industry."
The Cambrian AI Explosion Ramps Up - "There's been a lot of news lately on the AI chip front, so I wanted to share a short synopsis of what has been happening for anyone who may be distracted by the holidays. I will flesh these stories out further before my annual Cambrian AI Explosion blog in January here on Forbes."
Alphabet's Loon hands the reins of its internet air balloons to self-learning AI - "Alphabet's Loon, the team responsible for beaming internet down to Earth from stratospheric helium balloons, has achieved a new milestone: its navigation system is no longer run by human-designed software."
Adobe accelerates AI 'visual recommendations' - "Next week, Adobe is rolling out 'visual similarity recommendations' which offer AI-powered product suggestions based on what consumers are considering purchasing. And this on-the-fly use of visual interpretation and recommendation is just the start."
This Company Uses AI to Outwit Malicious AI - "Robust Intelligence is among a crop of companies that offer to protect clients from efforts at deception."
AI can turn spoken language into photorealistic sign language videos - "An AI that can produce photorealistic videos of sign language interpreters from speech could improve accessibility by removing the need for humans. Ben Saunders at the University of Surrey, UK, and his colleagues used a neural network that converts spoken language into sign language."
Labor Unions Work to Find Ways to Bargain With AI's Black Box - "Unions are trying to figure out how to protect their members from artificial intelligence programs they can't see and may not even be sure are really there. The U.K.'s Trades Union Congress, an umbrella group for British unions representing more than 5."
Amazon debuts Trainium, a custom chip for machine learning training in the cloud - "Amazon today debuted AWS Trainium, a chip custom-designed to deliver what the company describes as cost-effective machine learning model training in the cloud."
Concerns & Hype
How Mark Kelly used conversational AI to help win a Senate seat - "Amplify.ai, an enterprise-level conversational AI platform, has helped 2020 senatorial campaigns drive engagement with local constituents. Senators-elect Mark Kelly (D-AZ) and John Hickenlooper (D-CO), ... have publicly partnered with the company."
That's all for this week! If you are not subscribed and liked this, feel free to subscribe below!
Copyright © 2020 Skynet Today, All rights reserved.