Last Week in AI #146: How AI can improve therapists, AI restores audio and video for the Beatles documentary, US blacklists SenseTime ahead of IPO
AI can analyze therapist-client conversations to provide better care, Peter Jackson used AI to improve Beatles footage, US blacklists SenseTime citing human rights concerns, and more!
Lack of access to quality mental health treatment is a growing problem, and AI-based software might be able to help. Companies working in this space have trained language models on therapist-client conversations to classify and analyze these conversations. For example, these models can identify when the conversations are following CBT (Cognitive Behavioral Therapy) techniques and guidelines, and when they're not. Such detailed analysis can help therapists deliver better care more efficiently. In the future, AI-assisted therapy may help patients "mix and match treatments" by identifying the "active ingredients" of psychotherapy and providing "precision medicine in psychology and psychiatry."
Our take: It is notable that these companies working on AI-assisted therapy are not working on chatbots to replace therapists, which I think many in both the field of AI and psychology would agree is counterproductive. Instead, these are AI-based tools that therapists and institutions can use to improve their treatments, a more realistic and promising approach. My main concern, which the article does note, is how doing this analysis requires access to private therapist-client conversations. Obtaining, storing, and using such data in ethical ways seem rather challenging.
The new 3-part documentary "The Beatles: Get Back" features about 8 hours of footage of The Beatles working in their studio and performing a concert. Putting it together not only required sifting through dozens of hours of 50-year old footage but also post-processing this footage to make it look much better. Even more importantly, the audio had to be crystal clear. Director Peter Jackson discussed how this was done in an interview with Variety:
“To me the sound restoration is the most exciting thing. We made some huge breakthroughs in audio. We developed a machine learning system that we taught what a guitar sounds like, what a bass sounds like, what a voice sounds like. In fact we taught the computer what John sounds like and what Paul sounds like. So we can take these mono tracks and split up all the instruments we can just hear the vocals, the guitars. … That that allows us to remix it really cleanly.”
AI techniques for improving the quality of old film footage have been demonstrated to work well for years, with many clips such as this one to be found on YouTube. However, this marks the first time production with this level of prominence and prestige has heavily relied on such AI techniques.
Our take: This is a great example of how AI can be used to benefit filmmaking, and I am sure many documentaries in the future will use machine learning tools to similar ends. The techniques used for audio restoration seem especially novel, and I am excited to see such innovation enabling this documentary to exist.
Clearview isn't the only controversial facial recognition out there. SenseTime, a Chinese facial recognition provider, had plans to go public on the Hong Kong stock exchange on December 17. The US accuses the company of enabling human rights abuses in Xinjiang, and the US Treasury placed SenseTime on a list of "Chinese military-industrial complex companies", adding SenseTime to an investment blacklist. Along with the sanctions and investment setbacks, SenseTime is struggling to keep its IPO afloat as the pricing of its stock offering did not take place as scheduled. SenseTime has spoken up to say the US has a "fundamental misperception" of its business while stressing its commitment to AI ethics and alleging it has merely been caught amidst geopolitical tensions.
Our take: It's a good idea for the US--and US companies--to disengage with technology companies / organizations that do not respect the same values that the US does. While I'm all for diplomacy, there is something to sticking by one's principles. Given the Chinese government's deep investment in and engagement with its technology industry, I would give some credence to the US's claims about SenseTime. However, I do hope that doesn't spill over into too general skepticism regarding Chinese companies. It would be great if geopolitical tensions didn't hamper the development of technology and the benefits that can accrue from that development. According to Jeffrey Ding, who writes the ChinAI newsletter, it is still unclear who makes up SenseTime's "much-hyped" AI ethics council. If the lack of transparency is to hide SenseTime's violation of its own core bylaws regarding the composition of the council, it would certainly damage the company's credibility on the ethics front.
Check out our newest editorial! If you are a paid subscriber (thanks!) you can read all of it, and if you are a regular subscriber (also thanks!) you can read the beginning of it as a free preview. Enjoy!
NYU Researchers Paving New Path for Robotics - "The ability to make decisions autonomously is not just what makes robots useful, it's what makes robots robots. We value robots for their ability to sense what's going on around them, make decisions based on that information, and then take useful actions without our input."
New Harvard institute to study natural, artificial intelligence - "Harvard University on Tuesday launched the Kempner Institute for the Study of Natural and Artificial Intelligence, a new University-wide initiative standing at the intersection of neuroscience and artificial intelligence, seeking fundamental principles that underlie both human and machine intelligen"
A Quadruped Humanoid Robot Might Be Able To Do It All - "Swiss-Mile's robot can stand on two legs, walk on four legs, and drive like a car"
DeepMind tests the limits of large AI language systems with 280-billion-parameter model - "The company’s new AI language model is named Gopher"
Deep-learning model speeds extreme weather predictions - "Researchers trained the Fourier Neural Operator (FNO) deep learning model—which learns complex physical systems accurately and efficiently—to emulate atmospheric dynamics and provide high-fidelity extreme weather predictions across the globe a full five days in advance."
DeepMind makes bet on AI system that can play poker, chess, Go, and more - "Unlike the other game-playing systems DeepMind developed previously, Player of Games can perform well at both perfect information games (e.g., the Chinese board game Go and chess) as well as imperfect information games (e.g., poker)."
Google's teaching AI how to see and hear at the same time - "A step closer to human-level perception"
AI could end foreign-language subtitles - "AI companies are developing methods to translate and synthesize voices in ads, movies and TV."
‘He touched a nerve’: how the first piece of AI music was born in 1956 - "Long before Auto-Tune and deepfake compositions, university professor Lejaren Hiller premiered a concert recital composed by a computer and became an overnight celebrity"
Facebook Says Its New AI Can Identify More Problems Faster - "The “Few-Shot Learner” system doesn't need to see as many examples to identify troublesome posts, and works in more than 100 languages."
Forging New Pathways: Boys & Girls Clubs Teens Take AI From Idea to Application - "Students in AI Pathways Institute use NVIDIA Jetson to build projects aimed at addressing everything from aiding patients with Alzheimer’s to monitoring pedestrian safety."
Clearview AI on track to win U.S. patent for facial recognition technology - "The government is moving to award a lucrative patent for a “search engine for faces,” a technology that has members of Congress and privacy advocates up in arms."
Uber spinout Serve Robotics gets $13M seed round to expand sidewalk robot deliveries - "Serve Robotics, the autonomous sidewalk delivery company that spun out from Uber-owned Postmates in March, has closed an expanded seed round at $13 million."
Group Backed by Top Companies Moves to Combat A.I. Bias in Hiring - "The organization has created a format for evaluating the technology, which is often used to screen job candidates."
Robotic Research raises $228M to scale autonomous driving tech - "Maryland-based Robotic Research has spent the better part of two decades developing autonomous driving technology for the Department of Defense."
Mercedes Beats Tesla to Hands-Free Driving on the Autobahn - "Daimler AG’s Mercedes-Benz won regulatory approval to deploy a hands-free driving system in Germany ahead of Tesla Inc., gaining an edge in the race to offer higher levels of automation in one of the world’s most competitive car markets."
A Cartel of Influential Datasets Are Dominating Machine Learning Research, New Study Suggests - "A new paper from the University of California and Google Research has found that a small number of ‘benchmark’ machine learning datasets, largely from influential western institutions, and frequently from government organizations, are increasingly dominating the AI research sector."
Inside Tesla as Elon Musk Pushed an Unflinching Vision for Self-Driving Cars - "The automaker may have undermined safety in designing its Autopilot driver-assistance system to fit its chief executive’s vision, former employees say."
For truly ethical AI, its research must be independent from big tech - "We must curb the power of Silicon Valley and protect those who speak up about the harms of AI"
How We Determined Crime Prediction Software Disproportionately Targeted Low-Income, Black, and Latino Neighborhoods - "The next development in data-informed policing was ripped from the pages of science fiction: software that promised to take a jumble of local crime data and spit out accurate forecasts of where criminals are likely to strike next, promising to stop crime in its tracks."
Machine-learning system flags remedies that might do more harm than good - "The system could help physicians select the least risky treatments in urgent situations, such as treating sepsis."
The Carbon Footprint of Machine Learning - "They estimated that training OpenAI’s giant GPT-3 text-generating model is akin to driving a car to the Moon and back, which is about 700,000 km or 435,000 miles."
How bias creeps into the AI designed to detect toxicity - "studies show that technologies similar to Jigsaw’s still struggle to overcome major challenges, including biases against specific subsets of users."
Alarmed by Tesla’s public self-driving test, state legislators demand answers from DMV - "Tesla is developing driverless cars on California’s public roadways using its own customers as test drivers and shrugging off test-reporting requirements — and, so far, the state’s Department of Motor Vehicles has been largely content to look the other way."
Algorithms are Making Many of Your Decisions – and You Might be OK With That - "The odds are good that at least a few algorithms helped you find this article. After all, algorithms – which are essentially systems or processes that help make a choice – have been around nearly forever."
Artificial Intelligence in the cockpit: Safety First or Profit First? - "Keynote speech for the European Cockpit Association 2021-11-26 in Brussels on the occasion of their 30th anniversary. "
Your Face Is, or Will Be, Your Boarding Pass - "Tech-driven changes are coming fast and furiously to airports, including advancements in biometrics that verify identity and shorten security procedures for those passengers who opt into the programs."
Biden administration again looks to increase AI R&D funding at civilian agencies - "At $1.7 billion, the requested funding represents an 8.8% increase over enacted fiscal 2021 civilian AI R&D investments of $1.5 billion."
America Needs AI Literacy Now - "Can artificial intelligence (AI) replace a doctor in the operating room? Are some AI algorithms inherently biased, or are they merely trained on biased data? If you’re not sure about the answers to these questions, you are not alone."
DOD to create new chief digital and AI officer post - "The Department of Defense will create a new leadership position to oversee the Joint Artificial Intelligence Center, Defense Digital Service and the DOD’s chief data officer. The idea is to centralize oversight of data and AI initiatives under one official at the highest levels of the Pentagon."
Broken Promises & Empty Threats: The Evolution of AI in the USA, 1956-1996 - "This essay adopts a periodization used in the Japanese AI community to look at the history of AI in the USA. "
Creating personalised data stories with GPT-3 - "In this post I demonstrate how GPT-3, a new and advanced language model, can construct engaging and unique stories from user-specific data, with relative ease"
Check out our audio discussions of last week’s AI news stories!
Copyright © 2021 Skynet Today, All rights reserved.