Last Week in AI #151: AI early warning system for new Covid-19 variants, increased federal spending on facial recognition, AI used to detect fake art
An "early warning system" for COVID-19 variants, Feds' spending on facial recognition tech expands, A.I. Research Is changing the way conservators and collectors think about attribution, and more!
Over the course of the pandemic, we've been alerted to a number of concerning variants coming out of regions including the UK, Brazil, and recently South Africa. Some of these have swept up the globe, but more than just those few might endanger lives. Flagging such variants early would allow for more time to respond to variants of concern. BioNTech has teamed up with London-based AI company InstaDeep to develop such an early warning system. They claim that in tests, the system picked up 12 of the 13 variants the WHO designated as dangerous and did so two months, on average, before the WHO reached its conclusion. The actual system uses four metrics derived from the DNA sequence of a variant to score its potential danger.
Our take: Another great example of how AI systems can help in crises. As always, it's worth remembering that these systems are tools and that their judgment is far from beyond question, but they can be incredibly useful and help us make decisions and act more quickly than we might on our own.
As we’ve covered in our most recent editorial, facial recognition company Clearview AI has often been in the spotlight over the past two years – usually because of various groups filing lawsuits against it. Nevertheless, it is still alive and well, and we now know that the FBI signed a contract to use its services just last month. While the contract itself is small (only valued at $18,000), it does set an important precedent for the agency as a whole. Other U.S. government groups such as the U.S. Immigration and Customs Enforcement (ICE) and many local police departments have also signed contracts with the company.
Privacy and civil rights advocates have criticized Clearview AI and those who worked with it, and have fought for regulations to shut it down. While laws have been proposed in Congress, none have passed yet; on the other hand, more than 27 states and localities have passed some sort of ban on facial recognition use.
Our take: There is an argument to be made that facial recognition really can aid police departments and groups such as the FBI do their jobs more effectively. However, I personally oppose them doing so until laws are passed that regulate what these organizations can and cannot do, and that requires transparency on how they use the technology. Moreover, it is important that police officers are well informed about the limits of the technology; as we’ve covered, there have already been three documented cases of wrongful arrests due to incorrect use of facial recognition, and more are likely to surface with time.
‘We Were Blown Away’: How New A.I. Research Is Changing the Way Conservators and Collectors Think About Attribution
A new study applies deep learning to identify artworks created by different artists not by looking at the content of a painting but by looking at the 3D texture of its surface. The difference among painters was much more obvious in their surface topologies, even at small scales like at the level of the diameter of a single bristle. The researchers argue that this method is more accurate and general than previous methods, which directly analyzed the colored image of a painting. In the future, the researchers hope to extend this approach beyond oil and acrylic paintings to drawings, watercolors, and sculptures.
Our take: This is an interesting example of improving the performance of an AI system on a task not by developing more advanced algorithms but by using a more informative type of data. While the initial study is still limited in scope and the algorithm is unlikely to identify forgeries all on its own, it may still be an important tool for future conservators when studying past artworks.
Apple ML Researchers Introduce ARKitScenes: A Diverse Real-World Dataset For 3D Indoor Scene Understanding Using Mobile RGB-D Data - "Understanding indoor 3D scenes are becoming increasingly important in augmented reality, robotics, photography, games, and real estate. Many state-of-the-art scene interpretation algorithms have lately been driven by modern machine learning approaches."
Has AI found a treatment for Fragile X? - "New funding for AI clinical prediction platform points to a streamlined pharma future."
AI makes it possible to simulate 25 billion water molecules at once - "Computer simulations of clouds of atoms and molecules must always trade scale for accuracy, but a new technique shows that both are possible at once using AI and clever coding"
Voyager AI Supercomputer Gives Investigators New Deep Learning Experimental Platform - "There is a lot of interest in using AI for wildfire management"
AI won’t replace psychologists, it will only make them better - "The application of machine learning models and AI in behavioral therapies is an innovative field that is evolving rapidly, with the potential to significantly improve initial assessment, ongoing progress monitoring, and clinical outcomes"
This AI Software Nearly Predicted Omicron’s Tricky Structure - "New algorithms that decipher complex sequences of amino acids offered an early view of the coronavirus variant. They could point the way to future drugs."
Rytr Reviewed: How the GPT-3 ‘AI Writing Assistant’ Performs In Real Life - "As consumer and corporate interest in artificial intelligence has grown, the number of companies purporting to offer AI-powered software has exploded. Microsoft has a web page devoted to the various ways it has baked AI into Office and Bing."
Future developments in AI could make your credit score obsolete - "AI is enabling better credit systems and early wage access"
How AI and language intelligence are helping employees learn - "With the ability to aggregate key information from a vast number and variety of sources, AI opens the door to better insights, better collaboration, and, ultimately, better results."
Nvidia’s AI-powered scaling makes old games look better without a huge performance hit - "Deep Learning Dynamic Super Resolution was added in the latest driver for RTX cards"
Robotic Arms Are Using Machine Learning to Reach Deeper Into Distribution - "Advances in technology are moving automated picking systems from the lab to labor-strapped warehouses"
Serve Robotics’ new autonomous sidewalk delivery robots don’t require human assist - "Serve Robotics, an Uber spinout that builds sidewalk delivery robots, is deploying its next generation of robots that are capable of completing some commercial deliveries without a human in the loop, according to the startup."
Why Multi-Functional Robots Will Take Over Commercial Robotics - "Single-task robots will soon make way for multi-application robots of the future"
Placer.ai, a location analytics startup, finds $100M at a $1B valuation - "Now, as we shift back into more “normal” behavior, a startup that’s helping to better understand where and how we are getting around has picked up a significant round of funding."
JD.com opens robotic shops in the Netherlands as Chinese e-commerce giant tests new model in Europe - "The stores merge online ordering with pickup shops, where robots prepare parcels for collection"
Nuro unveils third-generation autonomous delivery vehicle - "One of the most eye-catching features of the new vehicle is its external, front airbag. This airbag deploys if the vehicle collides with pedestrians and bikers, and is meant to keep those around the vehicle safe."
SeekOut will expand AI-powered recruitment tools with new $125M - "SeekOut is a powerful AI-driven talent search engine helping more than 1,000 companies get a competitive advantage in recruiting hard-to-find and diverse talent."
Self-driving truck company Torc Robotics picks Austin for tech hub - "Torc Robotics is opening an Austin engineering hub to accelerate development of its self-driving truck technology. The 21,000-square-foot office will join the company's Blacksburg, Va., headquarters and its Albuquerque, N.M. test center."
Labelbox Raises $110 Million Series D Led by SoftBank Vision Fund - "Labelbox, the leading training data platform for enterprise machine learning applications, announced the closing of a $110 million Series D funding led by SoftBank’s Vision Fund 2."
Second Nature raises $12.5M to coach salespeople with AI-powered avatars - "Second Nature hosts a simulator with avatars that have conversations with reps based on talk tracks and sales playbooks, measuring how deeply reps cover key topics."
Waymo and Trucker J.B. Hunt Partner Up. This Isn’t Just About Self-Driving Vehicles. - "Friday, the companies announced a deal designed to bring self-driving technology to J.B. Hunt’s (ticker: JBHT) fleet of trucks"
AI is quietly eating up the world’s workforce with job automation - "The debate around whether AI will automate jobs away is heating up. AI critics claim that these statistical models lack the creativity and intuition of human workers and that they are thus doomed to specific, repetitive tasks. However, this pessimism fundamentally underestimates the power of AI. "
Great Resignation Versus Increasing Investment In AI, Robotics And Automation: A Troubling Trend - "An increase in lower-level jobs, and a decrease in those willing to apply for them, is creating a powerful (and potentially troubling) decision matrix - tipping the scales in favor of automation, robotics and AI (artificial intelligence). "
Guardrail failure: Companies are losing revenue and customers due to AI bias - "New survey finds that 80% of U.S. firms found problems despite having bias monitoring or algorithm tests already in place."
Text autocompletion systems aim to ease our lives, but there are risks - "the AI behind these tools has shortcomings that could introduce biases or influence the language used in messaging in undesirable ways."
When It Comes to Health Care, AI Has a Long Way to Go - "Medical information is more complex and less available than the web data that many algorithms were trained on, so results can be misleading."
DMV ‘revisiting’ its approach to regulating Tesla’s public self-driving test - "For years, Tesla has tested autonomous vehicle technology on public roads without reporting crashes and system failures to the California Department of Motor Vehicles, as other robot car developers are required to do under DMV regulations."
How countries are leveraging computing power to achieve their national artificial intelligence strategies - "With the rise of artificial intelligence and machine learning, there is an increasing demand for these powerful computers and, as a result, processing power is rapidly increasing. "
Jeff Dean: AI isn't as smart as you think -- but it could be | TED - "What is AI, really? Jeff Dean, the head of Google's AI efforts, explains the underlying technology that enables artificial intelligence to do all sorts of things, from understanding language to diagnosing disease -- and presents a roadmap for building better, more responsible systems that have a dee"
“Chatbots: Still Dumb After All These Years” - "Fifty-two years later, we’re still waiting. "
Github Copilot Wants to Play Chess Instead of Code - "I used GitHub Copilot to write English instead of code and found out it can do some surprising non-code-related tasks."
Copyright © 2022 Skynet Today, All rights reserved.