Last Week in AI #177: OpenAI commercializes DALL-E 2, Sony AI beats human competitors in racing game, Gmail getting smarter searches, and more!
Discover more from Last Week in AI
Last Week in AI #177: OpenAI commercializes DALL-E 2, Sony AI beats human competitors in racing game, Gmail getting smarter searches, and more!
Copyright concerns with commercialized AI generated models, Sony AI learns to outrace humans in race simulation game, Gmail will search by user intent beyond simple keyword matching
Top News
Commercial image-generating AI raises all sorts of thorny legal issues
Last week OpenAI moved DALL-E 2, the image generation tool, into Beta (the company hopes to expand its current user base to 1 million) while granting users the “the right to reprint, sell, and merchandise” images they generate with DALL-E. This is useful for users who wish to use the generated images for commercial purposes, like making illustrations for children's books. However, as DALL-E was trained on 650 million image-text pairs from the Internet, the model can easily generate materials that include trademarked logos, characters, and other IPs that may constitute copyright infringements. Other openly available AI image generation models face similar problems. Also, it’s not clear if OpenAI violated any IP laws for just training on these Internet images and then commercializing their model. While the UK is exploring allowing commercial use of models trained on public but trademarked data, the U.S. may not follow suit.
Our take: I’m not a lawyer and I have no legal expertise in copyright law. But if I were to comment on these 2 matters - training commercial models on Internet data and using the model to generate trademarked IP - the first one seems less problematic than the second. The first one seems analogous to search engines scrapping links and metadata from websites (although Yelp famously had issues with Google putting Yelp reviews in Google Maps). The second one however does significantly lower the bar for generating new copyright-infringing materials, and OpenAI would make a profit on them (the company charges money per image). We will probably have to wait for future lawsuits for these matters to settle.
Sony’s racing AI destroyed its human competitors by being nice (and fast)
Over the past few decades, intelligent systems have beaten the best humans at a variety of games and tasks. Now, Sony AI's Gran Turismo Sophy can beat some of the best human sim-racers. It wasn't a straightforward path--first optimized for speed on empty tracks, GT Sophy lost in races where there were other human racers. But Sony's team realized that pure speed and aggression was not the way to win a race that required intelligence as well, and so trained GT Sophy “to balance its aggression and timidity, picking the most appropriate behavior for the situation at hand.” Sony plans to put GT Sophy in a future version of Gran Turismo, and thinks the sim-racing community could learn a lot from watching it race.
Our Take: It's worth noting that GT Sophy has access to information about the cars and track in a game that no human will know precisely. Humans stare at a screen, but Sophy sees Gran Turismo's virtual environment as a static map and "has access to certain information about what the car is doing in its environment, including three-dimensional velocity, angular velocity, acceleration, the load on each tire, and tire-slip angles." Nonetheless, this is an interesting announcement and it's still possible we might see a future version of Sophy that is capable of beating humans while only operating on pixels.
Gmail is getting a machine learning boost for smarter searches
Gmail is revamping search by incorporating machine learning in their algorithms, a step forward from their current technology based on rule-based techniques like keywords and filters. The update will allow users to search for emails with lesser context and fewer keywords; put more simply, users will have to spend less effort in unearthing that niche, age-long email of which they might only vaguely remember the gist, but, sadly, no identifiable keywords. Machine learning can retrieve that niche email by detecting and matching users’ “intent” rather than requiring specific keywords. This technology will be available to users in approximately two weeks.
Our Take: Searching for emails might sound like a small task, but it can be very important when it’s needed. Using machine learning methods to search emails with very limited information will be more than appreciated by users who may not remember keywords that are specific enough to retrieve that email. As the update suggests, one aspect of improving the results includes performing searches based on user behavior. While this sounds useful, it could be a cause of concern for user privacy.
Other News
Research
Microsoft launches simulator to train drone AI systems - "Microsoft has launched a platform to train the artificial intelligence (AI) systems of autonomous aircraft. Project AirSim is, in effect, a flight simulator for drones, which companies can use to train and develop software controlling them."
AI tech to automate process of denture design and enhance treatment efficiency without compromising accuracy - "Loss of permanent teeth is usually caused by dental diseases or trauma and is common in the global population, especially among the elderly due to aging and relatively poorer oral health."
NSF Funds Machine-Learning Research at UNO and UNL to Study Energy Requirements of Walking in Older Adults - "For a healthy young adult, walking for ten minutes consumes less energy than the calorie content of one slice of bread. However, as we grow older, our bodies become less energy efficient, turning simple daily activities like walking around a block into a daunting effort."
Robot Outperforms a Surgeon in a Precision Training Task - "Who is better at performing surgery: an experienced surgeon or a robot?"
UT hosts machine learning demonstration - "The training supported by state lawmakers is meant to help peak interest in students and ready them for a career in machine technology."
Open source platform enables research on privacy-preserving machine learning - "The biggest benchmarking data set to date for a machine learning technique designed with data privacy in mind has been released open source by researchers at the University of Michigan."
Berkeley shows off accelerated learning that puts robots on their feet in minutes - "Robots relying on AI to learn a new task generally require a laborious and repetitious training process. Berkeley researchers are attempting to simplify and shorten that with an innovative learning technique that has the robot filling in the gaps rather than starting from scratch."
Novel method allows robots to learn in the wild - "The robot watched as Shikhar Bahl opened the refrigerator door. It recorded his movements, the swing of the door, the location of the fridge and more, analyzing this data and readying itself to mimic what Bahl had done."
DeepMind & UCL’s Stochastic MuZero Achieves SOTA Results in Complex Stochastic Environments - "The novel approach achieves performance comparable or superior to state-of-the-art methods in complex single- and multi-agent environments while maintaining the superhuman performance of the original MuZero"
Clinicians can build trust with machine learning through experience, study says - "A new study published in Nature Partner Journals Digital Medicine revealed the barriers to machine learning adoption within clinics; however, the study suggests that clinicians can build trust with machine learning through experience."
Art and Artificial Intelligence - "Faculty from the LSU College of Art and Design are using robots and artificial intelligence in combination with art and landscape. Associate Professor of Art Hye Yeon Nam and Assistant Professor of Landscape Architecture Brendan Harmon are collaborating on two different projects."
Applications
6 AI innovations for those aging in place - "AI and technology offer hope: The New York Office for the Aging recently partnered with ElliQ by Intuition Robotics to bring empathetic care companions into the homes of hundreds of seniors aging in place across the state."
Samsung launches a new AI-enhanced app to fix your blurry photos - "Samsung's Gallery app is feature-rich and packs powerful editing options like Spot Color, Object Eraser, and Shadow Remover. Most of these options are buried deep in the photo editor, but they work very well once you figure out where they are."
Machine learning identifies gun buyers at risk of suicide - "A new study from the Violence Prevention Research Program (VPRP) at UC Davis suggests machine learning, a type of artificial intelligence, may help identify handgun purchasers who are at high risk of suicide."
Banks Turn to AI to Help Dodge Enforcement Spotlight - "Regulators increasingly expect banks and other entities to use smarter systems to help thwart potential criminal activity"
Virginia Department of Transportation Preparing to Use Artificial Intelligence to Predict Traffic - "This marks the first time that AI will be used to predict traffic disruptions in Virginia. "
Diabeloop rounds up €70M to take AI-powered insulin delivery tech global - "After decades of having to juggle multiple devices and fingerprick systems to take their own blood sugar readings and use them to calculate insulin dosages, people with diabetes now have access to technologies that’ll do virtually all of the work for them."
Robotics and AI are going from cage to stage - "A lot of promising companies come out of work by researchers at universities, or even grad students who have struck on some new innovation."
AI speeds sepsis detection to prevent hundreds of deaths - "Patients are 20% less likely to die of sepsis because a new AI system developed at Johns Hopkins University catches symptoms hours earlier than traditional methods, an extensive hospital study demonstrates."
How Robots Can Help Us Act and Feel Younger - "By 2050, the global population aged 65 or more will be nearly double what it is today. The number of people over the age of 80 will triple, approaching half a billion."
Business
Neural Sleeve is a bionic leg wrap that uses AI to correct walking patterns - "California startup Cionic and Yves Béhar's design studio Fuseproject have developed a bionic wearable that uses electric pulses and artificial intelligence to correct muscle movements in people with limited mobility."
Inside Midjourney, The Generative Art AI That Rivals DALL-E - "AI tools like DALL-E that generate images from short text descriptions are all the rage right now, so it only makes sense that other systems are starting to compete for space in the budding AI art market."
HiddenLayer emerges from stealth to protect AI models from attacks - "As AI-powered services like OpenAI’s GPT-3 grow in popularity, they become an increasingly attractive attack vector. Even shielded behind an API, hackers can attempt to reverse-engineer the models underpinning these services or use “adversarial” data to tamper with them."
Olive AI lays off 450 - "Olive AI this morning announced it would lay off 450 employees amid a shift in business strategy, according to a staff-wide email sent by CEO Sean Lane."
Tesla to increase cost of FSD beta software beyond its $12,000 price tag - "Tesla vehicles today come standard with its driver assistance system called Autopilot. For an additional $12,000, owners can buy FSD – a feature that Musk has repeatedly promised will one day deliver full autonomous driving capabilities."
Amazon-owned self driving firm Zoox seeks to test robotaxi in California - "Zoox, a self-driving technology firm owned by Amazon.com Inc (AMZN.O), said on Tuesday it was gearing up to launch its robotaxi business, self-certifying that its vehicle with no pedals or steering wheel meets U.S. federal regulations and applying for a permit in California to test-drive it."
Zesty.ai raises $33M for its AI-powered climate risk platform - "Zesty.ai Inc., a startup with a climate risk platform that provides data about wildfires, storms and other severe weather events, has raised $33 million in new financing. "
Baidu unveils latest autonomous electric vehicle: Apollo RT6 - "Baidu, a Chinese search engine and artificial intelligence firm, unveiled its latest electric autonomous driving vehicle on Thursday. The Apollo RT6 will be soon be part of Baidu’s robotaxi fleet, as China pushes forward with its autonomous driving ambitions."
Theator, an AI platform that analyzes surgery videos, closes out its Series A at $39.5M - "When it comes to video-based data, advances in computer vision have given a huge boost to the world of research, making the process of analyzing and drawing insights from moving images something that is scalable beyond the limits of a small team of humans."
Google fired engineer who said its AI was sentient - "Blake Lemoine, who claimed Google’s chatbot generator LaMDA was sentient, has been fired."
Concerns
Will AI Steal Submarines’ Stealth? - "Submarines are valued primarily for their ability to hide. The assurance that submarines would likely survive the first missile strike in a nuclear war and thus be able to respond by launching missiles in a second strike is key to the strategy of deterrence known as mutually assured destruction."
AI Helps the Powerful but Harms the Vulnerable, Mozilla Warns - "AI is great for rich and powerful people and for tech giants trying to boost profits. Otherwise, artificial intelligence and the automation it enables can be harmful, nonprofit Mozilla concluded in a report published Monday."
Analysis
AI Art Is Challenging the Boundaries of Curation - "In just a few years, the number of artworks produced by self-described AI artists has dramatically increased. Some of these works have been sold by large auction houses for dizzying prices and have found their way into prestigious curated collections."
The AI Battle Rages On - "How it could impact your job and 401k"
GitHub Copilot users feel more productive - "A GitHub study has found that the developers who report the greatest productivity gains from using GitHub Copilot, the company’s AI-based programming assistant, were those who accepted the most Copilot code suggestions."
Surreal or too real? Breathtaking AI tool DALL-E takes its images to a bigger stage - "When the Silicon Valley research lab OpenAI unveiled DALL-E earlier this year, it wowed the internet. The tool is seen as one of the most advanced artificial intelligence systems for creating images in the world."
Fun
If A.I. Generated Book Jackets Based on Literary Titles… - "This is why we don’t trust algorithms to design hardcovers. The internet has been having fun of late with DALL-E Mini, newly renamed craiyon, an A.I. program that generates images in response to user suggestions."
New AI-generated horsies - "Recently I've been experimenting with DALL-E 2, one of the models that uses CLIP to generate images from my text descriptions. It was trained on internet text and images, so there's a lot it can do, and a lot of ways it can remix the stuff it's seen online. I decided to have it generate new horsies."
Copyright © 2022 Skynet Today, All rights reserved.