Keeping up with an industry as fast-moving as\u00a0AI<\/a>\u00a0is a tall order. So until an AI can do it for you, here\u2019s a handy roundup of the last week\u2019s stories in the world of machine learning, along with notable research and experiments we didn\u2019t cover on their own.<\/p>\n This week, movers and shakers in the AI industry, including<\/a> OpenAI CEO Sam Altman, embarked on a goodwill tour with policymakers \u2014 making the case for their respective visions of AI regulation. Speaking<\/a> to reporters in London, Altman warned that the EU\u2019s proposed AI Act, due to be finalized next year, could lead OpenAI ultimately to pull its services from the bloc.<\/p>\n \u201cWe will try to comply, but if we can\u2019t comply we will cease operating,\u201d he said.<\/p>\n Google CEO Sundar Pichai, also in London, emphasized the need for \u201cappropriate\u201d AI guardrails that don\u2019t stifle innovation. And Microsoft\u2019s Brad Smith, meeting with lawmakers in Washington, proposed<\/a> a five-point blueprint for the public governance of AI.<\/p>\n To the extent that there\u2019s a common thread, tech titans expressed a willingness to be regulated \u2014 so long as it doesn\u2019t interfere with their commercial ambitions. Smith, for instance, declined to address the unresolved legal question of whether training AI on copyrighted data (which Microsoft does) is permissible under the fair use doctrine in the U.S. Strict licensing requirements around AI training data, were they to be imposed at the federal level, could prove costly for Microsoft and its rivals doing the same.<\/p>\n Altman, for his part, appeared to take issue with provisions in the AI Act that require companies to publish summaries of the copyrighted data they used to train their AI models, and make them partially responsible for how the systems are deployed downstream. Requirements to reduce the energy consumption and resource use of AI training \u2014 a notoriously compute-intensive process \u2014 were also questioned.<\/p>\n The regulatory path overseas remains uncertain. But in the U.S., the OpenAIs of the world may get their way in the end. Last week, Altman wooed members of the Senate Judiciary Committee with carefully-crafted statements about the dangers of AI, and his recommendations for regulating it. Sen. John Kennedy (R-LA) was particularly deferential: \u201cThis is your chance, folks, to tell us how to get this right \u2026 Talk in plain English and tell us what rules to implement,\u201d he said.<\/p>\n In comments to The Daily Beast, Suresh Venkatasubramanian, Brown University\u2019s director of the Center for Tech Responsibility, perhaps<\/a> summed it up it best: \u201cWe don\u2019t ask arsonists to be in charge of the fire department.\u201d And yet that\u2019s what\u2019s in danger of happening here, with AI. It\u2019ll be incumbent on legislators to resist they honeyed words of tech execs and clamp down where it\u2019s needed. Only time will tell if they do.<\/p>\n Here are the other AI headlines of note from the past few days:<\/p>\n Bill Gates may not be an expert on AI, but he is<\/em> very rich, and he\u2019s been right on things before. Turns out he is bullish on personal AI agents, as he told Fortune<\/a>: \u201cWhoever wins the personal agent, that\u2019s the big thing, because you will never go to a search site again, you will never go to a productivity site, you\u2019ll never go to Amazon again.\u201d How exactly this would play out is not stated, but his instinct that people would rather not borrow trouble by using a compromised search or productivity engine is probably not far off base.<\/p>\n Evaluating risk in AI models is an evolving science, which is to say we know next to nothing about it. Google DeepMind (the newly formed superentity comprising Google Brain and DeepMind) and collaborators across the globe are trying to move the ball forward, and have produced a model evaluation framework<\/a> for \u201cextreme risks\u201d such as \u201cstrong skills in manipulation, deception, cyber-offense, or other dangerous capabilities.\u201d Well, it\u2019s a start.<\/p>\n Image Credits:<\/strong> SLAC<\/p>\n<\/div>\n Particle physicists are finding interesting ways to apply machine learning to their work: \u201cWe\u2019ve shown that we can infer very complicated high-dimensional beam shapes from astonishingly small amounts of data,\u201d says SLAC\u2019s Auralee Edelen.<\/a> They created a model that helps them predict the shape of the particle beam in the accelerator, something that normally takes thousands of data points and lots of compute time. This is much more efficient and could help make accelerators everywhere easier to use. Next up: \u201cdemonstrate the algorithm experimentally on reconstructing full 6D phase space distributions.\u201d OK!<\/p>\n Adobe Research and MIT collaborated on an interesting computer vision problem: telling which pixels in an image represent the same material<\/em><\/a>. Since an object can be multiple materials as well as colors and other visual aspects, this is a pretty subtle distinction but also an intuitive one. They had to build a new synthetic dataset to do it, but at first it didn\u2019t work. So they ended up fine-tuning an existing CV model on that data, and it got right to it. Why is this useful? Hard to say, but it\u2019s cool.<\/p>\n Frame 1: material selection; 2: source video; 3: segmentation; 4: mask Image Credits:<\/strong> Adobe\/MIT<\/p>\n<\/div>\n Large language models are generally primarily trained in English for many reasons, but obviously the sooner they work as well in Spanish, Japanese, and Hindi the better. BLOOMChat is a new model<\/a> built on top of BLOOM that works with 46 languages at present, and is competitive with GPT-4 and others. This is still pretty experimental so don\u2019t go to production with it but it could be great for testing out an AI-adjacent product in multiple languages.<\/p>\n NASA just announced a new crop of SBIR II fundings, and there are a couple interesting AI bits and pieces in there:<\/p>\n Geolabe<\/a> is detecting and predicting groundwater variation using AI trained on satellite data, and hopes to apply the model to a new NASA satellite constellation going up later this year.<\/p>\n Zeus AI<\/a> is working on algorithmically producing \u201c3D atmospheric profiles\u201d based on satellite imagery, essentially a thick version of the 2D maps we already have of temperature, humidity, and so on.<\/p>\n Up in space your computing power is very limited, and while we can run some inference up there, training is right out. But IEEE researchers want to make a SWaP-efficient neuromorphic processor<\/a> for training AI models in situ.<\/p>\n Robots operating autonomously in high-stakes situations generally need a human minder, and Picknick<\/a> is looking at making such bots communicate their intentions visually, like how they would reach to open a door, so that the minder doesn\u2019t have to intervene as much. Probably a good idea.<\/p>\n<\/p><\/div>\n\n
Other machine learnings<\/h2>\n
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2023\/05\/similar-materials-2.gif\")
<\/p>\n![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2023\/05\/zeus-1.jpg\")
<\/p>\n