Machine learning

Prof. David Autor and his team of researchers found that junior patent attorneys who used AI for assistance showed no net improvement in their skills, reports Andy Mukherjee for Bloomberg. “When researchers subsequently tested everyone on an unassisted, offline task to gauge internalized judgment, the learning gains went entirely to senior lawyers,” writes Mukherjee. “Foundational expertise may be a prerequisite for extracting durable skill from AI-assisted practice,” the researchers noted.  

In The Wall Street Journal’s “Economics Newsletter,” reporter Greg Ip highlights a study by Prof. David Autor that found that AI did not significantly improve the performance of the 133 patent lawyers who participated. “Among junior lawyers, the performance of some got better, others got worse, and on average they were the same. Among senior lawyers, the lowest performers improved, the best didn’t,” writes Ip. “The conclusion: Lawyers don’t acquire durable new skills with AI.” 

In an opinion piece for The Wall Street Journal, Gabriele Mazzini, a research affiliate at the MIT Media Lab, explains why “policymakers and voters need to resist AI exceptionalism” amidst recent warnings of a potential AI apocalypse. “There is no sentient [AI] actor lurking in the weights to be reasoned with, feared or pacified,” writes Mazzini. “There are only human software engineers, product managers and corporate boards deciding which guardrails are worth the latency cost and which permissions can be skipped in the race to market.” 

Prof. Max Tegmark joins Fox Business's “The Claman Countdown,” to discuss warning lawmakers about the risks of rogue AI at the “Pro-Human Assembly” in Washington. “It’s very obvious that if you make these things [AI] vastly smarter than humans in the next few years, give them robotic bodies, have millions and billions of them all over earth, and don’t understand how they work, we could lose control over them. So, let’s not do that,” says Tegmark. “We have very clear ideas for how we can control AI that cures cancer and [fuels] all sorts of other powerful tools. We want to build tools, not beings that are going to replace us.” 

Prof. Max Tegmark speaks to Fox Business about the need for government regulations on AI. AI agents “have goals; they’re actively pursuing them and being incredibly creative about getting things done, even if it wasn’t the way you wanted,” says Tegmark. “Put that together with robotic bodies, and you’ve got yourself a species. If they’re vastly smarter than us, which is the goal of these companies, then they can not only do hacking, they can go online and hire a bunch of humans to work for them.”  

A recent study by MIT researchers found that AI chatbots easily generated misleading election-related answers, images and videos, writes Emily Condon for USA Today. “This, along with other findings from our study, suggests that in an ongoing election with little authoritative information for models to rely on, AI may not be as helpful in pushing back against unsubstantiated rumors and misinformation,” the researchers explain.  

Prof. Daniel Huttenlocher, dean of the MIT Schwarzman College of Computing, joins Fox Business to discuss why he feels humans should be held accountable for AI’s failures. “The key thing to remember is we cannot [and] we should not be treating AI as if it’s something that has control or has responsibility. That rests with us,” Huttenlocher emphasizes.  

A new report published in part by MIT FutureTech finds that AI does not significantly accelerate scientific research, writes Peter Hall for the Scientific American. “As the experiments become larger or more complex, so, too, do the models that simulate them, and the data to train those models is hard to come by,” Hall explains. “Robots handling real-world materials raise safety questions that are unknown to software.”  

Prof. Daniel Huttenlocher, dean of MIT’s Schwarzman College of Computing, joins Fox News’ “Fox and Friends” to discuss his views on the role of human responsibility in AI’s evolution. “With conventional software, when you program something, you give the machine precise instructions of what to do, so the humans developing it have a much more concrete understanding of what it can and can’t do,” says Huttenlocher. “With an AI system instead, it’s learned much more general skills and what you tell it to do is achieve some sort of a goal, but you don’t tell it the specific way it should achieve that goal. We can’t control it [AI] in the same ways that we control traditional software, but that doesn’t mean we shouldn’t control it and that we can’t control it, it’s just different.” 

Prof. Gary Gensler joins CNBC’s “Squawk Box” to discuss the AI race and shares his suggestions for government regulations. “I do feel that there’s a need for more responsible regulation of artificial intelligence,” says Gensler. “These existential risks [conflicts of interest, explainability, and bias], however you measure them, are much harder to control because you’ve got the competition between Chinese model developers, [and] U.S. model developers.”  

Prof. Dylan Hadfield-Menell speaks to The Boston Globe’s Joshua Miller about his concerns with the current state of AI systems and how they are optimized. “The parts [of AI systems] that are imitating the ways that humans use language, those parts of the systems actually are quite correctable. They’re still pretty unpredictable, but they do seem quite flexible,” says Hadfield-Menell. “On the other hand, there’s reinforcement learning, task-focused behavior, that can often be quite sticky. They [AI systems] seem to really want to push towards task completion.” He adds: “If you have a powerful optimization-driven system pointed at some goal, it can often do a lot of unexpected things that can have a lot of impact.” 

Prof. Shafi Goldwasser and Prof. Vinod Vaikuntanathan founded the Institute for Responsible Superintelligence (RESI), “to develop mathematical principles and protocols to ensure AI models behave as intended,” writes Aaron Pressman for The Boston Globe. “You patch them [AI models] up, you test them, you patch them to make sure that they can perform to the best ability, but also not do damage,” says Goldwasser. “That’s probably not enough attention, given the fact that these things are going to become what we call super intelligent.” 

In an article for The Wall Street Journal, Prof. Daniel Huttenlocher, dean of MIT’s Schwarzman College of Computing writes that: “AI is an extremely powerful and beneficial technology. But accountability for its actions remains with the people and institutions that grant it the power to act, and so does everything that follows: the duty to justify the arrangement, to watch it, to correct it and to stop it when necessary.”  

The Boston Globe’s Hiawatha Bray spotlights Prof. Max Tegmark speaking at the “Pro-Human Assembly” in Washington, an event advocating for government regulations on AI, sponsored by Tegmark’s Future of Life Institute. “We are here today because we know that what’s up next is a fork in the road,” said Tegmark at the assembly. “We know that the future of AI is not inevitable. It is for us to decide.”  

Members of the MIT community speak to Boston 25 News reporter Lisa Gresci about how they use AI, and their concerns with regulating the technology. Developers should consider, “how AI can help communities so [they are] not just building things, but are also deploying it [with purpose] out in the world,” says graduate student Anku Rani.