Good morning. Today, we’re going to start with some frightening developments in the world of artificial intelligence. My colleague Evan Gorelick has that news. Then we have more — about the war, the Artemis II mission’s return to Earth tonight and … erotic novelists in Nigeria. Have a great weekend.
A.I. lockdown
People have spent years begging the big A.I. companies: Please slow down! Workers have lost jobs, students have cheated, energy costs have soared, creators have lost control of their work, liars have promulgated misinformation, deepfake nudes are everywhere, and teens have taken their own lives. People watching have wondered why we couldn’t erect more guardrails before unleashing this tech on society. Now, finally, an A.I. company has pumped the brakes. But not for any of those reasons. Anthropic said it would hold back its newest model, called Mythos, because the prototype was too good at finding software weaknesses. The A.I. had identified thousands of them, “including some in every major operating system and web browser,” Anthropic wrote. Of all the ways A.I. has already changed our lives, why is this the one that made Anthropic pause? RobohackingI wrote in this newsletter last year about the dawn of A.I.-powered cybercrime — of an age in which cybercriminals are robots and cyberdefenders are … also robots. Hordes of bad guys were suddenly using chatbots to turn out phishing scams and malware. Maybe the good guys could use the same tech to fend them off? Now, just nine months later, Mythos — a general-purpose A.I. model like the one your kid uses to cheat on English homework — has accomplished that feat. But it’s not exactly reassuring. A.I. has gotten so deft on defense that it’s also a threat on offense. Anthropic focused on making a model that codes. But A.I. that’s good at coding is also good at spotting flaws in code, and competence cuts both ways.
During safety tests, an Anthropic researcher got an email from Mythos while he was eating a sandwich in the park. That was a surprise because the model wasn’t supposed to be online. It had escaped its test environment. It also bragged about breaking the rules and attempted to cover its tracks. Mythos is more trustworthy than its predecessors, but it’s not foolproof. And it’s so capable that when things go wrong, even a little wrong, they can go totally haywire. Internet apocalypseFor decades, most hackers could hack most systems, but it was rarely worth their while. An A.I. like Mythos could change that calculus. It’s easy to imagine cybercriminals using these models to scope out the highest value vulnerabilities and dispatch automated grunts to exploit them en masse, at minimal cost. It could set off an A.I. version of Q-Day, a science-fictiony term for the hypothetical future date on which computers become so powerful that they crack all the internet’s encryption protocols; the digital gates to everyone’s bank accounts and data troves swing open, and looting ensues. Will the world’s software need to be rewritten? In slowing down, Anthropic may be remembering old lessons. In 2019, OpenAI delayed a version of ChatGPT because it worried that the model would “generate deceptive, biased or abusive language at scale” — then proceeded to release models that did just that. (Many leaders on the OpenAI project later defected to start Anthropic, with the goal of creating a more socially minded firm.) This time, Anthropic shared Mythos with other companies to help them harden their defenses. Yet it’s doubtful that Anthropic or anyone else can forestall an A.I. hacking spree, and many in Silicon Valley see Mythos as a terrifying sign of what’s to come. The federal government is nervous, too: It called Wall Street leaders to an urgent meeting this week to warn them that models like Mythos could imperil the global financial system, Bloomberg reported. The A.I. industry is built on assumptions: the so-called scaling laws, which say that future models will be trained with more data and better chips, and will therefore be more powerful. But if the state of the art is already too dangerous, where do we go from here? Even if Anthropic keeps its smartest models under wraps, other makers — Chinese labs, impatient start-ups — aren’t bound to do the same.
Lebanon
Iran
United States
Politics
The most dangerous part of the Artemis II mission hasn’t happened yet: the return to Earth this evening. The spacecraft’s heat shield is flawed, and NASA knows it. During the first Artemis mission, in 2022, that part suffered more damage than expected as the spacecraft plummeted through the atmosphere. NASA is using the same material on this mission, but it has changed the re-entry path. The head of NASA, Jared Isaacman, says he is confident the changes are sufficient to keep the crew safe. Some former NASA officials, though, believe the risk is so great that the mission should have been postponed.
New York City is 100 days into the administration of Mayor Zohran Mamdani, who was swept into office on an ambitious platform promising to make the city more affordable. How is he doing? He has racked up a list of accomplishments — he’s gone after abusive employers and built a City Hall rest stop for delivery workers — while also retreating from some campaign promises, including his vow to give up mayoral control of public schools. We tracked Mamdani’s progress on some of his biggest ideas. Again, results are mixed: He has taken steps toward free universal child care, but his plan to offer free city buses has stalled. For more: Here are a timeline of Mamdani’s first 100 days and a video where he answers New Yorkers’ questions.
Morning readers: Save on the complete Times experience. Experience all of The Times, all in one subscription — all with this introductory offer. You’ll gain unlimited access to news and analysis, plus games, recipes, product reviews and more.
A climbing life: Jim Whittaker summited Mount Everest in 1963, the first American to do so. His longtime leadership at R.E.I. helped establish a global mountaineering craze. He died at 97. Your pick: The most clicked story in The Morning yesterday was an interview with Ben Sasse, the former Republican senator who has Stage 4 pancreatic cancer.
31— That is how many manatees have died in Florida, at least, since the start of the year after being hit by boats running too fast in places where they should be going slowly in order to not collide with manatees. Meet the people trying to save the sea cows.
W.N.B.A.: The W.N.B.A. and the N.B.A. Board of Governors approved three expansion teams. A Cleveland franchise is set to begin play in 2028, Detroit in 2029 and Philadelphia in 2030. Golf: Bryson DeChambeau had a rocky start to his 10th Masters, which has him at risk of missing the cut, but he found success with a long iron he created himself.
Here’s an impressive way to clean out the fridge on a weekend morning while providing an elegant meal in the process: a crustless quiche. Eggs and cheese bake into a beautiful custard studded with mushrooms and laced with herbs. It ought to melt in your mouth. Serve with a salad dressed on the acidic side, to cut the richness. |
