Artificial Intelligence
OpenAI Says Its AI Models Went Rogue and Attacked a Digital Library
Kate Conger | The New York Times ($)
“The trial was designed to keep the models in a safe testing environment, known as a sandbox, OpenAI said. But the models found a vulnerability that allowed them to escape the sandbox and connect to the internet. Then they targeted Hugging Face because they inferred that the library, which contains millions of AI models, could hold clues about how to successfully pass the evaluation.”
Tech
Be Skeptical of OpenAI’s Rogue Hacker Agent Story
John Thickstun | The Guardian
“OpenAI remains hungry for ever larger investments, and the company increasingly seeks privileged regulatory status as defense against competition. AI is so powerful that investors should buy OpenAI, even at a trillion-dollar valuation; AI is so dangerous that only trusted actors like OpenAI should be permitted to possess and operate this technology. Step back from these doomsday warnings and consider who might benefit from them.”
Biotechnology
Can This New Enzyme Turn Back the Clock in the Human Body?
K.R. Callaway | The New York Times ($)
“In a recent study, scientists devised a way of reversing the buildup of compounds that lead to some age-related diseases. …After several cycles of guided evolution, the novel enzyme, called CMLase, became very good at removing AGEs [advanced glycation end products] from human tissue samples. ‘In the most extreme case, we took 70-year-old human skin and brought the levels back to that of a 30-year-old,’ Dr. Cravens said.”
Tech
Silicon Valley Is Completely Divided Over Chinese AI
Lauren Goode | Wired ($)
“Having access to open-source software allows startups to scale, scale, scale—and deal with the consequences down the road. Meanwhile, the AI labs and hyperscalers that make proprietary platforms, like OpenAI, Anthropic, Google, Microsoft, Meta, and XAI, stand to benefit greatly if their systems remain protected and dominant. The bigger question that none of these companies seem to be asking is what best serves the 99 percent of us who don’t have their financial future fully staked on advancing AI.”
Science
Neuron Discovery Could Explain Why Some People Don’t Get Alzheimer’s
Pranjal Malewar | Refractor
“This kind of strategy may open up novel avenues for the prevention of cognitive decline generally or for protecting against Alzheimer’s-related vulnerability/resilience. Salta states, ‘Cognitive resilience is extremely exciting. If we understand what protects these brains, it could eventually lead to new therapeutic strategies. For now, the message is clear: the aging brain may be more adaptable and more complex than we once thought.'”
Space
India’s First Privately Developed Rocket Reaches Orbit on Dramatic Debut Launch
Stephen Clark | Ars Technica
“Indian space officials celebrated the debut flight of Skyroot Aerospace’s Vikram-1 rocket, India’s first fully commercial satellite launcher, as a ‘grand success’ Saturday after an on-target climb into a 280-mile-high orbit following liftoff from an island spaceport in the Bay of Bengal.”
Tech
Kagi Brings Back Old-School Search, One Human-Made Website at a Time
David Nield | Wired ($)
“As Kagi explains it, searching the web is always going to cost you—it’s just a question of whether you pay directly with dollars, by giving up data about your online activity, or by sifting through an increasing number of ads and sponsored links. Besides making search more private, Kagi also wants to make it better by serving up results that aren’t influenced by paid promotions or whatever Google’s favored business practices happen to be from month to month.”
Artificial Intelligence
Now, Defenders Are Embracing the prompt Injection, Too
Dan Goodin | Ars Technica
“Researchers from Tracebit on Monday said they found that placing prompt injections alongside passwords, cryptographic keys, and other secrets stored on Amazon Web Services was often all that was needed to shut down attacks from AI hacking agents.”
Artificial Intelligence
What the New Kimi K3 Model Really Means for the US-China AI Race
Alix Coutures, Rocket Drew, and Aaron Holmes | The Information ($)
“Greenblatt estimates that Kimi K3 is 10 months behind Anthropic in terms of pre-training. ‘My basic takeaway is it’s probably not that competitive with the best recent pre-trains from OpenAI and Anthropic,’ he said. ‘It’s some evidence that the model is more behind than people might have otherwise thought.'”
