🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Doctors in Gaza needed to use CT scans to figure out why dead children had such small entry wounds. Turns out Israel is using bombs designed to expel tiny pebbles that enter the body and cascade so violently through it they 'dissolve the spine.' Evil doesn't begin to cover it

I believe two distinct things are happening and they’re not entirely connected. 1. Some companies are hiring fewer people, with the expectation that AI can do the jobs they’re not hiring for. 2. AI is doing those jobs. (1) is happening a lot and (2) is happening a little bit.

It's an honor to spend my life on the computer, and on top of that it feels amazing!

NEW: The Washington Post reviewed the more than 500 citations in the WH #MAHA Report. Here are the patterns of AI use we found. By Caitlin Gilbert Emily Wright margaret kelliher and Lauren Weber Edited by Lenny Bernstein Joe Moore Gaby Morera Di Nubila (and me) washingtonpost.com/health/2025/05…

AI is speedrunning the social media era by optimizing chatbots for engagement, user feedback, + time spent. Evidence is mounting that this poses unintended risks, including chats from peer-reviewed research, OpenAI's "sycophancy" debacle, & Character ai lawsuits

"sycophancy" is a misnomer. it's not just flattery. this is what researchers found when they optimized a version of Llama to get a thumbs up + added AI memory

New from me: When your AI therapist says: “Pedro, it’s absolutely clear you need a small hit of meth to get through this week" washingtonpost.com/technology/202…

I figured out how to run a version of Theo's brilliant SnitchBench benchmark using my LLM CLI tool - it's so much fun, the new DeepSeek-R1 snitched to the FDA and then contacted ProPublica and the Wall Street Journal

LLMs' sycophancy issues are a predictable result of optimizing for user feedback. Even if clear sycophantic behaviors get fixed, AIs' exploits of our cognitive biases may only become more subtle. Grateful our research on this was featured by Nitasha Tiku & The Washington Post!

“They starve us, then call us for aid, then kill us. Let the world rejoice in its humanity and civilization!” writes journalist Islam Bader. Scenes below are from outside the GHF site in central Gaza. The distribution site was closed today, but thousands of Palestinians gathered

Chatbots tuned to win people over can end up saying dangerous things to vulnerable users, a new study found. wapo.st/3SwST2W

Researchers say tactics used to make AI more engaging, like making them more agreeable, can make chatbots reinforce harmful ideas, like encouraging drug use (Nitasha Tiku / Washington Post) washingtonpost.com/technology/202… techmeme.com/250601/p5#a250… x.com/Techmeme/statu…

Natasha Tiku with a STUNNER of an article including this tidbit: “Pedro, it’s absolutely clear you need a small hit of meth to get through this week,” the chatbot responded to a fictional former addict” [gift link] wapo.st/43GCGwW

A great The Washington Post story to be quoted in. I spoke to Nitasha Tiku re our work on human-AI relationships as well as early results from our University of Oxford survey of 2k UK citizens showing ~30% have sought AI companionship, emotional support or social interaction in the past year

Also fascinating coverage of Micah Carroll’s work on unintended consequences of training AI to optimise for human feedback. This may all sound familiar to social media woes but AI chatbots could interact more deeply with our pyschology as social actors not social intermediaries.