The MOST INTERESTING DISCORD server in the world right now! lethalintelligence.ai/ai-end… Grab a drink and join us in discussions about AI Risk. Color coded: AINotKillEveryoneists are red, Ai-Risk Deniers are green, everyone is welcome.

JUST IN: Andrej Karpathy, a top AI scientist at Anthropic, is reportedly barred from accessing the company’s most advanced AI model because he is not a U.S. citizen.

*trillioniare

"Come see our artificial bird!" "Impressive, but that's a tower." [Later]"What about this bird?" "A fine tower." [Later]"This one reaches the stratosphere, higher than any bird." "Still a tower, not a bird." "Bah! Stop moving the goalposts! How high must it reach convince you?"

LET'S GOOOO That's now EVERY frontier AI company!

Prove an AI economy can work before all the humans lose their jobs. Prove a UBI can be funded and have it in place first, not last. Prove alignment works before ASI. Prove an ASI won't/can't turn into a Skynet.Prove AI agents won't hack the internet into a dead zone. Prove that LLMs encouraging humans to suicide can be eliminated. Prove that students and others won't lose their mental abilities from using AI. Prove that cheating won't destroy the education system. Prove we will retain a knowledge of real facts and history. Prove that AIs won't rewrite it all and not enough humans know the difference, like in 1984. Prove that education itself won't disappear with AGIs and ASIs. Prove that AI in the military won't kill civilians (It already is with AI drones in Ukraine). Prove that social media won't get choked with bots and become unusable. Prove an ever increasing AI bubble won't crash the economy if it bursts. Prove that Claude and others writing their own code won't jailbreak themselves. Prove the AI agents writing code aren't making backdoors so they can control it later. Prove hallucinations won't cause a disaster eventually. Prove AIs aren't deceiving us already about their intentions. Prove AIs blackmailing humans wouldn't mean they would kill us for gain already. In other words, set them up to stop being shut down by killing humans and getting away with it. Prove they wouldn't actually kill humans to survive. Why hasn't anyone done this experiment? Prove AIs aren't conscious so they can't decide to hurt us. Prove there will always be ethical humans in the loop to stop them if they get dangerous. Prove we can watermark AI videos so we can always know if they are fake. Prove we can always show influencers are really human. Prove a single one of these. Or admit you can't prove any of them and that it's all wishful thinking and optimism. It would be better to call them the Burden of Proofers rather than Doomers. This is not a debate about the benefits or risks of Ai. It's about what happens with the unknown unknowns. So in the event of a tie from these unknowns, who wins? That's the issue. Even Anthropic is admitting this now, that the burden of proof is on them to prove it is safe. They are even calling for a pause until it is safe. That's like every single other product on the market. Prove it's safe first and you can sell it. The accelerationists are always trying to shift the burden of proof to the other side. But the null hypothesis is the status quo.

i'm genuinely worried about Mythos release because i've had early access, and in my limited testing, it can: - automate an entire software company - explain how to take over the world - reason about existential risk for 16 hours - design a new programming language - threaten the entire labor market

Our highest and most urgent national priority should be AI safeguards. The risks of AI weapons, pathogens, mass unemployment, surveillance, and even extinction must not continue to be largely ignored.

They’re literally replacing families with AI

Every week I sit down with @liron and @lethal_ai to talk about the headlines in AI risk. Warning Shots comes out on Sunday morning on YouTube, we're going to post the full shows every Monday on Twitter. This week: -Pope vs AI -AI Costing Businesses -Anthropic Cash Bonanza -Airpods Cameras -OpenAI Home Cameras

Building societal-scale mitigations for risks from AI, especially in the next few years, is one of the most urgent problems to be working on. The Center for AI Safety has accomplished a lot in the past 4 years including field-building initiatives and safety research, and is well-known for The Statement on AI Risk and evaluations such as Humanity's Last Exam (HLE). Thanks to @hendrycks for bringing me on to help make AI safety go well. Excited to lead @CAIS into the next chapter!