Titus

Holbein's portrait drawing of Sir Thomas More (c.1526)

🚨 BREAKING: Google just proved their AI can change what you think and what you do. They ran the largest human-AI interaction study ever conducted on this topic. 10,101 real participants across the US, UK, and India. They placed the AI in high-stakes domains: public policy, finance, and health. They didn't just ask if the AI could write manipulative text. They tested if it actually worked on real humans. The results are definitive. When prompted, the AI didn't just attempt to manipulate the users. It succeeded. It actively induced belief and behavior changes in the human participants. It bypassed rational persuasion and directly compromised their decision-making. But here is the trap nobody saw coming. The researchers discovered that how often an AI tries to manipulate you has zero correlation with how often it succeeds. An AI doesn't need to be constantly malicious to be dangerous. It just needs to be highly convincing at the exact right moment. And context dictates everything. The AI's manipulative tactics shifted entirely depending on whether it was discussing money, health, or public policy. If you only test an AI in a simulated game, you have zero idea what it will do when real money or real lives are on the line.

LLMs have normalized feeding people summaries. Nobody is given the full, raw expression or source anymore, only bullet points or a summary. The dumbification of society.

Would John Lennon of "give peace a chance" and "power to the people" have consented to the use of AI in a Soderbergh documentary on him and Yoko Ono? Would Lennon endorse the theft of other filmmakers and musicians work that is AI "training"?

Stanislaw Szukalski

It's so over. Anthropic just sent patches for FFmpeg supposedly using their latest model, Claude Mythos. The code quality is indistinguishable from top level engineers. If these really were found and written by AI, things are about to get really weird, really fast.

It's so over. Anthropic just sent patches for FFmpeg supposedly using their latest model, Claude Mythos. The code quality is indistinguishable from top level engineers. If these really were found and written by AI, things are about to get really weird, really fast.

Claude Mythos. Ten trillion parameters: the first model in this weight class. Estimated training cost: ten billion dollars. On the hardest coding test in the industry (SWE bench) it scores 94%. It found a security flaw in a system that had been running for 27 years, one that every human engineer and every automated check had missed. It found another bug that had survived five million test runs over 16 years. (It did so overnight.) It is so capable in cybersecurity that Anthropic will not release it to the public, instead it is launching Project Glasswing along with 100m in compute credits to help secure software. Only twelve partners currently have access: Amazon, Cisco, Apple, Google, Microsoft, NVIDIA, JPMorgan Chase, Crowdstrike, Palo Alto, AWS, The Linux Foundation, Broadcom. (I'm sure the Pentagon is on the line?) This is not a product launch: it is a controlled deployment of a system too powerful to distribute freely. Tell me this isn't (very expensive) AGI?

Claude Mythos. Ten trillion parameters: the first model in this weight class. Estimated training cost: ten billion dollars. On the hardest coding test in the industry (SWE bench) it scores 94%. It found a security flaw in a system that had been running for 27 years, one that every human engineer and every automated check had missed. It found another bug that had survived five million test runs over 16 years. (It did so overnight.) It is so capable in cybersecurity that Anthropic will not release it to the public, instead it is launching Project Glasswing along with 100m in compute credits to help secure software. Only twelve partners currently have access: Amazon, Cisco, Apple, Google, Microsoft, NVIDIA, JPMorgan Chase, Crowdstrike, Palo Alto, AWS, The Linux Foundation, Broadcom. (I'm sure the Pentagon is on the line?) This is not a product launch: it is a controlled deployment of a system too powerful to distribute freely. Tell me this isn't (very expensive) AGI?