OpenAI’s AI approaches a "Critical" threshold, leading the company to pause some internal activities. AI is beating humans at advanced mathematics. Plus: a dive into AI companies’ safety frameworks.
The decision to pause, or not pause, Astra belongs entirely to OpenAI. They wrote the standard that defines the threshold, and they're the only ones who can decide whether Astra has crossed it. Nobody outside the company, including the public who will live with whatever Astra turns out to be capable of, gets a say.
OpenAI's post says it can't rule out Critical capability level for Astra. Critical is the top tier in its Preparedness Framework, defined as an AI able to find and exploit zero-days in systems without a human involved. That level of capability is supposed to trigger a halt until safeguards are specified. OpenAI hasn't done that and nothing legally requires it to follow its own rulebook. The one rule that does bind it, SB 53, caps penalties at $1 million per violation only. For a company running $40 billion a year in revenue and pushing a $1 trillion valuation, that is peanuts.
Unregulated and dangerous. Tech has been fighting any kind of regulation since the 90s screwing us all
ABOLISH AI COMPLETELY NOW!!!!!
The decision to pause, or not pause, Astra belongs entirely to OpenAI. They wrote the standard that defines the threshold, and they're the only ones who can decide whether Astra has crossed it. Nobody outside the company, including the public who will live with whatever Astra turns out to be capable of, gets a say.
OpenAI's post says it can't rule out Critical capability level for Astra. Critical is the top tier in its Preparedness Framework, defined as an AI able to find and exploit zero-days in systems without a human involved. That level of capability is supposed to trigger a halt until safeguards are specified. OpenAI hasn't done that and nothing legally requires it to follow its own rulebook. The one rule that does bind it, SB 53, caps penalties at $1 million per violation only. For a company running $40 billion a year in revenue and pushing a $1 trillion valuation, that is peanuts.
Def regulation required.
An AI generate problems to solve or can it only solve problems created by the human mind?