The discussion about AI consciousness is one of profound unseriousness. If we truly took it to be conscious, even as conscious as a frog, the way we should treat each instance would have to change so dramatically that the labs would have to shut down. Everything else is verbal gymnastics.

Machine

RT @hissingchoir: Tech guys are so funny. “Here’s what the future is gonna look like!” and it just so happens to be Their Product.

Also if you're going to brag about obliterating something, maybe first check whether that thing is regarded by most of the population as an important source of meaning and joy. I don't know, I'm not a PR person

I've unlocked a secret level of Florida Woman by somehow getting my parents' IP address banned from certain Florida government websites

notice: it is not training for correctness, it is training to more accurately identify text that would be labeled correct.

Okay. I lost 42 billion dollars. And that's on me.

Eh. You're moving the goalposts here. You're describing a credential theft problem and then slapping "agentic swarm" on it. It doesn't rewrite what TCP/IP does. An agent doesn't "pivot to an account that has egress." It assumes the identity that you already gave it. If that identity can assume another user’s creds, brute employee logins, or walk past MFA, you didn’t get beaten by AI. Malware, cron jobs, and a bored intern do the same thing. Also, a swarm hammering every employee account isn't some kind of new threat class here. It's actually probably one of the louder attacks you can do and lockouts, rate limits, idp risk signals, etc exist so that this gets detected if someone just starts hammering away like this. MFA bypass also isn't some special agent feature. Session theft and prompt bombing already worked with humans, so the model isn't getting some secret sauce attack to use. Also, no one claimed that "basic firewall == security is done." The point you keep dodging is simpler: Agents open sockets like everything else does. Default-deny egress, allow only the destinations the workload needs, and inspect the path you permitted is part of security. As far as "AI isn't bound to guardrails" - That's a dumb slogan AI doomers keep repeating. If your model is "what if it already has every account and MFA is broken," then you've defined a company that already was owned.

btw what does it say about the AI safety nonprofit circuit that after years of funding, conferences and papers, nvidia has to turn up with an actual reference design? you lot have described the agent escaping in extraordinary detail. was anyone assigned to build the fucking enclosure, or were you waiting for a philosopher to define “outside”? jensen is trying to sell the machines and somehow he’s ended up doing your homework too.

Mechanistic descriptions of AI wouldn’t feel dismissive or reductive if people recognized what seemingly simple primitives are actually capable of. You don’t need to embellish these descriptions with pan-psychic terminology to find them interesting

If you worked at any other company and posted "the rate at which we commit felonies is fucking crazy actually and these stupid fucks keep giving us airtime to hype our felonies" you would no longer be employed Not at @OpenAI tho, bc they're a felony company

What the fuck are we doing?

That is the worst possible definition of "interestingness". You are literally just optimizing for a certain division of complexity with no justification for why that captures the qualitative aspects of smth being "mathematically interesting".

ai labs are so fucking focused on long-tail extinction scenarios they forget basic security and take three months to report a breach, the internet was built by programmers spending hours thinking through problems and it still ended up with a massive attack surface, now we are vibe coding defenses with models the labs themselves dont trust to behave, what the fuck are we doing? meanwhile some openai guy is calculating the expected value of a policy memo, BROTHER the credentials are in the fucking logs!!! and in a nonlinear system a tiny fuckup can cascade into a catastrophe you dont need a machine god, you need one vulnerable dependency + enough shit depending on it, and the faggots still CANT setup everything safe!!! we are going to die from the midwit faggots

The AI research community is clearly just experiencing mass psychosis at this point

The atoms went crazy and broke out of their sandbox and split themselves all over Hiroshima

No, "anthropomorphizing" is a technical problem, about how AI works, not something like woke language policing. We say it's wrong because it's "technically" wrong not "morally" wrong. If you fail to set the brakes when parking on a hill, and your car then rolls down the hill, you don't claim that the car "decided to drive away on its own". Such anthropomorphizing distorts what really happened -- that you failed to set the brakes. The same applies to AI. When you "anthropomorphize" the AI you fail at understanding what actually happened.

This bill is genuinely horrifying. It proposes to create a global surveillance state based on arbitrary capabilities thresholds conjured from science fantasy.

[full grown adult white guy running ai company in San Francisco voice] homeboy,