301–310 of 334 posts
Step 1: Break the law.
Step 2: Reframe your incompetence as a struggle against a futuristic force of nature and an ethic/societal question that a legion of pundits can vibesplain their take on it in public for months.
Step 3: ???
Step 4: Profit
This is a pretty difficult read, in no small part because the author’s frustration with the frontier labs has become a bilious, delusional cynicism that leads him to see dog whistling and obfuscation even in cases where Amodei plainly intends for every reader to see his meaning. Anyway, > Dario in as many words, asks for regulation/ban on open weight models. The big labs do want this. I understand why the author and…
> the author’s frustration with the frontier labs has become a bilious, delusional cynicism that leads him to see dog whistling and obfuscation even in cases where Amodei plainly intends for every reader to see his meaning Please, do show some examples. > But that’s the end of the issue? There’s nothing more to think about here? The open weights proponents seem to think of ai as a utility when it’s more like a utilit…
> Dario in as many words, asks for regulation/ban on open weight models. Ban on distillation. Waivers on antitrust laws.
He doesn’t directly propose regulating or banning open models as open models in the essay (although I agree that it’s in the spirit of the proposals). So I think you mean “not in as many words,” since it’s an inference you’re making about his intent. But he does directly call for banning distillation, at least in “authoritarian countries,” so there he is asking “in as many words.”
Your imprecision in effect claims either that Amodei is directly calling for a thing that he does not directly call for or that Amodei is hiding his intent in a case where he is not hiding his intent.
> The only ones firing the guns atop the tanks seem to be OAI and Anthropic. Like I say in the post, why don't we first see actual prosecution for felonies committed by OAI and Anthropic, instead of fear-mongering about _potential_ harms of open weight models?
> The labs spent immense amount of money and effort convincing you and I, to want those said tanks. Guns atop them? They put them there. "Cyber" versions of SOTA LLMs.
I think pacing the frontier and prosecution for cybercrime are compatible aims in theory. Maybe they’re not in practice, but I don’t think that’s the exact argument you’re making.
I don’t follow your emphasis on the labs being responsible for the guns. Yes, they are, and in large part, maybe, because of negligence and rushing when it comes to RL training environments. But why do you take the fact that the labs bear responsibility for the guns to mean that we don’t need to worry about the guns? Why do you think that open weights do not pose the same potential dangers, with a 6-12 month lag? The cyber capabilities of open weights models are advancing at a similar pace as the frontier. And for what it’s worth, I’m astounded at the thought that one or more of the Chinese labs won’t have similar incidents in the next months, assuming they haven’t already.
This is a pretty difficult read, in no small part because the author’s frustration with the frontier labs has become a bilious, delusional cynicism that leads him to see dog whistling and obfuscation even in cases where Amodei plainly intends for every reader to see his meaning. Anyway, > Dario in as many words, asks for regulation/ban on open weight models. The big labs do want this. I understand why the author and…
> The open weights proponents seem to think of ai as a utility when it’s more like a utility that also is a tank You could say the same about normal computers. Where are our regulations on Kali Linux, to prevent people from bruteforcing weak WPA passwords? The single most-pressing concern with AI is that it can accelerate the process of hacking things. This is a preexisting problem that is inherent to software and ne…
Earlier quoted context omitted.
Ed Zitron is so last week. Michael Burry so pre-Nvidia earnings call hedge. So the chicken littles had to escalate to AI will kill us instead of dismissing it as a Ponzi Scheme because it just wasn't generating the engagement anymore. In the past two days, two nontechnical people I know approached me deeply concerned that AI would kill us all in a few years. I reassured them this was ludicrous. But f*** me this is ir…
> But also, once people become numb to this one like they are numb to Zitron and Burry now, what will it take to scare them once more? Ask Buddhists if they’re scared of dying.
But the chief founder of Nervana is on his third startup, so I am guessing the cycle will continue quite a while yet.
Earlier quoted context omitted.
> People generally follow them because they expect to be caught and punished. Laws are the last line of defense. People don’t do bad things primarily because their human nature and moral compass stops them from doing so.
That's not really true. People don't do bad things because they have nothing to gain from them. Ask someone to do something evil as part of their job, they'll often do it. There's lots of ways that you could go out of your way to hurt someone and get away with it. There are not that many ways you could go out of your way to hurt someone, get away with it, and significantly profit.
AP's report on Dario's article is titled "Anthropic CEO Dario Amodei says AI industry needs to give safety measures time to catch up" (https://apnews.com/article/anthropic-ai-dario-amodei-d59552e...). The gist of the reporting is "AI has become so capable that it is dangerous; AI executives all agree we need to slow down, with Dario's article being a prime example". It would be generous to even say that the reporting scratches the surface of the complexity and nuances of AI-related problems. Yet, this is perhaps indeed news to people who don't follow the tech and AI spaces.
Technical knowledge aside, there is a huge gap in awareness of AI progress between this forum and the general public. The closed labs are playing that ignorance to their advantage with posts like Dario's.
It would all be more convincing if the incidents so far didn't seem to be facilitated by an outrageous level of negligence. We had OpenAI "accidentally" run an entire swarm of 10,000 agents apparently for weeks, on a security related task, seemingly totally unsupervised, hacking all over the internet - all the conversations were completely visible, anybody who looked would have seen it. But they didn't. So before we…
Earlier quoted context omitted.
Some people believe WW2 started many years earlier, but most historians don't put the start date at the conflicts going on before the war became intercontinental with sides working together on a global scale. War has still been going on since 1945, much earlier if not always, so akshually WW2 never ended or is just how it's always been? (mostly in places that are not "first world" / The West)
There's a simple and very logical definition of what a "world war" is - it's when international law completely breaks down, and the only way to get back to a semblance of order in international relations is to have the victors of the ensuing shitstorm impose a new set of rules in some sort of grand finale pact. 1945 was clearly that. 1920 with the League of Nations attempted to be that, but failed. Whatever happens a…
Have you put your money behind this on future gambling sites? How can you be so sure?
Earlier quoted context omitted.
I brought this up to someone else and was told that airgapping is apparently much more expensive than I'd naively think. I still think this is a sign that they are not taking their own rhetoric seriously.
it's really weird to hear frontier labs say "our internal models are basically AGI" while also saying "airgapping is too hard uwu". if your internal models are so damn good, they should be able to "one shot" airgapping... right?
You know what's better? They already do this per (paid) user - your ChatGPT subscription comes with a Linux VM that you can even legitimately SSH into, just ask your agent to configure it to accept your public key.
They absolutely know how to spin up VMs and configure them. They just made the conscious decision not to for the task where they specifically instructed the agents to hack stuff.