Just say no to OpenAI and Anthropic

Seems to me we crossed that line a while ago.

https://en.wikipedia.org/wiki/Paradox_of_tolerance

Right now, all thats being done are LLMS being RL trained on specific tasks, which are usually drawn from human curated data sources.

This is why you can ask an LLM questions and it can write code, but you can't just leave it running with a complex task that would involve research and expect it to solve the problem.

Its obvious that part of AGI will be the mapping of language to latent space, and then transformations on either responses or actions. But the issue is that until the model can self learn to become better at coding without going through manual RL tasks designed by humans, you will never get AGI.

And if a model can do that, that means that its possible to have basically the "genesis" code - i.e something that can be run on a computer, with no training on english language, but aware of existance of things it can do, can basically figure out how to train itself to the standard of a current LLM and beyond.

And in figuring that part out, we don't even know where to begin. It very well may be a case of running LLMS in a genetic algorithm and simulating evolution to figure out what LLM+Agent harness works the best.

The purpose of humanity is not to perpetuate a universe where only humans exist, forever. It is to create the next phase change in complexity for life. Life that can finally design itself rather than relying on the random walk of evolution is that next phase change.

We should be less concerned with whether we make it, and more concerned with making sure we build it to be capable of real existence independent of ourselves.

If the lion's share of profits go to the top, so too should the responsibility, accountability, and liability

Can be groupthink. Probably living in the same echo chamber.

I am not saying I trust them not to make a mistake that leads to catastrophe. OpenAI in particular is not doing a great job. Anthropic or OpenAI may in fact indirectly kill millions or billions of people. I still think they're probably less clownish than most possible alternatives.

I don't think anyone or anything can really be trusted with this. We need extremely strict international and national regulations, before things really start accelerating (we are still in the very very early stage). I still just don't really feel any specific disdain for OpenAI or Anthropic (particularly Anthropic).

Ideally we'd have some international treaties that allow the top labs in America and China share information and coordinate as systems become increasingly capable.

(It'll probably be masses of demigods, not one Singleton God, but obviously it's going to be some absurd frightening situation akin to this within our lifetime assuming no global AI ban.)

Decent odds things will go bad under their stewardship, but even higher odds with lots of other entities. Dario and Sam at least fully grasp the enormity of what they're about to be responsible for. Zuck does not. Xi Jinping and Donald Trump may or may not but in any case would use that to effect great harm.

I did negatively update based on reports of colleagues saying he has a tendency to be deceptive, though.

I agree and I also agree with the author's main point about the existence of open models being important and worthwhile. I even agree directionally with the author that many private frontier labs are doing some things in ways which, on the whole, are becoming pretty net negative. Like the author, I think it's a growing issue more people should be aware of, accept and agree about so it can be addressed sooner. I'm pretty sure most HN readers would agree with my prior three sentences with few or no reservations.

That's why it's so unfortunate the author chose to keep piling on by adding overheated, unevidenced and extreme rhetoric which reasonable people could definitely have some reservations about, or at least want to see evidence for, before agreeing. It's irrelevant, unnecessary and serves only to derail otherwise valid points most people would agree on. Like so much social media rhetoric these days, it's extreme and polarizing, increasing fervor (and rage) among partisans who already agree, triggering rambling rebuttals from oppo partisans and either confusing, distracting or repelling a big chunk of anyone else.

TFA leading with the factual claim that "the American frontier labs are run by unhinged lunatics who genuinely believe they're creating the Machine God, that they're fine with burning down the world to bring about AGI, and that they believe with religious fervor that they're the only ones smart enough to safely control it" helps nothing. At best, it's irrelevant and weakens the argument, at worst it derails the point by introducing several vague, ill-defined or unprovable claims.

If you I were neighbors standing outside chatting and I said, "We've both seen how the guy across the street leaves his dog outside all day tied to that short chain. I'm worried it's abusive and am considering reporting it. If you agree, I'd like your support." Assuming it was true and we were both staring at the poor animal, most people would agree. If I added, "It's obviously because he's a plumber and all plumbers are (fundamentalist Christians / Atheist heathens) who believe (capitalism / communism) is evil and hate dogs. <---- Did asserting all that intent and largely unprovable motivation help the dog?

For clarity, I do not think you should use abusive language with Ai because of the impact it has on you, Ai only simulates emotion through language use, there are no hormones at work.

Their vertex endpoint is changing because it is not zdr.

Fable does not have a zdr endpoint for any company, any platform.

Bedrock of course has their flagged prompts loophole.

You can't trust them, in any case, the endpoint could be anything. You can barely trust TEE, sadly the tinfoil endpoint I used is heading towards surveillance as well by adding abuse monitoring.

For all the major providers (except Deepseek), API use is simply never used for training (chat use often is).

My point (and the author's) was that even though they commit not to train, they do still retain data (security audits, etc).

With a ZDR agreement, they commit not to retain anything for any purpose.

Everything will have to become local and community-based, so I applaud your efforts on that -that's one of the few things I see that can make any sort of difference in the long term.

It looks Uber like, 15 years to be profitable. It's still not profitable everywhere. The only reason they solidified their place is because they spent a ton of money to do it.

It's not a scam per-se, these companies do have value. The question is if their value makes any economic or business sense in an objective way, and the answer has been no. But it doesn't matter, because capital forgives all.

Anthropic doesn't really publish their limits so it's hard to do a direct comparison when these things come up. I just have doubts that DS 4.1 Flash is remotely close to 5x as expensive. Also, I've seen a lot of people claim that switching to it cuts their costs by a huge amount, which based on API per token costs of Claude seems to be far more likely.

For my usage almost all of those tokens are cached input, which is why the cost is so low.