Hacker Newsnew | past | comments | ask | show | jobs | submit | lifeformed's commentslogin

Some reason this kind of argument always by default assumes the US will be the good guys.

I didn’t say that. I said the US is ahead, and should consider its strategic adversaries and their behavior.

Hoping everyone chooses to do the right thing is not an effective strategy for mitigating bad things.

This assumes that children's learning is solely about correctness and obtaining information. There is a whole universe of other considerations, like interacting with other people, with adults, with fallible authority figures. Having a conversation with a teacher and getting the wrong answer teaches you far more about being a human being and living in a human society, than getting facts back from a screen. Yes, teachers being incorrect is something that should be improved, but reducing human interaction is not an improvement.


Is it actually that hard to make good models or is it just about the amount of resources you have to do training? (This is an actual question, I really don't know.) I'm sure it's not trivial but does it really take world class secret knowledge to build off of the known existing techniques? I feel like there's tons of low hanging fruit still to explore, and time and resources are the limiting factor.


The gap between grok and Gemini to Claude and chatgpt suggests that yes it is that hard.


All of the 11 grok co-founders alongside Elon quit: https://techcrunch.com/2026/03/28/elon-musks-last-co-founder... so that will have hampered grok.

Like Zuckerberg, top talent may not work with a polarising character if they disagree with his behaviour. Space focused talent don't have many choices aside from SpaceX but ai companies are a plenty and a top AI person can pick and choose.


The fact that you need top talent also suggests that it is indeed that hard


I dunno, if most of the top of a company quits it's extremely disruptive even if everyone else in the company is competent.


I suspect that Grok has been ironically lobotomized by pressures to correct its political views.

Similarly, I could imagine the Gemini folks working in a significantly more complex corporate climate, with different parts of Google pushing for different capability focuses. They are only lagging behind less than a year, so it isn't too large of a gap yet.

That said, the fact that Anthropic is currently the top dog suggests that talent and execution is incredibly important. A year ago none of my normie friends new them, and when i suggested using Claude looked at me like when I recommend Linux.


That shouldn’t affect Grok’ coding ability. How often are people discussing politics with Claude code? Writing decent code is just hard and it’s not just Grok.


It affects their ability to hire and retain talent.


If training a good model requires talent then that’s the answer to the question this thread is trying to answer: is training a good model actually that hard?


Talent to do.. what? This could mean a lot of things.

Navigating astronomically huge fundamentally not so hard but still really tangly and hairy projects requiring both excellent short- and long-term vision in an overheated domain with angry people and lots of money is a skill all of its own.


Talent to train high quality LLMs, especially coding LLMs.


Why would these be independent?


More specifically, political lobotomy shouldn’t affect coding ability.


You’d be quite surprised, I think. Fine tuning a model on one axis can have drastic impacts on another that as a human we would expect to be completely unrelated.


I have never seen anyone argue that this cannot be overcome with more high quality RLVR data.

The practical reality is that the Chinese and American models might have very different politics. But the most relevant factor in model performance is the quality and volume of training data, not ideology of the base model. Unless you are suggesting something very particular about the way Grok was neutered.


It's all a bunch of weights isn't it? Why wouldn't fiddling with some parts of the weights have cascading effects?


Yet empirically it does


Not true, aggressive post training makes models notably dumber.


> That shouldn’t affect Grok’ coding ability.

If you are spending all your time having to re-train because the boss doesn;t like the output, it will hamper coding


> A year ago none of my normie friends new them, and when i suggested using Claude looked at me like when I recommend Linux.

Isn’t that still the case? Normies haven’t even heard about Claude, in my experience.


In my experience it has improved a bit, but 90% of my normies still have no idea. (It was 100% before)


Not hard to be a fast follower. Lots of companies are ~6-9 months behind. Reaching the actual bleeding edge is much harder.


>Is it actually that hard to make good models

Didn't take DeepSeek long. Or XAI to launch grok.

If they have a top team and the money then appears to be a matter of a year or two? And one startup mentioned is Japanese not Chinese so they won't be banned from buying US tech.


Maybe a really lightweight fast LLM could moderate messages in realtime. Not sure how pricey that would get though.


No matter how cheap per request, someone will figure out a way to DoS that endpoint, and it will be extremely pricy unless you have effective rate limiting.


The OpenAI moderation endpoint is free and it doesn't count toward the usage limit.


Extremely if not behind substantially cheaper anti spam measures.

Further it may still make sense to use human reports to gate some automation even if it slows response.


Well I'm sure someone will tell it to.


But a messaging program like Signal is fairly similar in scope, but only has like a couple dozen of devs, compared to Messenger's thousands(?).


Signal could definitely use a bunch more devs, if only to fix all the UX bugs I hit on a daily basis.


You haven't used messenger much then, if you're comparing the complexity of the two.


I guess I don't understand which are the features Messenger has that Signal doesn't that requires a billion dollars a year to maintain.


I don't use it. What am I missing?


The truth is that if you get in early enough on a hype train and cash out in time, you will make money. That's enough of a rational basis to participate. The ostensible purpose of it all is basically irrelevant, except as a signal to participate.


It takes away from real human artists who do their part to slowly advance human culture. Music will not develop without human artists. Maybe for this moment in time AI can fulfill some people's musical desires, but it's not going to keep up with the times. The point of art, in a general sense, is humanity. Automating away your artistic needs is like automating away your social needs. It's a one way "relationship" that is superficial and self-indulgent. It's a step towards an empty world.


Why would music not develop without human artists? This isn't true at all, "AI" isn't necessary LLM as well, there is plenty of ways for AI to innovate, and let's be real, most musics from humans are a bit of copy-cat nowadays, ton of AI music actually made me vibe personally and stuff I haven't heard before.

Have you tried a day of listening solely to AI music? I feel you might change your mind, sure sometimes there is some serious off-tune (feels like an hallucination from the model) but we know this is temporary.

PS: I'm conscious of what it does to humanity, but there is also facts that AI does produce great songs, that's 2 different discussions.


That's the creator's perspective. From a listener's perspective, it's "do I enjoy it" or "do I not enjoy it". Everything else is intellectualization.


But that means nothing. There's no raw "enjoy", except maybe drugs, and I have my doubts about that.


> There's no raw "enjoy"

What do you mean?


Painters said the same thing about cameras


And turns out there is still room to enjoy both photography and paintings as their own art forms.


I think this article is agreeing with you. It's saying that it's not an "intelligence" failure in the traditional "military intelligence" sense of the word we expect, but rather it's an "intelligence" failure in "runaway idiocy" sense.


then they should have called it a "mental faculties" failure.

this is an attempt to make it look like the CIA or NSA fucked up, when we know damn well they knew Iran had been prepping to fight the US for decades.

this is a Presidential failure, or more likely, a Presidential Advisor failure. and "failure" in the sense that Stephen Miller or Steve Bannon have been pushing for this since day 1.

there is a reason no president ever decided to take a go at them until now, and we're seeing why


Read the last line of the article again.


burying a statement in the last line is disingenuous CYA at best and explicit disinformation at worse.

they could have put it in the headline, or first paragraph.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: