Anthropic, Google, OpenAI and XAI Granted Up to $200M from Defense Department - https://news.ycombinator.com/item?id=44565416 - July 2025 (130 comments)

I wonder how many politicians and political appointees understand that these LLMs are just doing word math and aren't "intelligent"?

Network and relationships supersede facts and domain knowledge in most parts of life unfortunately (or fortunately, depending on perspective).

One may be smart and logical, but can you social engineer like Sam Altman or distort reality like Musk? People are the underlying OS, everything above is fancy trimmings.

The exact same number of politicians and political appointees that understand the need to create artificial demand for these services.

Where's the difference to politicians? ;)

I agree, I suspect most don't know.

Not quite the same. SQL returns pretty predictable results and we can talk about the data.

LLMs are treated differently and often opaque and the feds seem determined to remove anyone doing critical thinking.

Let me fix the headline to remove clickbait:

US Government announces $200M Grok contract

Let me fix it; US government pays $200M for mechahitler

Funny

I don’t get the whole MechaHitler outcry, LLMs have been doing weird things for years now. Sometimes randomly this, sometimes randomly that. Even Google and Microsoft had many cases were AI gave weird responses

Yeah weird how LLMs randomly seem to reflect the politics of their owners, so random I guess we'll never know, weird.

It does seem that it is egregious enough it should have been caught in QA before being released to the public.

Everything else around it is a dopamine high from joining in on a hate train against Musk.

Grok was tweaked, it got directions to be crazier.

Grok has seemed unusually prone to bizarre behavior, even if you don't count the times that its system prompt was modified (like the infamous "please talk about white genocide" incident in May). Part of the problem seems to be that it's being trained on recent Twitter posts, resulting in a weird sort of feedback loop when users discuss Grok's behavior.

Difference being the CEOs of Google and Microsoft didn't do a Nazi salute before their bots started extolling the virtues of Hitler's ideology. Every damn time.