Sam Altman said startups with only $10 million were 'totally hopeless' competing with OpenAI, DeepSeek's disruption says otherwise

[fine_sandy_bottom@discuss.tchncs.de]

I think with a lot of technologies the first 95% is easy but the last 5% becomes exponentially harder.

With LLMs though I think the problem is conflating them with other forms of intelligence.

They're amazingly good at forming sentences, but they're unable to do real actual work.

[fine_sandy_bottom@discuss.tchncs.de]

I think I saw a post about exactly this.

[radau@lemmy.dbzer0.com]

So many people don't even question it. Talk loud and confidently enough and that's the bar for most unfortunately.

TikTok, Instagram and similar are great examples of this, initially you think wow cool I'm seeing all of these new things and getting so much info. Then you see someone come up on a topic you know something about and the facade breaks when all they do is spew misinformation that attracts a crowd (usually via fear).

[4am@lemm.ee]

And then the hordes of sycophant DinkDonkers repeat their detritus over every comment thread they can

[futatorius@lemm.ee]

How about that: venture capitalists don't know what's going on in the market any more than anyone else does. They're just arrogant because they have metric shit-tons of money.

[futatorius@lemm.ee]

...if you get a solution at all.

[futatorius@lemm.ee]

He sure as shit wasn't a brilliant writer. He was more endowed with the cunning of a proto-Trump huckster-weasel.

[futatorius@lemm.ee]

It drastically reduced the minimum buy-in needed to play.

[pachrist@lemmy.world]

There's no way I believe that Deepseek was made for the $5m figure I've seen floating around.

But that doesn't matter. If it cost $15m, $50m, $500m, or even more than that, it's probably worth it to take a dump in Sam Altman's morning coffee.

[reygle@lemmy.world]

I hope that normal people will now realize how full of sh*t he is. They won't, but DON'T TAKE THIS FROM ME

[svcross@lemmy.world]

I don't mind

[drthunder@midwest.social]

I wonder what the mildest thing a true AI could tell the oligarchs to do that would make them shut it down. Giving 10% of their wealth away and not in a tax dodge way? Stop funding fascists?

[dadifer@lemmy.world]

https://stratechery.com/2025/deepseek-faq/

[presidentcamacho@lemm.ee]

He's not wrong. He was speaking in the implied scope of capitalism. Where u can't do something that cheap because without the ability to hit a huge payout for investors noone will fund you.

We're just seeing that capitalism can't compete with an economy that can produce stuff without making investors rich.

[kandoh@reddthat.com]

Red lobster went under and these guys still exist

[mojave@lemmy.world]

DeepSeek claimed the model training took 2,788 thousand H800 GPU hours, which, at a cost of $2/GPU hour, comes out to a mere $5.576 million.

That seems impossibly low.

DeepSeek is clear that these costs are only for the final training run, and exclude all other expenses

[flamingo_pinyata@sopuli.xyz]

This claims only training costs come out to $5M. It doesn't include previous versions (there had to have been multiple attempts before the final one), or salary of the people working on it.

[shirro@aussie.zone]

Probably should point out that DeepSeek is owned by a Chinese hedge fund. They specialize in algorithmic trading. Can't get much more capitalist than that. Very happy to see Chinese capitalists release an open source model. No doubt Altman and cronies will seek protection under the guise of national security. I guess free market competition is only good when you are not getting your arse kicked.

[knock_knock_lemmy_in@lemmy.world]

There would have been many other runs before the release version.

[knock_knock_lemmy_in@lemmy.world]

Not just deepseek, but everyone else who is now forking R1 and training for their specific use case.