On DeepSeek and export controls

Post by astrange » Wed, Jan 29, 2025, 8:12 PM UTC

codekilla wrote: post #78

The author seems to be of the opinion that the creators of DeepSeek will either be unable to, or will not see the value of optimizing the 'second stage' RL component of the 'new' (post pre-training RL) way of training frontier foundation models. Every competent programmer in China is now looking for low level ptx optimizations for EVERY SINGLE STAGE of the pipeline. They will now, likely not publish any of it.

Post by astrange » Wed, Jan 29, 2025, 8:15 PM UTC

kklisura wrote: post #95

Fernicia wrote: post #74

Earlier quoted context omitted.

To say the two are remotely comparable is either insincere or naive. Like I'm not saying the US always acts utterly virtuously, but they aren't running concentration camps and chemically castrating entire ethic groups. They also aren't imprisoning people for speech, which is important when we're discussing who we want controlling AI in the future.

DeepSeek model not providing answer on the Tiananmen Square and ChatGPT providing answer on ethnic cleansing of Palestine are two sides of the same coin.

Post by verdverm » Wed, Jan 29, 2025, 8:18 PM UTC

bcrosby95 wrote: post #91

suraci wrote: post #17

Earlier quoted context omitted.

otherwise it will become to a multipolar world(or 'bipolar', said the article) that would be horrible, a multipolar world? what a nightmare

Powerful people are playing chess and the rest of us are pawns - whatever could go wrong?

Post by astrange » Wed, Jan 29, 2025, 8:20 PM UTC

CamperBob2 wrote: post #68

throwaway_32u10 wrote: post #20

I'm kind of tired of the entire narrative that "US good, everyone else bad", as if only US deserves to hold powerful technology because it will be used for "the greater good", rather than employ it in military applications, as if Anthropic didn't partner with Palantir. And it amazes me how many people can't seem to see past the "US is good, everyone else are bad" smoke mirror.

He's unlikely to be writing from a genuinely ideological point of view. He's playing a zero-sum game, in that every chip that goes to China is one that Anthropic and/or its cloud provider doesn't get its hands on. So it makes sense to him to argue for export controls using whatever rhetorical flexes and flourishes he can come up with.

Post by astrange » Wed, Jan 29, 2025, 8:57 PM UTC

CamperBob2 wrote: post #168

astrange wrote: post #167

Earlier quoted context omitted.

There's no zero sum games in a growing industry like this.

As long as the semiconductor fabs are running at capacity, yes, it's one big zero sum game. If you win one chip, I lose one chip, and vice versa. This situation is temporary, of course. China previously had a large incentive to get their own leading-edge nodes into production, and now they have a Manhattan Project-size incentive.