ihexx

if you were intentionally trying to doget as close as possible to a nazi symbol while still retaining enough plausible deniability, this is what you'd converge to

solution: paper towel to grab door handle, dispose in bin outside toilet

could shave

hahahahahahahaa come on now, we all know those savings are passing to shareholders not consumers

you hate to see it

allegedly. According to OpenAI.

The thing is, DeepSeek is only relevant because its R1 model was the first in the world to match openAI's o1 model on reasoning. This was at the time openAI was thought to have a huge lead there. Here's the thing: OpenAI specifically blocked every way to access the reasoning traces from o1 via the API and their apps. Deepseek invented their own algorithm to generating reasoning traces, and open sourced it, and now every AI lab uses it and can compete with OpenAI.

The claim that Deepseek copied them doesn't hold water: because HOW, when they blocked all access to reasoning specifically to prevent that?

And if the argument is: they just copied the answers, that is a lame one because there's whole public datasets built off chatgpt answers that the open source community has been using for years; that's a nothingburger story and does not explain why deepseek was so good. Remember for 6 months, not a single lab was even close to matching o1 until deepseek showed the world how. And then suddenly everyone could.

I think that OpenAI put that out there to feed into the 'china copycat' narrative to fearmonger USA legislators into harsher hobbling of the chinese AI industry via blocking GPU exports. OpenAI pushed for this, Anthropic pushed for this, Google pushed for this.

aah thanks that's kinda what I'm looking for; partially solves it, but it doesn't count as 'reading' a post if you just view it on your feed but not expand it

but who aligns the aligners of the aligners?