Lmk when cancer is cured. This s*** is a waste of time til then
Cancer vaccines soon. Cure may not happen for a long time. But I'll take every advancement.
Definitely doesnt add up how cheap they produced this for, theres no f***ing way they made this for less than a 2 bedroom condo in new york city lol
And then they’ll realize how totalitarian it is over there and run back to America
chinese american scientists were being harrassed by american authorities and returned to china lmao ur such an idiot stop talking about china
Mfs don’t realize they’ll get thrown into labor camps if they talk s*** about Xi the way they do Trump
u don’t know anything about china everything u say about em is just a reflection of ur poorly formed ideology
China has researched and collected data much longer than USA
If China don’t beat US in AI race, it’s embarrassing
No they haven’t lmao
Like what are you even talking about and what are you basing this on
No they haven’t lmao
Like what are you even talking about and what are you basing this on
No they haven’t lmao
Like what are you even talking about and what are you basing this on
Common sense
Definitely doesnt add up how cheap they produced this for, theres no f***ing way they made this for less than a 2 bedroom condo in new york city lol
He and other heads are gonna say this because they have investments at risk they need to justify and are unnerved by the uncomfortable discussions this has led to.
Deepseek didn't include hardware in training costs and said they used some type of Nvidia chip. It doesn't literally cost $6M thats just what the 2 months of training added to. That's not what really matters. The issue they really have is that a Chinese lab democratized AI and went open-source in a sector filled with closed-AI companies greedy for hundreds of billions in new investments. Why pay thousands a year for pro models when this is basically free
They've been publishing research papers on their advancements for a year, released their major model 3 weeks ago and a new one last week, but now all of a sudden everyone has a hardware excuse right as the market is reacting lol yeah right
This whole fiasco today and some of the narratives that have miraculously appeared have been outright embarrassing.
Trump saying its a good thing if its actually cheaper cus now "you'll be doing it too" and not paying as much lmaoo
Trump saying its a good thing if its actually cheaper cus now "you'll be doing it too" and not paying as much lmaoo
https://twitter.com/TheInsiderPaper/status/1884022839667564753they sound so shook after they did that huge ass press release at the WH
they sound so shook after they did that huge ass press release at the WH
“put that name down in your books, because I think you’re gonna be hearing a lot about it”
“This is a resounding declaration of confidence in America’s new potential”
They use a mixture of experts architecture which sort of uses specialized models for different inputs, using fewer parameters for each input and therefore lowering compute costs
Also with it being open source you can run the model locally and avoid paying api fees altogether
MoE has been around in frontier models way before deepseeks models came out
honestly if you have the background I would just read the technical paper on r1 they dropped or at least just read a breakdown or something
would clear up a lot of these misconceptions
What I’m confused about is outside of this pushing pricing down across-the-board..
Isn’t an open AI in an advantageous position since this is open source technology and they have the capital, talent, and existing infrastructure to take advantage of these new innovations and use deep seek’s open source model to improve their own solutions?
there is no real strategic advantage to doing this, they could undercut prices without open sourcing stuff
ig it helps in terms of building repute/prestige
the other argument is that it kind of implicitly boosts r&d by giving talented but under-resourced researchers the opportunity to build on your work and open source their own results, which helps counteract the fact that a company like openAI probably has more funding/talent
like prior to the r1 paper no one outside of people actually working in these labs knew how reasoning models worked, there was broad speculation but nothing concrete
This. I'm busy coding games and tools without it taking me hours to program.
I'm not wasting my time trying to trick the ai.
It's a powerful tool to use for your projects and to make your life easier.