Reply
  • Updated Jan 27

    cnbc.com/amp/2025/01/24/how-chinas-new-ai-model-deepseek-is-threatening-us-dominance.html

    “A little-known AI lab out of China has ignited panic throughout Silicon Valley after releasing AI models that can outperform America's best despite being built more cheaply and with less-powerful chips.

    DeepSeek, as the lab is called, unveiled a free, open-source large-language model in late December that it says took only two months and less than $6 million to build, using reduced-capability chips from Nvidia called H800s.

    The new developments have raised alarms on whether America's global lead in artificial intelligence is shrinking and called into question big tech's massive spend on building AI models and data centers.”

  • America spending BILLIONS on AI for financial speculation while China can do it at a fraction of the cost even accounting for wage differences

  • Jan 25
    ·
    1 reply

    They have h100s.. we dk how many, but I’ve read 50,000

  • Deepseek is fire.

  • Jan 25
    ·
    2 replies

    Can't believe China is gonna be our salvation. America is trash now

  • Jan 25
    ·
    1 reply
    Bo Ceephus

    Can't believe China is gonna be our salvation. America is trash now

    I can believe it

  • Jan 25
    ·
    1 reply
    Sloth

    I can believe it

    u shouldnt

  • Jan 25
    ·
    1 reply

    We will be their slaves honestly

  • maxx

    u shouldnt

    don’t believe your eyes or reality, trust and hold onto preconceived biases based on “rational facts”

  • Jan 25
    ·
    5 replies

  • Jan 25
    ·
    1 reply
    radiosonde

    We will be their slaves honestly

    first world projection because the imperial core sees the global south as slaves to extract value, that doesn’t mean they want you to become slaves

  • teku izing

    first world projection because the imperial core sees the global south as slaves to extract value, that doesn’t mean they want you to become slaves

    Stop it

  • AI that’s reinforcing its own learning is damn near the last puzzle piece

  • Jan 25
    ·
    2 replies

    ngl you have no idea what youre talking about lol but yeah deepseek is cool

  • Jan 25
    ·
    2 replies

    the real crazy thing is this is completely open-source and from free research available, their work is published and literally anyone can benefit from it.

    open source won again

    and deepseek is genuinely a huge marvel when comparing to gpt. solving things at a rate much faster, and providing answers that are more accurate lolz

  • Tkken

    ngl you have no idea what youre talking about lol but yeah deepseek is cool

    dang whatd he say wrong?

  • Jan 25
    ·
    1 reply
    Tkken

    ngl you have no idea what youre talking about lol but yeah deepseek is cool

    I haven’t said anything besides reposting a news article from CNBC, unless you extrapolated more information based off the title of the thread or article and care to share?

  • ghosting ®️
    Jan 25

  • Still no real time access. Ai is stupid

  • Jan 25
    ·
    1 reply

    乔丹笑

  • Jan 25
    ·
    5 replies
    teku izing

    I haven’t said anything besides reposting a news article from CNBC, unless you extrapolated more information based off the title of the thread or article and care to share?

    well if you ctrl+f stargate in that article you get zero results so that should hopefully answer your question

    and you cant "obliterate" a joint venture that was announced a literal week ago they only just started building out data centers for stargate

    also yes deepseek r1 beats o1 on a subset of benchmarks but openai is already rolling out o3-mini, then a better model next month. openai's moat is clearly not that large but they arent getting "obliterated" by anyone

    deepseek is cool because theyve built some highly compute efficient models to get around limited access to hardware and managed to make them comparable to foundation models that were already released by top ai labs. its very neat stuff but they arent ahead of the pack in any sense

    its a pretty dynamic race, it was looking like claude had caught up to gpt-4o and then this whole paradigm of varying/scaling up test time compute emerged and things shifted

    idk I could get into more technical detail but it feels like this is some kind of politically motivated thing on your part so whatever

  • I said this yesterday

    Money aint gonna improve AI. Its all about algod

  • Tkken

    well if you ctrl+f stargate in that article you get zero results so that should hopefully answer your question

    and you cant "obliterate" a joint venture that was announced a literal week ago they only just started building out data centers for stargate

    also yes deepseek r1 beats o1 on a subset of benchmarks but openai is already rolling out o3-mini, then a better model next month. openai's moat is clearly not that large but they arent getting "obliterated" by anyone

    deepseek is cool because theyve built some highly compute efficient models to get around limited access to hardware and managed to make them comparable to foundation models that were already released by top ai labs. its very neat stuff but they arent ahead of the pack in any sense

    its a pretty dynamic race, it was looking like claude had caught up to gpt-4o and then this whole paradigm of varying/scaling up test time compute emerged and things shifted

    idk I could get into more technical detail but it feels like this is some kind of politically motivated thing on your part so whatever

    I love Claude. Hopefully they catch up

  • Jan 25
    ·
    2 replies

    Blame Disney

  • Bo Ceephus

    Can't believe China is gonna be our salvation. America is trash now

    they’re gonna say tiktok is influencing you

1
2
...
11