• 0 Posts
  • 55 Comments
Joined 2 years ago
cake
Cake day: June 26th, 2023

help-circle














  • That’s your opinion/agenda, not a legitimate argument in the conversation about AI efficiency. The discussion is on how best to achieve a goal, and you’re saying that it shouldn’t be achieved. Even if you’re right, you’re still going off on a separate tangent.

    You’re the vegan who butts in on the conversation about how best to sear a steak and says meat is murder. You’re welcome to your opinion on meat and you may even be right, but it is of absolutely no value or interest to the people talking about methods for cooking meat.



  • So what happens when OpenTuna runs out of fish to steal and there are no more boats?

    Information doesn’t stop being created. AI models need to be constantly trained and updated with new information. One of the biggest issues with GPT3 was the 2021 knowledge cutoff.

    Let’s pretend you’re building a legal analysis AI tool that scrapes the web for information on local, state, and federal law in the US. If your model was from January 2008 and was never updated, then gay marriage wouldn’t be legal in the US, the ACA wouldn’t exist, Super PACs would be illegal, the Consumer Financial Protection Bureau wouldn’t exist, zoning ordinances in pretty much every city would be out of date, and openly carrying a handgun in Texas would get you jailtime.

    It would essentially be a useless tool, and copying that old training data wouldn’t make a better product no matter how cheap it was to do.


  • Yes, but that doesn’t mean it is more efficient, which is what the whole thing is about.

    Let’s pretend we’re not talking about AI, but tuna fishing. OpenTuna is sending hundreds of ships to the ocean to go fishing. It’s extremely expensive, but it gets results.

    If another fish distributor shows up out of nowhere selling tuna for 1/10 the price, it would be amazing. But if you found out that they could sell them cheap because they were stealing the fish from OpenTuna warehouses, you wouldn’t argue that the secret to catching fish going forward is theft and stop building boats.


  • chiliedogg@lemmy.worldtoMicroblog Memes@lemmy.worlddeepseek
    link
    fedilink
    English
    arrow-up
    2
    ·
    edit-2
    2 months ago

    Which is fine in many ways, and if they can improve on technical in the process I don’t really care that much.

    But what matters in this case is that actual advancement in AI may require a whole lot of compute, or may not. If DeepSeek is legit, it’s a huge deal. But if they copied OpenAI’s homework, we should at least know about it so we don’t abandon investment in the future of AI.

    All of that is a separate conversation on whether or not AI itself is something we should care about or prioritize.


  • chiliedogg@lemmy.worldtoMicroblog Memes@lemmy.worlddeepseek
    link
    fedilink
    English
    arrow-up
    2
    ·
    2 months ago

    I think the big question is how the model was trained. There’s thought (though unproven afaik), that they may have gotten ahold of some of the backend training data from OpenAI and/or others. If so, they kinda cheated their way to their efficiency claims that are wrecking the market. But evidence is needed.

    Imagine you’re writing a dictionary of all words in the English language. If you’re starting from scratch, the first and most-difficult step is finding all the words you need to define. You basically have to read everything ever written to look for more words, and 99.999% of what you’ll actually be doing is finding the same words over and over and over, but you still have to look at everything. It’s extremely inefficient.

    What some people suspect is happening here is the AI equivalent of taking that dictionary that was just written, grabbing all the words, and changing the details of the language in the definitions. There may not be anything inherently wrong with that, but its “efficiency” comes from copying someone else’s work.

    Once again, that may be fine for use as a product, but saying it’s a more efficient AI model is not entirely accurate. It’s like paraphrasing a few articles based on research from the LHC and claiming that makes you a more efficient science contributor than CERN since you didn’t have to build a supercollider to do your work.