32 Replies to “What The Heck Is This Thing?”

  1. On Bannon’s show this am, it’s being talked about as more of a (another?) back door into your data (and your government’s data) to work for their data farm. I think he’s stating the obvious but it’s good the news gets out there.

    Those smart Tv’s, the smart refrigerators, the smart whatever you allow to access your data… so many trojan horses

    I wonder if Hillary’s data is still connected in some Mainland databank … just humming

    1. Remember the Stuxnet? I laughed when the Israelis did this, and thought it was the perfect means to infect Iran’s nuke industry through porn. I fully expect the Chinese to infect us (yes I know, they already gifted us that 5 years ago), that if war (a kinetic and actual war) begins with Taiwan, none of our refrigerators will function, our connected cars will be bricked, nothing will function as it should, all of our telephones will ring for the next 90 days, non-stop. (it’s an example..)

        1. Almost any sw can be hacked. What of it? Open source software has a community securing it, proprietary sw has a big corp, guess which one has a better track record?
          At least Linux doesn’t come pre-hacked, like Windows, Android and MacOS.

          1. > Which one has a better track record?

            Proprietary closed source by a mile. The vast majority of people using open source code have no idea what it’s doing and have never looked at the code, much less performed any kind of security audit. There are catastrophic vulnerabilities in foundational open source libraries that have persisted for decades.

          2. You are full of it, and I can’t help but remember when you blamed piracy for Adoby’s TOS taking possession of peoples work to train their AIs.
            Making you a stooge.

          3. “Open source software has a community securing it, proprietary sw has a big corp,”

            Guess which one can get it’s socks sued off if they f- it up. Yep, BigCorp will get sued. Fear of being sued is a powerful motivator.

            Guess how long it has been since anybody read through the code in some of those libraries -everybody- uses…

  2. – “It’s not magic, just incredibly clever engineering”
    – “Competition increases dramatically”

    Gee … where have I heard these kind of phrases before ? Hmmmm ? Oh yeah! That’s how Trump speaks. Those are Trumpisms!

    Let me suggest that Silicon Valley AI spent wayyyyyyyy too much time ensuring their AI was anti-white and was chock full of TDS than they spent on a rational, efficient architecture. It matters not how big a pile of garbage you shovel IN … you will always get garbage OUT. It’s never been more “magic” or complicated than that.

    However … Perhaps Trump is way out of his depth on this topic and should have listened to his “First Buddy” Elon who warned against “investing” $1/2 Billion taxpayer dollars in it?

    1. Well Donald … this didn’t age well … and FAST

      Trump downplayed Musk’s skepticism about the $500 billion AI project — attributing it to personal animosity rather than a critique of its viability. Trump told reporters that Musk’s remarks stemmed from his dislike of Altman, emphasizing that all participants in the deal are “very, very smart people” and comparing Musk’s feelings to his own personal grudges.

      https://www.msn.com/en-us/technology/tech-companies/trump-aides-furious-with-elon-musk-for-trashing-500b-ai-project-he-gives-zero-f-ks/ar-AA1xOojn

      Turns out those “very, very smart people”were outsmarted by a small group of unknowns from quingkong China. Next time … best listen to Musk and not to Larry Ellison, Sam Altman and other Tech Billionaires who HATED your GUTS and conspired to ruin your election chances a couple months ago.

      1. Watched the whole thing. That is truly frightening. Trump needs to detach from Sam Altman and Larry Ellison ASAP.

        After listening to this grieving, yet clear minded woman … I want to give her a hug and pledge all my own prayers that she finds justice for her son.

  3. If this is true (DeepSeek is better from a $5M investment than OpenAI is from $100M and DS runs on a laptop while OpenAI needs Nvidia’s $2T market cap GPU) then America was building a toilet out of solid gold and the Chinese just made one from porcelain.

    1. Sam Altman is a scammer. Testified that he wasn’t interested in money or going public. Then is spotted driving a multi-million dollar Koenigsegg Regera.

      1. I’m willing to bet that Altman’s swimming pools, at his multiple homes, are all larger and have more waterfalls than yours do.

    2. To think all that round tripping nvidia did may come to naught. Then again it will probably be the retail investor who takes the fall.

  4. …and it’s all open source.
    China respecting consumers more than the west?
    That’s unpossible!!!

    This could change alot. A small, powerful AI that you can train yourself, run locally, incorporate into other sw…

    2. OTOH … tell me again why we had to cancel TikTok? Yet now DeepSeek is the #1 app download on Apple??

      If they’re clever-enough to kick Silicon Valley in the balls … how clever are they at spying on and stealing all your data

  6. Appreciate the effort, but most of the information in that thread was either outright wrong, or semi-accurate.

    Deepseek is a major step forward in training techniques, but it’s not as momentous as he makes it sound.

    The “gaming gpus” he mentioned are extremely powerful processors of the same basic type they use for training AI models. GPUs started out entirely dedicated to playing videogames, and they were later adapted to use within AI.

    The 32/8 bit remarks — just standard quantization that we’ve been doing for several years. They trade off some accuracy for memory. Old news.

    “”multi-token” system — highly simplified and misleading description of how tokenization and inference works. There are many similar systems out there.

    Anyhow enough of that. This guy could easily have just read their own damned documentation rather than patching together this blob of mis- and semi- information.

    For those who are actually interested in this topic:

    https://arxiv.org/abs/2501.12948

    https://huggingface.co/deepseek-ai/DeepSeek-R1

    1. As a layman … I find it nearly impossible to read those articles because of all the tech jargon. I suppose this is the language of programmers … but I am still an analog human, living in an analog world. Witness my devotion to vinyl

    2. People who aren’t software engineers can be easily impressed by fairly basic systems engineering principles.

      But then, most “software engineers” don’t know any systems engineering principles, which is how you get the “throw more hardware at it” mentality until you reach physical scaling limits.

    1. Closed down 16.97% woo hoo !! Under the 17% barrier.

      So investors lost nearly 1 out of every 5 dollars invested. Ouch. Any further slide is gonna bring serious pain

  9. AI IS BS.

    What is the point? All these things need to be trained and, and they will just immitate each other. AI systems are plagiarism systems.

    1. To repeat, AI is just a large data base with a search engine and language interface. Now, for specific applications, the data base can be smaller. For an AI system to brag, the data base is the internet; the search engine is nothing new. Nor is the language interface. The “training” is just feeding it examples. Thus, a new raga can only be “invented” once the AI system has been given samples of ragas and they be identified as such. If you just relied on the internet, the world’s largest general purpose data base, then it would take a much longer time to find samples of ragas and then the system would only repeat them, with at best small variations.

      Sorry, no intelligence.

  10. AI that answers a question, then deletes that answer, and says it never said it. That folks is fact, not fiction. That is your Chinese DeepSeek. Carry on.

Proudly powered by WordPress & Skyrocket Themes

MENU

Navigation