[–] 8 points 2 weeks ago

A few years back I was talking to an acquaintance that worked in the NSA for about a decade, he was far to casual when he said "Oh, yeah we have back doors in Linux and all over the open source world that haven't been discovered.".

I mean, yeah, that tracks but did he have to be so proud of his achievement. He was one of those, 'see if it can be done don't question if it should' kind of people.

I suppose it is a part of their NOBUS "NObody But US" strategy but with the code being in the open, that is a dangerous game to play with it being discovered by others. Nowadays with LLM's able to do mass analysis of code, I do wonder if there is some regrets in putting those things in there?

But also I suspect a lot of it probably needs physical access to the machine as just sending a few packets to a machine probably isn't enough to get it into an executable state.

Either way, interesting guy, wouldn't trust him with anything!

  • source
  • [–] 24 points 2 weeks ago (4 children)

    As much as I dislike China, I'm hoping they can do to RAM what they've done to things like solar panels and EVs.

    Just flood the market. Make RAM cheap again!

  • source
  • [+] -11 points 2 weeks ago

    What we call hallucinations is operating on a lot of assumptions while you don't have enough data, instead of saying "I actually don't know what am I doing". People do it too.

    It's not like that's not a problem, sure, but the further this tech goes, the better it is at not hallucinating hard.

  • source
  • parent
  • context
  • [+] -11 points 2 weeks ago (2 children)

    Confidently incorrect. It has reasoning layers. It creates new text based on the context of the previously written conversation. That conversation could be in the context of things that weren't in it's training dataset, and you still will receive some sort of an answer.

    You don't need a datacenter to run LLMs. A decently strong PC can run useful stuff already. The real fight is not anti-AI, the real fight is locally run AI (means of production you own) vs datacenter AI (means of production you rent, that can be restricted any moment).

  • source
  • parent
  • context
  • [–] 11 points 2 weeks ago (2 children)

    Having a centralised and planned RAM economy seems a bit communist, TBF. It's just the small detail about it all being to the benefit of one man that's bothersome.

  • source
  • [–] 13 points 2 weeks ago

    This data center build out can't last forever, they are so overextended, obviously the government will bail them out afterwards, or either and or both parties. But that should at least halt these build outs for the most part, and lower computer costs, (although the manufacturers are likely geared to llm specific equipment and would have to retool, and the market can remain irrational longer than we (me, myself, and I,) can remain sane, apparently.)

  • source
  • parent
  • context
  • [–] 20 points 2 weeks ago (3 children)

    It's not AI, it's an LLM model. It will never become AI, because llm models are incapable of thinking, of reasoning, it's giving you an answer based on all the answers ever written they have on file. The hallucinations can't be eliminated.

    Don't believe the hype. These data centers need to be destroyed, after due process by law. They will get bailed out by the government then agree to run social scores on us, that's what is going to happen.

  • source
  • parent
  • context
  • [–] 4 points 2 weeks ago* (last edited 2 weeks ago)

    Hardware is the means of production and software leverages that. AI is part of that software suite working in tandem with hardware to be productive.

    I would argue computers in general are both the means of production, as well as the radio networks like LoRa that can let them communicate without large infrastructure buildouts, intstead relying on truly neutral mesh connections with no central servers. Thus is covers both production and communication in one device.

  • source
  • parent
  • context
  • view more: β€Ή prev next β€Ί