18 Replies to “I, For One, Welcome Our New Self-Driving Overlords”

  1. I’m old, but at the end of AI research has to be some product we value. I am not sure where it is going beyond doing fake videos of important people saying ridiculous things. Maybe a robot arm that has perfected Five Finger Mary might have some universal appeal but past that I think they are reaching.

  2. I would say, based solely on my (admittedly limited) understanding of the basics of digital computers, that some people who should have bloody well known better have just realized that this LLM software is -never- going to do what they thought it could. Now they are moving to limit the damage they have invited by buying into their own hype.

    Huzzah.

    Now, this is not to say that LLMs have no uses, or that they can’t make money. There will be stuff they can do. Navigating messy environments appears to be one of those things, so robots are going to get a lot smarter.

    But, LLMs are -never- going to do all the things their inventors hoped for.

    A minor example thought experiment. If A=B, and B=C, then A=C. Humans -understand- this statement. They can also derive this statement by observation, a generalization that describes something from the real world.

    Some very powerful Silicon Valley people thought that LLMs would be able to do that too, but they can’t. And they never will. They can state it, they can use it as a rule in an algorithm, they can reproduce it by scanning text, but they can’t understand it. It is just a text string, no different than “Mary had a little lamb.”

    I view this as a failure of the American educational establishment. Specifically English and Philosophy. They never told these kids how to figure out what a thing -IS-, or even to consider that important to know. So here we are, and knowing what an LLM really -IS- turns out to be worth billions and billions of dollars.

    1. currently there is something in the analytical most most complex human brain (as well as certain individuals of other mammalian species, yes l do believe some dogs, moose, elephants, lions chimps, mice etc can be deemed ‘genius’ compared to their peers) that is and will REMAIN missing in the digital realm which at its most BASIC level is just BINARY digits aka ‘bits’. a choice between ‘1’ or ‘0’, on/off, open/closed whatever the specific technology happens to be. not much room for variance there.
      when l started out in IT one could actually SEE a bit, it was a temporary polarized round magnet, the bit value depended on the N-S or S-N polarity.
      biologically? more than that. consult the medical experts.

      p.s. hey mr SPORT carbuncle, hows that for ‘gibberish’? ha ha ha !!!

  3. I’ve used Grok a lot and I’m impressed. Its understanding of English is excellent, even when I bugger up the spelling and grammar in my questions. And it gets humour and sarcasm. It could probably pass the Turing Test for intelligence. HAL is pretty much realized.

    1. “Its understanding of English is excellent…”

      Grok has no understanding of English. Or humor, or sarcasm. It -reproduces- what it is trained on, and it predicts the next letter in a text string or the next pixel in an image. It understands nothing.

      That’s why all the six-fingered hands, and clock faces that always read ten to two. That’s why those guys are cancelling billions of dollars worth of server farms. They figured it out before you did.

      “HAL is pretty much realized.”

      You remember what happened to HAL, right?

    2. It’s not difficult to pass a Turing test for intelligence…if the human running the test is vain and stupid. ELIZA could do as much, and all it did was repeat back what you had just said to it if it could (“Why do you think the human running the test is vain and stupid?”) or some canned response if it couldn’t (“I see.”)

      Grok and others of its species are useful for writing copy that doesn’t require any originality or thought, such as pornography or anti-Trump propaganda recycled from the mainstream media. They will never replace humans who know what they’re talking about.

      The only people who believe they might are people who don’t know what they’re talking about and don’t care to be told what they’d rather not hear by people who do.

      Yeah, let’s fire all the grumpy old White men who know what they’re doing and refuse to be dictated to by a DEI manager young enough to be theor daughter and replace them with ELIZA! Think of the fortunes the firm will save…!

    3. It turns out that the Turing Test is a terrible judge of general intelligence.

      With a skilled examiner – say, one familiar with the Voight-Kampff protocols – it’s not hard to get even a sophisticated LLM to flunk a Turing Test.

  4. AI might mirror what happened with the development of the internet:

    1. People make all sorts of claims about it.
    2. Stocks related to it soar in price.
    3. These stocks then crash, bringing the rest of the market down with it.
    4. The technology turns out to be as revolutionary as claimed, but in unexpected ways.
    5. Some stocks related to it recover and do well, but not necessarily the original ones.

    2. Perhaps since the development of markets rather than the internet. Pretty sure AI is just a modern version of the tulips.

      LLM is just pattern matching not a true ai and better/faster/more data is never going to lead to a true ai.

      The main issue that the development of it and the stock values gained outweigh the possible markets.

      as I remember the numbers corrrectly if AI completely replaced ALL porn, well thats about 1/3 of what ai is currently valued at.

    3. At the moment, AI is cold fusion. We have cold fusion right now. It just requires more energy in than it puts out, which makes it useless.

      LLMs can do useful things, but to do so they require orders of magnitude more resources than any other option.

  6. Carney’s vision of Building Infrastructure was never about roads or pipelines. It was about Data Centers. The new, cool thing that would make urbanites swoon.
    So if the Liberals are entranced by it, you know it’s going to go belly up.
    See also…EV’s, Windfarms, Gun Grabs, etc.

  7. Software engineer from 18 to 57. Previous bubbles were 4th generation language would make software engineers obsolete. Dot-com bubble. Dozens of companies losing vast piles of money while kids cruise around South of Market on scooters imagining enormous IPOs.

    Some jobs will go bye-bye to AI. Many white-collar workers are only slightly useful. Blue-collar jobs are going nowhere although they may be augmented by AI tools.

    1. To put it another way, computers made the jobs of the Girl Fridays obsolete, 40 years ago. If the top brass had wanted to cut costs, all they had to do was let the girls go.

      Alas, because Girl Friday by this point too often went in society as Mrs. Corner Office, she was immune from cost-saving measures. Corner Office couldn’t as easily outsource the work done by the guys who did, well, all the real work as the services provided by Girl Friday without his personally risking financial ruin.

    2. to clarify, by ‘going nowhere’ do you imply ‘no future there’ or do you mean ‘staying put, firmly established
      not leaving where they happen to be now’?

Proudly powered by WordPress & Skyrocket Themes

MENU

Navigation