The Register Home Page

back to article Navigating the truth maze in a world of clever machines and cleverer marketers

All the worries and fears about AI boil down to one. How do we know how well it's working? The kind of benchmark that IT normally worries about isn't without importance. How fast a particular data set is learned, how quickly prompts can be processed, what resources are required and how it all scales? If you're creating an AI …

  1. xyz Silver badge

    What is the truth?

    Seemingly it's whatever the missus says.

    1. Inventor of the Marmite Laser Silver badge

      Re: What is the truth?

      Yes dear.

  2. Headley_Grange Silver badge

    "Working" ?

    "How do we know how well it's working" - depends who's asking. For the people developing and selling AI "working" will be defined only by how much money it's making for them and how quickly they can achieve market dominance by buying or bankrupting their competitors. Given the history of search and advertising I don't think that the regulators have got a chance.

  3. Fonant
    Holmes

    Truth is a function of time

    ^ One of my favourite concepts.

    My personal guess is that AI is a fad that will pass. "Hey, we've made computers really good at pattern matching to generate plausible output" is all it really is. A fun toy, but not useful for serious stuff that matters.

    1. juice

      Re: Truth is a function of time

      > My personal guess is that AI is a fad that will pass

      I don't think it'll pass - there's a lot of genuinely useful stuff coming out, especially when it comes to user-assist stuff. But equally, I do suspect the hype bubble will crash soon.

      There was a recent article by Cory Doctorow which summed things up pretty nicely.

      https://locusmag.com/2023/12/commentary-cory-doctorow-what-kind-of-bubble-is-ai/

      The tl;dr version is that LLM models are both costly and unreliable. So you can't sell to the profitable high-end (where reliability is key), and it's too expensive for the low end (where price is key).

      And once people realise that, the VC money is going to dry up pretty quickly...

    2. John Miles

      Re: Truth is a function of time

      I suspect it will be here to stay but, like a lot of hyped things, not in the ways it is being hyped - e.g. I don't recall Java being thought of as a back-end language when it hit hype mill - more an applet language which would change the way we saw applications

    3. Headley_Grange Silver badge

      Re: Truth is a function of time

      I think the LLM side of it is annoying and it will get even more annoying. Already you can spot the eBay item descriptions that have been written by AI. Even without errors it's noticeable and starting to grind on me. If it persists then in a couple of years we'll be looking back on today's management-bullshit-speak as if it were Byron.

  4. Inventor of the Marmite Laser Silver badge

    How long before AI for general purpose is viewed with the same near-universal contempt as Clippy?

  5. Anonymous Coward
    Anonymous Coward

    I wonder

    Are we on the edge of an emerging parallel with Hari Seldons psychohistory in Asimov's Foundation trilogy?

    1. Anonymous Coward
      Anonymous Coward

      Re: I wonder

      "Are we on the edge of an emerging parallel with Hari Seldons psychohistory in Asimov's Foundation trilogy?"

      Short Answer:

      NO !!!

      Longer Answer:

      Not a hope in Hell !!!

      'Clever' pattern matching has its limits hence the 'hallucinations / make up a pattern that matches' problem with LLM's of late !!!

      :)

      1. Anonymous Coward
        Anonymous Coward

        Re: I wonder

        OP here. Very likely you're right, which is why I caveatted the question. Who knows what could ultimately be developed though

        1. Doctor Syntax Silver badge

          Re: I wonder

          In that case you should have posed the question in the headline to comply with Betteridge's law.

  6. Howard Sway

    Redundancy is a powerful weapon against error

    Money is an even more powerful one. Make the LLM peddlers liable for loss or damages resulting from erroneous output from their systems, and watch the marketing hype dry up instantly, replaced by "well, it's still just at an experimental stage, you shouldn't rely on the output or take it as being 100% true".

    1. Anonymous Coward
      Anonymous Coward

      Re: Redundancy is a powerful weapon against error

      And do the same with manufacturers of full self driving cars

  7. amanfromMars 1 Silver badge

    Seems like 2024 is definitely gonna be a really strange and surreal year.....

    ..... with weird winners and apathetic losers aplenty

    amanfromMars [2401021728] ........ airs on https://www.zerohedge.com/news/2024-01-01/hartnett-correct-ai-baby-bubble

    A.I. is a Baby Stock Bubble

    Oh please, you cannot be serious. Go tell that pile of nonsense to the Marines to hear of their worries and fears for the future because of considerably smarter than ever thought to be possible AIs ..... for there be a real canny few of them recognising their almighty overwhelming advantage in comparison to any Earthly competition or opposition from any indigenous native human population.

    amanfromMars [2401021005] ...... shares freely on https://www.nationaldefensemagazine.org/articles/2024/1/2/large-language-models-have-pitfalls-for-national-security

    [Thank you. Your comment will be displayed soon after reviewing.]

    Speaking Truth to Power Provides Wisdoms for Giant Leaps for Mankind

    Generative AI is an architecture, Svitlana Volkova, chief AI scientist at Aptima Inc., said during a panel discussion at the National Training and Simulation Association’s Interservice/Industry Training, Simulation and Education Conference in November.

    “It’s a neural architecture generating pre-trained transformers,” .... ......... https://www.nationaldefensemagazine.org/articles/2024/1/2/large-language-models-have-pitfalls-for-national-security

    That is a common rooky mistake to make  ...... to not understand and realise it is much more a vast remote neural ecosystem presenting and embedding into earlier human-trained training colonies SMARTR inhabitants/virtual machine entities for the relatively anonymous exercising and practically autonomous retraining of generative pre-trained transformers in an augmented alien environment with both incorporated private and public media platforms posting and hosting daily realities for existential crisis management ........ which one does well to realise and accept is the current Earthly existence descriptor for present day 21st century humanities?

    And shared here as a question, rather than servered as a loaded statement, to encourage enquiry into whether such is to be thought remarkable and/or feared capable of providing future solutions to emerging problems.

    Happy New Year ...... which, believe it or not, is certainly guaranteed to be completely different from any other previously planned for year ..... whether one believes such virtualised situations are possible and the true state of existential Earthly realities or not.

    Que Sera, Sera is a Grand Fait Accompli.

  8. Michael Wojcik

    Wait, what?

    All the worries and fears about AI boil down to one. How do we know how well it's working?

    Wow. Just wow. This is so wildly off it's barely even wrong.

    If you want to generalize sophomorically, you could lump, say, explicability into your question. Even, arguably, alignment, if you don't mind reducing a hugely complex area of research and debate into a useless mishmash of confusion. But "how well it's working" certainly says nothing about, oh, learned helplessness, or about competitive cognitive technologies.

    I realize the Register is a generalist tech publication, and so we can't expect highly-specialized coverage of all technical aspects of the industry. But the coverage of (so-called) AI has often been ... not great, let's say. And yes, I know this is an opinion piece; and yes, it's more sophisticated and thoughtful than most of what we get from the commentators here (which has largely been an incessant chorus of the same ignorant catchphrases); and yes, Rupert makes some good points. But could we avoid the hypergeneralization? That opening paragraph is like claiming "all worries and fears about" EVs are just range anxiety.

    1. amanfromMars 1 Silver badge

      Re: Wait, what?

      All the worries and fears about AI boil down to one. How do we know how well it's working?

      If the truth be told, indications as to how well AI is working is relative and directly proportional to human efforts trying to regulate and gain leading function of it ........ as El Reg and Lindsay Clark inform us

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like