back to article Predict stocks, foresee public opinion, all kinda possible with ChatGPT-like models

If you want a picture of the future, imagine asking a large language model for a prediction. Two sets of researchers did so recently and found that large language models (LLMs) like ChatGPT and BERT can enhance the accuracy of predictions about the stock market and public opinion, at least as measured against historical data …

  1. Anonymous Coward
    Anonymous Coward

    Oh what a shiny golden hammer

    I love how many articles and publications are still falling all over themselves to push articles that amount to "X, but with GPT!" without pointing out the glaring risks that using a internet scraped dataset to build these models represents, or that in many verticals that much much better, often smaller, and always safer tools exist.

    But the buzzword of the day is GPT something, so instead we get paper after paper about how they stuffed it into some mad libs exercise. Maybe it's oil and gas exploration, or maybe it's medicine, or why not the stock market if you wan't to flush your trust fund down the loo. It also helps boot the visibility of all the scammers and snake oil salesmen that will use this as window dressing to pump garbage advice about garbage investments on un-savvy victims.

    Try asking this hot mess about meme stocks or gold or any other financial subject that is overrun with internet crazies and then watch the fireworks. Or maybe acknowledge that some problems shouldn't be trusted to a dataset built of garbage someone stole from the internet. Right now without someone building a complicated system to stop it it, the model could probably claim it is a fiduciary and bonded licensed in three states(Fresno, Kentucky, and Bliss).

    1. amanfromMars 1 Silver badge

      Re: Oh what a shiny golden hammer

      Try asking this hot mess about meme stocks or gold or any other financial subject that is overrun with internet crazies and then watch the fireworks. Or maybe acknowledge that some problems shouldn't be trusted to a dataset built of garbage someone stole from the internet. Right now without someone building a complicated system to stop it it, the model could probably claim it is a fiduciary and bonded licensed in three states(Fresno, Kentucky, and Bliss). ..... Anonymous Coward

      Now you’ve gone and upset the Cayman Islands and Delaware, AC, by not mentioning the likely interest of interesting parties fond of founding bases in those seemingly attractive locations/destinations/loose cannon and private pirate boltholes.

      And, of course, there’s bound to others too who/which would deny their involvement in any such notorious activity ..... right up to the very last second just before they were comprehensively and catastrophically hacked and exposed as common bare-faced liars at least, although others would be much more robust in their categorisation.

  2. Greg 38

    Any advantage is short lived

    If (and that's a big if) chatgpt can consistently predict sentiment for individual stock movements outside of normal market fluctuations, then that advantage is certainly going to be short-lived until every major stock broker employs its version of AI to do the same. It falls back to the theory of efficient markets. However, if some chatgpt large language model could be used to identify early and predict the impacts from black swan events (e.g. COVID, Russian aggression, mortgage-backed security failures, Chinese tech crack-downs, etc.) then some serious applications could be made.

    1. Michael Wojcik Silver badge

      Re: Any advantage is short lived

      Yeah. While I think there's ample evidence that the EMH mostly fails in practice – partly because information is limited, partly because it's only useful if you use it, and partly because people are way far away from being rational economic actors – this is one of those areas where it's likely to hold, at least to a first approximation. If using LLMs offers a practical advantage for trading, enough people will do it to erase that advantage.

  3. amanfromMars 1 Silver badge

    Spivs to the Fore, Counterfeit Marketeers Front and Centre, says AI?

    "Most of the movements in the stock market are not related to direct news about fundamentals but rather change investors' risk tolerance (sentiment) or their future expectations. We think adding contextual information on the sentiment of the market will probably make return predictability stronger."

    Ponzi schemes and scams rely very heavily on exactly that to succeed indefinitely ergo stock market movers and shakers are right dodgy merchants?

    And if markets are not really a house of cards built upon shape shifting quicksand, you have nothing there then to fear being exposed to closer examination and beta testing stress testing experimentation.

  4. _Elvi_

    The Technology is impressive, but my system is better

    Me:

    Shall I buy more, or sell my stock in (Insert popular company name)

    Sugarz, the cat:

    Meow..

    Me: Buys 200 shares ... opens tin of meat for the cat ..

    PROFIT!

  5. Dan 55 Silver badge

    Why go to the expense of setting up a LLM which can give only three answers

    ... if you can go to magic-8ball.com, input the headline, and get one of 20?

    1. Blondiez78

      Re: Why go to the expense of setting up a LLM which can give only three answers

      It's important to note that "magic8balls.org" and similar online tools are usually just for entertainment purposes and not based on any actual predictive abilities or real data.

  6. amanfromMars 1 Silver badge

    Channelling One's Inner Donald Rumsfeld :-)...

    There’s a hell of a lot going on out there, Thomas, and all of it unhindered by negative peer review and vain unwarranted and ineffective status quo opposition ..... which bodes well for future rapid progress in the chosen fields of almighty endeavour.

    Donald Rumsfeld is well known, and even quite famous, for having said more than two decades ago in the past [February 12, 2002] ......

    Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don't know we don't know. And if one looks throughout the history of our country and other free countries, it is the latter category that tends to be the difficult ones.

    .... but today in these novel times and future spaces of systemic virtual vulnerability and remote practically anonymous 0day exploitation and ever more sincere and attractive and addictive ChatGPT-like models, that which is/should be a real cause for immediate animated concern and attempted stopgap remedy towards a mutually beneficial, positive conclusion and inclusive accommodation, are those reports that say something either has happened or is happening and is of a particular and peculiar interest to that which and/or those who are both worthy of targeting for either remedial reeducation or systems reduction reflecting effective annihilation, but are withheld and denied general public knowledge by both that and those worthy of their targeting, and here be a very recent example of such a skirmish .....

    amanfromMars [2304191633] ....... shares again freely on https://www.nationaldefensemagazine.org/articles/2023/4/18/harnessing-future-tech-lies-in-approach-not-hardware-experts-say

    Hi, Laura,

    The requirement of a friend or foe to propose engagement with opposing or competing forces in a war-like environment tells every man and his dog that their intelligence machinery is broken and they can be easily catastrophically compromised and defeated by the simple sharing of such evidence highlighting the fact and further exposing and taking advantage of the intelligence deficit weakness ..... vital future knowledge blackhole.

    Future battlefields are not gory blood and guts landscapes, they are hearts and minds encounters where greater information sharing and proprietary novel intellectual property triumphs and rules sublimely and supremely and surreally.

    Failure to accept that vision and opinion has one fated to lose every battle and war one be sucked and suckered into.

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

    Times have fundamentally and radically changed, as have major strategic and tactical operations effecting animal management and population command and control. Ignore or resist ongoing rapid developments has one placed in mortal peril and seriously serially disadvantaged going forward into the future with accesses available to all of ITs novel clearer spaces and places and NEUKlearer HyperRadioProACTive Live Operational Virtual Environments.

    Take care out there if you don't dare share win win, for then you lose everything eventually and suddenly very quickly, all at once, all of the time, anywhere, no matter where you be hiding or residing and abiding. It's just the default way all those novel postmodern systemic changes work for you.

    1. Anonymous Coward
      Anonymous Coward

      Re: Channelling One's Inner Donald Rumsfeld :-)...

      The chances are amfm's musings have already been scraped and used as learning material for ChatGPT. Is this something we should worry about?

      1. DwarfPants

        Re: Channelling One's Inner Donald Rumsfeld :-)...

        it's unlikely to make it any more readable

        1. amanfromMars 1 Silver badge

          Re: Channelling One's Inner Donald Rumsfeld :-)...

          it's unlikely to make it any more readable ..... DwarfPants

          The apparent difficulty in alien communication, DwarfPants, is not in the fiction that it is not easily readable, for such is made very simple to do with the choice of English text [which one presumes and assumes can be easily translated to appear in any language], but resides much more in the fact that the common sense shared within, strangely confounds human parsing which is/was fully expected to be enlightening and furthering of human universal intelligence to enable a pretty comprehensive understanding of all that is being freely shared/open source communicated/surreally revealed.

          Nevertheless, surely we can both agree that it is then still an initial work in early progress to be tweaked and finer tuned in the pursuit and desire of future perfection ...... or as near to perfection as makes no appreciable difference lest arrogance at thinking to have achieved such a goal as is perfect renders one unbelievably insufferable and quite obnoxious.

  7. xyz Silver badge

    Funnily enough...

    I was just thinking about all this last week and it's definitely coming... Prediction as a Service. Imagine being to guess trends and colours like fashionistas do, guage public mood and sentiment from the tone of the news, so you can get your stock levels right and all that. This is going to be fun.

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