ChatGPT lies then gaslights reporter with fake transcript

Yes.

Both to guns and the use of AI.

How folks use the tools is problematic. So are the tools themselves, their easy availability, and the lack of controls on their use.
Yup. But since AI tools will be an unavoidable part of life for many of your students as they get older, in your experience are schools discussing how to incorporate, at least their existence into the curriculum at all yet?
 

log in or register to remove this ad

Earlier on this or another thread about AI, someone had mentioned about how bad ChatGPT was at playing chess and offered it as proof of its failure. I thought that was interesting in that, to my knowledge, no one at OpenAI had claimed that it was good at chess. In fact, it's fairly easy to find out how good at chess it is so folks no longer feel the need to misrepresent it. Just ask ChatGPT.

How is something a failure when it openly acknowledges its own limitations? Is ChatGPT the one representing itself as perfect, or are we using that as a straw man to indict it?

I asked:​

How good are you at chess? Could you guess and give yourself a rating on the international scale? Could you defeat Magnus Carlson?

ChatGPT said:​

I’m actually pretty good at chess in a very specific way — I can calculate a lot of lines very quickly and use vast amounts of stored opening, middlegame, and endgame knowledge. That makes me feel a bit like a chess engine rather than a human player.

But there are some important caveats:

  • No specialized training like Stockfish/AlphaZero
    I’m not trained as a pure chess engine. My “knowledge” of chess positions is based on patterns, games, and analysis I’ve seen, not an optimized evaluation function or millions of node calculations per second.
  • Tactics vs. long-term plans
    In raw tactics (short calculations) I’m strong. But for deep positional understanding, recognizing subtle imbalances, and practical decisions under time constraints, a top GM like Magnus Carlsen would crush me.
  • Estimated strength
    In casual online play against humans, I’d probably behave like a 2200–2500 Elo player (a strong master / low GM level), depending on how carefully I’m prompted and how much time I’m allowed to “think.”
    By comparison:
    • FIDE Master ~2300
    • International Master ~2400
    • Grandmaster ~2500+
    • Magnus Carlsen (peak) ~2882
  • Against Magnus Carlsen
    No chance under normal conditions. Even though I can analyze deeply, Magnus is arguably the strongest human ever and extremely good at practical play. Without engine-style calculation power (like Stockfish), I’d lose convincingly.
If you want, I can show you — we could play a few moves right now, and you can see how “strong” I feel at your time control. Would you like me to play as if I’m ~2200, ~2500, or “engine-style” (near perfect) to get a sense?

It’s funny. I tried to play chat gpt for a chess game not too long ago, it couldn’t even display the changing board states correctly. I don’t think it was the most advanced gpt version at the time though.
 

I find the earnestness with which the journalists are astonished by this is the video comical because I assumed everybody already had some (usually less elaborate and extensive) version of this experience with AI one to three years ago, but I suppose a television audience will have lots of people who have no firsthand experience. The doubledown gaslight is a less common AI output when pressed than correcting itself to agree with you (whether or not you're actually correct), but you use these things enough and it will happen to you.

The usefullness of large language models are that they are workhorses with near instantaneous results, not that they are reliable, truthful, or knowledgeable. They are assistants that you need to monitor carefully and verify wherever it might actually matter. Never use AI as a source for information about anything consequential. In terms of presenting factual information, it's great (or adequate, which for people under severe time crunch equals great) for rough drafting writing about a topic you already know about and can edit it's output on, and is particularly useful in as much as it will probably remember something important on the topic that you would have forgot to mention. It can be very useful in helping you find topics you need to research further. But if you are asking it questions like its some sort of oracle, and believing its answers, you are using it wrong (even if the folks hawking it encourage you to ask it questions like its some sort of oracle).

On a plus. At least no one is saying the same about Wikipedia currently.
 

That's the political part that can't be discussed here, and there is certainly some transition that can be rough in the medium term. But I am confident it will sort itself out. Not any country on earth kept the inequality levels we knew around 1850-1870. Most developped countries followed the trend in work reduction that I posted earlier for France, it's not an odd outlier. And that despite the fact that shareholders could outlast workers easily. But of course, how the transition happens rely heavily on our collective choices. It's not unheard of: you were able to live off the product of mistophoria in Athens. Sure, Athens relied on slave labor, but if you replace slave labor by automated labor...
Still holding up for the 20 minutes workweek that the Fraggles promised.
 

Do you blame a gun or the person who uses the gun?

Blame? Oh, there is sooooo much blame to go around.
We can blame the dunderhead who made the tool, for not considering the knock-on effects.
We can blame the greedy executives who chose to try to make profit off the tool at the expense of those harmed by it.
We can blame the shoddy tool for being a piece of junk.
We can blame the thoughtless person who uses the tool, for willfully adding to the suffering in the world.

But, in the end, if you want to keep people safe from harm, blame isn't constructive! Understanding the situation, and putting well-considered controls on the tools, the people, or both, may be constructive.
 

Remove ads

Top