News for hackers. Stuff that matters.
Re:Don't argue with me. $100k is fair, and you know it by cadamsdotcom on Friday November 13, 2026 @12:43AM (#50478850)
My recommendation is to have ChatGPT write code to do calculations like this, and fetch data for them that it plugs in so you can see it as constants in said code.

Just like you wouldn't hammer in a screw. Use different tools for the jobs they're best at: code where it's best, and LLMs where they're best.

AI is a powerful tool because it can use tools and can also build tools for you to use or for it to use. But it is not a calculator.

Re:Don't argue with me. $100k is fair, and you know it by p-e-w on Friday November 13, 2026 @12:50AM (#50478886)
> AI is a powerful tool because it can use tools and can also build tools for you to use or for it to use. But it is not a calculator.

It’s both, and a million other things. Of course AI can do complex calculations today. I can’t even remember the last time I saw an AI make a mistake on something like that.

The problem is that people’s mental image of what AI can do always has an inertia of 6-12 months. Like you can still see people talking about how image generation models supposedly can’t draw hands, even though today we have 2K AI-generated video with perfect hands.

Re:Don't argue with me. $100k is fair, and you know it by cadamsdotcom on Friday November 13, 2026 @01:27AM (#50479167)
It can only do that because it's fed millions or billions of synthetic calculations.

Code has determinism and runs in milliseconds.

Re:Don't argue with me. $100k is fair, and you know it by speerer on Friday November 13, 2026 @02:44AM (#50479770)
I think it might get the last one of those questions correct.
Re:Don't argue with me. $100k is fair, and you know it by jmmcd on Friday November 13, 2026 @05:24AM (#50481296)
But the model writes a python or js program to calculate, it doesn't use next-token to guess.

(BTW the last one was meant to be (3-1).)

Re:Don't argue with me. $100k is fair, and you know it by jmmcd on Friday November 13, 2026 @06:47AM (#50482159)
But if you add a few digits, the model writes a python or js program to calculate, it doesn't use next-token to guess.

(BTW the last one was meant to be (3-1).)

Re:Don't argue with me. $100k is fair, and you know it by rmunn on Friday November 13, 2026 @01:36AM (#50479238)
I've seen AI-generated videos posted two weeks ago with six fingers on someone's hand. I don't know what model they used, but it still happens. Less often than it used to, granted, but it still happens.
Re:Don't argue with me. $100k is fair, and you know it by cadamsdotcom on Friday November 13, 2026 @01:44AM (#50479304)
Smaller models save money so even more slop can be generated.

Creator did you a favor making their slop so detectable ;)

Re:Don't argue with me. $100k is fair, and you know it by sampullman on Friday November 13, 2026 @01:49AM (#50479337)
The latest models can do fairly complex arithmetic without tools now, but they're definitely not calculators.
Re:Don't argue with me. $100k is fair, and you know it by anonymous908213 on Friday November 13, 2026 @03:20AM (#50480086)
> ChatGPT won't solve problems with code unless asked,

It helps to clarify what model you're talking about. Maybe you mean the free model, Luna? Using whatever their frontier paid model in the chat interface is at the time on xhigh, it defaults to writing python scripts for everything vaguely calculation-shaped for at least the past year. I can literally remember a specific number I asked about last Nov and it used a script. I just asked this exact question from GP and it returned $310m after running a python script unprompted.

Re:Don't argue with me. $100k is fair, and you know it by cowlevel on Friday November 13, 2026 @04:10AM (#50480499)
ChatGPT has been secretly doing that under the hood for many months now.
Re:Don't argue with me. $100k is fair, and you know it by pmarreck on Friday November 13, 2026 @10:09AM (#50484674)
Early on, pretty much as soon as it was able to run arbitrary Python code it created (note: it can use other languages now too! Ask it!), I instructed it to NEVER guess at any question that is best served by computing it or counting it.

So I never actually saw the "strawberry" problem since it always just wrote a program to count the "r"'s!