Feb 10, 2026 · 27 min · 15 segments
Summary In this episode, Autumn and Noah explore the intersection of AI and mathematics, discussing why AI struggles with math, the differences between calculus and algebra, and the historical…
Noah GiansiracusaHostAutumn PhaneufHostI think the easiest way to think about it is AI is not one thing, just like humans are not one thing.
That's like saying why are people bad at math? Terrence Tao, I would say, is pretty darn good at math.
Many people, often myself, when I'm at a restaurant trying to figure out the tip, are quite bad at math.
So kinda what happened was it really depends on how you're training these AI systems.
What you're intending them to do, unsurprisingly impacts what they're good at.
So early on, we were playing with chatbots and large language models that are really just based on predicting words and having conversations, and there's all these funny foibles, like they couldn't count the number of Rs in the word strawberry.
And that's because they weren't trying to learn rules and principles and logic.
You know, you can have inequality and add something to both sides, and it still equals.
They just kinda guess and look what's plausible, and it turns out that works great for a conversation where things are all kind of imprecise.
But there are systems that are fine-tuned or custom-tailored or really designed to do math, and within the last couple months, these have really been taking off and doing amazing things.
There's even been some evidence I've seen on some really cool discussion on Twitter/X and probably other platforms of them really doing, like, I don't quite wanna say novel research, but helping mathematicians do some novel research.
One found this, like, obscure reference in a German journal from the 1970s that had, like, the key lemma that helped them solve a, a problem.
So AI now is very good at math, but that doesn't mean if you're using a chatbot it'll be good at math.
It's sorta like saying, "Why are hammers so bad at putting screws in, and why are screwdrivers so bad at working with nails?" You need the right tool for the right job.
I think the easiest way to think about it is AI is not one thing, just like humans are not one thing.
That's like saying why are people bad at math? Terrence Tao, I would say, is pretty darn good at math.
Many people, often myself, when I'm at a restaurant trying to figure out the tip, are quite bad at math.
So kinda what happened was it really depends on how you're training these AI systems.
What you're intending them to do, unsurprisingly impacts what they're good at.
So early on, we were playing with chatbots and large language models that are really just based on predicting words and having conversations, and there's all these funny foibles, like they couldn't count the number of Rs in the word strawberry.
And that's because they weren't trying to learn rules and principles and logic.
You know, you can have inequality and add something to both sides, and it still equals.
They just kinda guess and look what's plausible, and it turns out that works great for a conversation where things are all kind of imprecise.
But there are systems that are fine-tuned or custom-tailored or really designed to do math, and within the last couple months, these have really been taking off and doing amazing things.
There's even been some evidence I've seen on some really cool discussion on Twitter/X and probably other platforms of them really doing, like, I don't quite wanna say novel research, but helping mathematicians do some novel research.
One found this, like, obscure reference in a German journal from the 1970s that had, like, the key lemma that helped them solve a, a problem.
So AI now is very good at math, but that doesn't mean if you're using a chatbot it'll be good at math.
It's sorta like saying, "Why are hammers so bad at putting screws in, and why are screwdrivers so bad at working with nails?" You need the right tool for the right job.
Every episode on Radar is fully transcribed, speaker-labeled, and rich with metadata. Here is a taste of this one. Try Radar for free to see the rest.
3 of 6
The dark side of pricing
How your data quietly changes prices
Algebra vs. calculus, simply
Understanding change starts right here
AI isn’t one thing
Why some AIs flub math, others excel
+3 more clips · 11 min 58 sec of audio in all
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
All 6 clips — the highlight moments, each cut as its own audio, with a title and a speaker
All 15 segments — the transcript broken into labeled sections, every ad read marked
All 25 topics — jump to every other episode discussing the same subject
Every related episode — other shows Radar links to this one
No account is needed to search Radar.
