Terence Tao just revealed why he isn't worried about AI.
The UCLA professor tested ChatGPT's reasoning against complex problems.
It performed like a mediocre student, with ONE disturbing pattern.
His reasoning challenges everything we think about AI dominance:🧵
The UCLA professor tested ChatGPT's reasoning against complex problems.
It performed like a mediocre student, with ONE disturbing pattern.
His reasoning challenges everything we think about AI dominance:🧵
Terence Tao is widely considered the world's greatest living mathematician.
UCLA professor. Fields Medal winner. Breakthrough Prize recipient.
When he speaks about the future of mathematics, the entire field listens.
His take on AI is surprising:
UCLA professor. Fields Medal winner. Breakthrough Prize recipient.
When he speaks about the future of mathematics, the entire field listens.
His take on AI is surprising:
Last month, Tao tested ChatGPT's o1 reasoning model on complex analysis problems.
The AI could answer difficult questions - but only with "a lot of hints and prodding."
Like working with a struggling grad student.
But here's where it gets interesting:
The AI could answer difficult questions - but only with "a lot of hints and prodding."
Like working with a struggling grad student.
But here's where it gets interesting:
"One key difference between graduate students and AI is that graduate students learn," Tao explains.
"You tell an AI its approach doesn't work, it apologizes, it will maybe temporarily correct its course..."
"But sometimes it just snaps back to the thing it tried before."
"You tell an AI its approach doesn't work, it apologizes, it will maybe temporarily correct its course..."
"But sometimes it just snaps back to the thing it tried before."
This isn't a minor flaw. It's fundamental.
When you correct a human mathematician, they internalize the lesson.
When you correct AI, it performs temporary compliance, then reverts to its original error patterns.
This creates a ceiling that AI can't break through:
When you correct a human mathematician, they internalize the lesson.
When you correct AI, it performs temporary compliance, then reverts to its original error patterns.
This creates a ceiling that AI can't break through:
Think about what this means for expertise.
A human expert builds on every correction, every failure, and every insight.
Their understanding compounds over years.
AI resets after each conversation.
No wonder Tao sees a different future than most predict:
A human expert builds on every correction, every failure, and every insight.
Their understanding compounds over years.
AI resets after each conversation.
No wonder Tao sees a different future than most predict:
"AI and mathematicians will more likely always be collaborators," he says.
Not because of sentiment or job protection.
But because AI excels at scale, while humans excel at learning.
Here's Tao's vision:
Not because of sentiment or job protection.
But because AI excels at scale, while humans excel at learning.
Here's Tao's vision:
"You might have a project and ask, 'What if I try this approach?'"
"And instead of spending hours and hours actually trying to make it work..."
"You guide a GPT to do it for you."
The human provides direction and insight as AI handles the computational heavy lifting:
"And instead of spending hours and hours actually trying to make it work..."
"You guide a GPT to do it for you."
The human provides direction and insight as AI handles the computational heavy lifting:
Tao believes AI will enable mathematicians to explore large-scale, previously unreachable problems.
Problems requiring massive calculation but human insight to direct.
And this extends to other jobs that people have previously thought would be replaced by AI.
Problems requiring massive calculation but human insight to direct.
And this extends to other jobs that people have previously thought would be replaced by AI.
A bit about me:
I spent 15 years in tech and blockchain & built a Web3 company to 7 figures without VC funding or tokens.
Now I share insights on tech entrepreneurship and building in Web3.
Follow me @constantine_rm for lessons from the trenches, not theory.
I spent 15 years in tech and blockchain & built a Web3 company to 7 figures without VC funding or tokens.
Now I share insights on tech entrepreneurship and building in Web3.
Follow me @constantine_rm for lessons from the trenches, not theory.
https://twitter.com/916759565596798976/status/1942229640283935116
Video Credit:
Terence Tao: Hardest Problems in Mathematics, Physics & the Future of AI | Lex Fridman Podcast #472
Terence Tao: Hardest Problems in Mathematics, Physics & the Future of AI | Lex Fridman Podcast #472
• • •
Missing some Tweet in this thread? You can try to
force a refresh
