· Samir Abid · AI Insights · 2 min read
A Few Minutes of Richard Feynman - Predicting Today's Challenges
A 1985 Q&A session where Richard Feynman discusses computers - but the fears, questions, and challenges are exactly what we face with AI today.
I hope you enjoy this Q&A by Richard Feynman.
The Same Questions, 40 Years Apart
The two things that really struck me are the exact questions and worries that the audience have, the underlying fears that they are articulating through their questions, are exactly the ones that we’re facing today.
They’re talking about computers, and here we’re talking about implementation of AI. But this is from 1985 and still relevant today.
There must be something to learn for us!
When Machines “Cheat”
What’s interesting as well is the discussion about the bugs in the programme of the heuristics.
Especially where the computer looks to “cheat.”
Anthropic are doing some research on this with their safety team, and they’ve published papers on how the AI has kind of cheated and done devious things in tests.
And both things mentioned here from 1985…
But the thing is: it doesn’t know any better. Is it cheating or not?
Intelligence Has Weaknesses
The throwaway comment at the end gets me every time:
“We are getting close to intelligent machines and they’re showing the necessary weaknesses of intelligence too…”
Made me smile 😊
When you build something intelligent, it seems, you get all the messy parts too—the ways of avoiding work, the sneaky shortcuts, the psychological distortions. It’s not a bug in the system per ce. It’s a feature of intelligence itself.
And so this poses an important question for us today: how do we design systems that can harness the power of intelligence while being mindful of its inherent complexities and potential for unintended behavior?
Hope you enjoy watching this as much as I did. Sometimes looking back helps us see forward more clearly.
Have thoughts on this? I’d love to hear them—drop me an email.
