The original post appeared on LinkedIn. You can view it below on Linkedin or scroll below for the web version.
Should Machines Think Like Us or Not?
This weekend, Gourav shared an interesting essay on AI that got me thinking about a contradiction in how we're building artificial intelligence.
The paper argues that AI has hit a wall. We've fed machines almost all the human-created data available - books, articles, conversations, you name it. To get smarter, AI needs to stop copying humans and start learning through experience, like trial and error.
Think of it this way: instead of learning to drive by reading driving manuals, AI should learn by actually driving and figuring out what works.
Here's what really struck me. The authors said something profound:
"Human thinking probably isn't the best way for machines to think. They could discover much better ways to process information and solve problems."
Essentially: Let machines be machines, not human copycats.
But here's the weird part. Right after saying machines shouldn't think like humans, the authors keep using human examples to explain their ideas:
[1] "Machines should learn continuously like humans do over years"
[2] "They should set long-term goals like humans learning languages or staying healthy"
[3] "They should interact with the world like humans do through their senses"
Effectively, their entire framework is based on how humans and animals learn:
Take action → Get reward → Adjust behavior
It's like saying "Cars shouldn't move like horses" and then designing a car that gallops
My Question
Why this matters? - This isn't just nitpicking. How we answer this question shapes the future of AI:
Maybe the real breakthrough will come when we stop trying to resolve this paradox and just let machines surprise us with their own solutions.
What do you think - Should AI think like us, or should we let it discover its own way of being smart?
Link for the article in the comments
