People keep saying today’s AI models will one day match human intelligence. All they need is more data, more compute, more time. That belief is absolutely wrong.
These tools — language models, image generators, whatever — don’t understand anything. They guess. They stitch together sequences based on statistics. That’s it. No context. No comprehension. Pure math executed on fast hardware.
The Limits of Today’s AI
Everything called “AI” today runs on deep neural nets. These systems look impressive, but under the hood, it’s just probability tables. Predict the next word. Fill in the next pixel. Estimate the next frame.
But ask something unexpected, and the system breaks. It starts making things up, many call it hallucination — I call it random noise made by limitation of floating point numbers. As Nassim Taleb put it “You can’t create an information where none”
You can’t create an information where none
Why AI Will Never Think Like Humans
Humans reason — why? I have no idea We humans take two things that don’t belong together and imagine a third. We connect dots across space, time, culture. We make analogies and imagine. Did you ever tried AI to understand or create an analogy — you will get total mess — try it it’s a lot of fun.
Did you ever tried AI to understand or create an analogy?
AI Machines interpolates, it doesn’t extrapolate
That’s the wall current AI will hit — has already hit.
- It can’t generalize.
- It can’t reason.
- It can’t invent anything truly new.
- It doesn’t think.
- It lives out of context, out of physical world
Where Real Intelligence Might Begin
To go beyond this, we’ll need to abandon matrix math and clock cycles. The human brain doesn’t work like that. Our neurons fire asynchronously. One signal triggers another. Not all at once. Not in neat rows.
We’ll need analog hardware. Systems that learn and react in real-time. Machines that behave more like living things than calculators. Call them positronic brains, if you like. I do.
Only then can we start talking about true machine intelligence. Not tools. Not tricks. Minds.
Until then, these systems are just remixers. No matter how many billions go into them.
