“Programs written by humans.”
These aren’t “programs”, i e. sequential and conditional machine instructions serially written by humans. The neural network behind the LLM is data-driven based on typically billions of parameters or weights that were created during training.
And the models are beginning to self-emgineer their own improvements and even successors. When that is fully in motion, WATCH OUT!
The people who poo-poo AI at this point know nothing about it. And even less about what the situation will be five years from now.
Five years eh?
Deal.
In five years, if man is still alive and if woman can survive, we will know who is right.
That’s what they don’t understand. AI may only have the intelligence of a bacteria, but give the objective of self-learning and self-improving, it will get more intelligence as it learns from its mistakes with each new generation.
If you use the concept of a single grain of rice doubling once a day, within 30 days, you’d have a pile of rice bigger than what exists on the earth.
If AI improves even slightly every generation, it will take less time for real intelligence because software generations can be thousands of generations in a day.
And what do you know about it?
It only decides for itself based on humans programming the parameters.
That’s why it creates black Nazis or female Indian WW2 soldiers.