isakrynell :
LLMs like ChatGPT are not true AI, they are statistics applied to language. It’s about reproducing patterns found in the training data. The intelligence lies in the patterns themselves, and those patterns were made by human brains.
The biggest problem AI has is sifting out relevant data from noisy, open-ended, and incomplete data. For AI to do what we want, to reach human-level intelligence, it has to be able to do this at least as well as a human.
We are good at this because finding relevant information in noisy, open-ended, and incomplete data is why life developed a nervous system in the first place. Evolution did something very similar to what our training algorithms do in computers, but to get the “sifting software” running between your ears, it ran that algorithm on trillions of test subjects with full-spectrum sensory input, across an entire planet, for four billion years.
We have nothing even in the ballpark of that. We don’t have a planet-wide evolutionary system to run on, and we don’t have billions of years to do it. We are nowhere close to real human-level AI, and that’s the hard truth these companies don’t want to accept.
2025-11-06 04:25:03