52
Are modern LLMs closer to AGI or next word predictor? Where do they fall in this graph with 10 on x-axis being human intelligence.
(programming.dev)
Wondering if Modern LLMs like GPT4, Claude Sonnet and llama 3 are closer to human intelligence or next word predictor. Also not sure if this graph is right way to visualize it.
Now imagine you have 0 knowledge or understanding of anything.
You don’t understand there is a thing called vision. You don’t know about dimensions. You don’t understand that objects exist. That things have shapes, colors perhaps. That there are materials.
You don’t even have eyes per se, or any other sensor. You can’t see.
You don’t understand what thoughts are. What concepts are, nevertheless any concepts themselves.
If I showed you a side-plane picture of a car, directly plugged into your brain, do you think you’d recognize “cars” if we suddenly also gave you eyes and legs and told you walk around downtown?
From the moment you are born you are “trained”. From the moment the first strains of organic molecules closed themselves off from their environment, they’ve been “trained”.
again, whoosh. you missed the part where you train me before asking the question. Then i can extrapolate. And I need very few examples, as little as 1.
I'm talking from the perspective of having actually coded this stuff, not just speculating. A neural network can interpolate, but it sure as hell can't extrapolate anything that was not in its training.
Also, as a human, I can also train myself.