4 ms·
It's quite possible a human forms a full representation of a coherent thought before translating that thought into words, meaning no token by token prediction.
by RhodesianHunter 3y ago
It's quite possible a human forms a full representation of a coherent thought before translating that thought into words, meaning no token by token prediction.
- adam_arthur 3y agoAnd digital audio will never match analog. Can anybody tell the difference anymore? If a machine produces highly similar output does it matter? LLM's clearly exhibit behavior of having implicitly learned systems, which is the valuable part of intelligence. Humans infer systems through text too, by the way.
- RhodesianHunter 3y agoWe're not at the point where people can't tell the difference though. A human will often say "I don't know". The machine will outright lie using it's best probabilistic guess.
- doctor_eval 3y agoIt’s also possible that translating a coherent thought into words is done on a token by token basis. I’m surely not the only one who sometimes can’t “find the right word” in the middle of a sentence when trying to describe a thought or idea.
- RhodesianHunter 3y agoRight, but that would still be functionally quite different from what an LLM does, given that it has no full thought and is entirely predictive by token.