3 ms·
Debates on the nature of artificial general intelligence
- throw310822 2y agoLargely misses the mark, imho. The term AGI is simply badly conceived, as it mixes a quality (the generality) with a quantity (the human level); and in doing so leaves an enormous gap between narrow AIs (e.g. a chess-playing computer, a speech recognition algorithm) and the lowest-recognized "general intelligence", already flying at human level. As for LLMs having a simplistic reward function (optimizing for the best possible choice of the next word)- how is humans' reward function (pleasure, in its various forms) any better?