2 ms·
Totally agree that there are better ways to build a program to beat tic tac toe. I'd expect an LLM could probably write the code itself with a few turns of the
by d_legs 2y ago
Totally agree that there are better ways to build a program to beat tic tac toe. I'd expect an LLM could probably write the code itself with a few turns of the crank (it wrote most of the code I used to test this). The point here is to test the G in AGI: how well models can generalize to new tasks.
- beardyw 2y agoMy quibble is with the use of language models. Human intelligence is not predicated on the written word. You wouldn't use an LLM to create something that plays tennis, yet our human general intelligence covers that.