3 ms·
As the other reply has said, the article points to limitations of LLMs, but that doesn't preclude synthesizing a system of multiple components that uses LLMs. T
by jhanschoo 2y ago
As the other reply has said, the article points to limitations of LLMs, but that doesn't preclude synthesizing a system of multiple components that uses LLMs. To the extent that I'm bearish on AI capabilities, I'll note that program synthesis / compression / general inductive reasoning which we expect intelligent agents to do is a computationally very hard problem.