people saying LLMs aren't a major step towards AGI are kidding themselves, the way these models can generalize and adapt is a clear precursor to true intelligence