“Max Child believes that while AI is far from AGI in visual, audio, and multi-step tasks, text-in/text-out models could reach AGI-level thinking within the next two or three model generations.”