I don't think anyone can meaningfully predict the AI bubble. Expert predictions go anywhere from galactic colonization bubble (do they collapse at the edge of the galaxy as resources get scarce, or continue on to the rest of the universe?) to whole-world GDP bubbles, to big tech bubbles, startup bubbles, etc. I haven't seen convincing evidence for a probability density that isn't fairly smeared over all these options and wide timescales.

Unlike traditional bubbles we get qualitatively different outcomes at greater scale. A million GPT-2s is nothing like 1 GPT-4. Researchers are continually surprised by new abilities at greater scales.