I resent this notion that AI doesn't advance if we aren't making new larger and larger foundation models.

Even during that lull between GPT3 and DALL-E/CLIP, there was tons of truly wonderful advances in AI...