“I think I hate [the scientist] because deep down I suspect he may be right. That what he claims is true. That science has now proved beyond doubt there’s nothing so unique…nothing there our modern tools can’t excavate, copy, transfer. That people have been living with one another all this time, centuries, loving and hating each other, and all on a mistaken premise. A kind of superstition we kept going while we didn’t know better.”
— Klara and the Sun, Kazuo Ishiguro

The fear that human uniqueness could be reduced to something reproducible I think captures much of today’s anxiety around AI.

After my fretful post on the rise of AI in generative art, some signs seem to be emerging in Big Tech that might indicate the AI juggernaut may be slowing, or at least recalibrating.

• ChatGPT-5 launched with much fanfare, but many users came away underwhelmed, noting incremental improvements rather than a leap forward.
• Meta has also restructured its AI division, breaking it into smaller departments and reducing staff—moves that rattled investors and knocked confidence in its stock.
• Meanwhile, analysis from Epoch AI (via TechCrunch) suggests that gains from reasoning models—particularly those trained with reinforcement learning—could plateau within a year.
• I personally find 3D generation tools like TripoAI amazing, although they are still act as a springboard for further retopology and refinement, as this feedback suggests. Their recent v3 model release still does not negate the need for these steps.

Taken together, these shifts may hint at the more practical boundaries of AI’s potential.

An AI slowdown may comfort those worried about job displacement. However, I think we will never feel quite as unique as we once were.