The cool demos from OpenAI, Figure and the like make us hallucinate a future that will take much (much) longer to pan out because they ignore the domain-specific knowledge that is inherent to the domain they pretend to disrupt.

I’ll be impressed when ILM talks about it.