Post by Andreessen Horowitz

The dominant paradigm in AI today is organized around language and code. But a different set of fields has been maturing quietly in parallel: robot learning, autonomous science, and new human-machine interfaces. They sit in exactly the right position relative to the incumbent paradigm. Close enough to inherit its infrastructure and research momentum. Far enough to require non-trivial additional work — which creates a natural moat and more room for emergent capabilities. The pace of progress over the past 18 months suggests these fields could soon enter a scaling regime of their own. Read the full piece linked in the comments.