If we need a breakthrough to reach artificial intelligence, why should we assume it will take any longer than our current path?

There is a persistent comfort in believing that because our current AI models hit a ceiling, we are decades away from anything truly transformative. It is a way of outsourcing our anxiety to a future paradigm shift. But by applying Lindy’s Law to the history of technical breakthroughs, this piece makes a disquieting case: even if we require an entirely new architecture to reach AGI, we might be much closer to that threshold than we think. It is worth sitting with the possibility that the next leap is already hiding in plain sight.

Read at source: Astral Codex Ten