Hans Moravec’s 1998 graph of computer performance evolution has surfaced in the Twittersphere (via Hillary Haley). It’s sixteen years old now, but the story it tells hasn’t shifted much (which means the climax is quite a bit closer).
What’s happened to the curve? According to this account, it has leveled off significantly since 2002, but it was never easy to fix on exactly what to quantify. MIPS is generally derided as a metric, in part due to simple quantitative obsolescence (exceeding three orders of magnitude since 1998).
Moravec’s brutally quantitative, hardware determinism remains a credible predictive tool, however, especially if unplanned emergent effects are expected to dominate (overwhelming software engineering). Once history has thrown up enough synthetic brain capacity, things can begin to move in.