Scientific and technological progress
is accelerating so quickly that I regularly reassure my readers that the coming
decades will bring unprecedented scientific, technological and economic
progression on a scale so prodigious that they won’t be able to believe it. The
trouble is, given that most people who most urgently need to hear this don’t
read my blog - on account that statistically most people in the world don’t
read my blog - the message of encouragement isn’t getting out there fast
enough.
According to research I’ve read, current trends even in just AI show capacity growing more than 25× per year, vastly outpacing human research growth. Even if these rates slowed by a factor of 100, the combined cognitive labour of humans and AIs would still expand far faster than anything in history, potentially delivering hundreds of years of innovation within a single decade. Rapid gains in computational capacity, algorithmic efficiency, model scaling, and inference costs all contribute to the next phase of what I call the progression explosion, which will trigger a corresponding unprecedented surge in technological development, robotics, and industrial output.
There may, of course, be fresh things to contend with, with such acceleration - like misaligned AI, power concentration, entrenched authoritarianism (which is, alas, already happening), and other challenges posed by advanced digital minds. But they will probably be a spit in the ocean compared with the huge potential benefits - especially extreme abundance, medical breakthroughs, and rapid scientific and material progress.
And as I’ve blogged about before, these advances will create unprecedented possibilities for solving long-standing “future problems” like climate change. If AI-driven research acceleration really does condense centuries of innovation into years, then technologies we are still working on - ultra-efficient batteries, carbon-negative industrial processes, fusion breakthroughs, and advanced materials for energy storage - will arrive with prodigious application far sooner than most people imagine. Rapid scientific iteration, combined with autonomous experimentation, would allow AI systems to explore billions of design possibilities for catalysts, solar materials, carbon-capture membranes, and so forth, before you can say “Greta Thunberg hates cheeseburgers served at BP Garages”.
