First of three on AGI per se

Let me start today's #DailyPlanet with a stupid lexical game: what's the difference between AGI and AI?

The answer, of course, is G. haha. Not really, but there's a general point lurking behind this stupid game. That general point has to do with the "general" in Artificial General Intelligence. Whether we reach AGI - in the sense of intelligence that meets or surpasses comparable human capabilities - or it takes us decades, we should never underestimate the bitter lesson:

One thing that should be learned from the bitter lesson is the great power of general purpose methods, of methods that continue to scale with increased computation even as the available computation becomes very great. The two methods that seem to scale arbitrarily in this way are search and learning.

Should this increased computation be made available in a distributed fashion, embodied in hardware everywhere (and then networked) or should it be centralized in massive data centers and distributed virtually? Sutton has some thoughts to share here too:

The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. All these are part of the arbitrary, intrinsically-complex, outside world. They are not what should be built in, as their complexity is endless; instead we should build in only the meta-methods that can find and capture this arbitrary complexity.

I guess the basic question is: should that irredeemable complexity of the world be sampled and acted upon locally or should it only be sensed locally and processed centrally before being acted upon locally. The hyperscalers think the latter. The Chinese think the former. American or Chinese, we can think of two generalist pathways, that for the lack of a better (bitter?) alternative, I will call:

• General Predictive Intelligence (GPreI), aka, The American Way of AI

• General Productive Intelligence (GProI), aka, The Chinese Way of AI

The next two Daily Planets will be a very light exploration of GPreI and GProI.