Brockism or potentially Overhang Reductionism (see discussion in comments) is a proposed name for one of
four viewpoints represented in the famous
2023 societal debate about AGI
safety taking place at OpenAI. Thankfully, all four factions agree on the need to deal with x-risk, but disagree about how:
(1) The "normal" faction, which includes
Satya Nadella and almost every businessperson both in VC and on Wall
Street. Normals say (at least with their investment decisions, which speak infinitely louder than
words) that we can deal with x-risk
later.
(2) The "decel" faction (short for "decelerate"), which says to slow down AI research.
(3) The "e/acc" faction (short for "effective accelerationists") is a trendy, recent term for optimistic
techno-utopianism, in the milieu of Vernor Vinge's stories.
(4) The "Brockist" faction (named after Greg Brockman). Brockists (which may or may not include Brockman himself, as the idea was inspired by him but his own views have yet to be verified) believe that the way to reduce x-risk is to accelerate AI
software research while halting or slowing semiconductor development. They believe that if chips are too
fast, we could stumble into unwantedly making an unaligned artificial superintelligence by accidentally inventing an algorithm that makes fuller use of existing chips. The difference between what we currently do with current chips vs what we *could* do with current chips is what Brockists call the "capabilities overhang".
Significant circumstantial evidence for the Brockism may be found in the achievements of the retro-computing "demoscene", which uses innovative
software to produce
computer graphics on par with the late 1990's on some of the very oldest
personal computers.
