Find big use cases with the smartest models first, then trade intelligence for efficiency
CompleteSkeptic · x · 2026-09-16
The author argues the right sequence is to first discover major use cases with really smart models; only then does it become practical to study how much intelligence can be traded for efficiency to make models small enough to run locally.
More from AGI Musings
- Ben Antieau guest post on Terence Tao's blog: mathematics needs both 'fast math' and 'slow math' in the LLM era — littmath · 2026-09-16
- Researcher warns merging with off-the-shelf AI will drive cultural homogenization — nptacek · 2026-09-16
- ASI x-risk estimates ignore externalities we impose on alien civilizations — EigenGender · 2026-09-16
- Censorship Couldn't Kill Literature — the Post-Literate Society Might — julianweisser · 2026-09-16
- iamtrask: trust networks could unlock a billion times more data than AI trains on — iamtrask · 2026-09-16
- iamtrask: decentralized AI would beat centralized systems on quality and cost — iamtrask · 2026-09-16