Difference between revisions of "Secret sauce for intelligence vs specialization in intelligence"
(Created page with "what is the relationship between the "you can't specialize in 'intelligence'" argument and "there are a small number of insights for AGI" (a.k.a. secret sauce for intelligen...") |
|||
| Line 3: | Line 3: | ||
if there are only a few insights till AGI (and assuming that compute isn't a bottleneck), then you can specialize in intelligence. | if there are only a few insights till AGI (and assuming that compute isn't a bottleneck), then you can specialize in intelligence. | ||
| − | if there are many insights till AGI (and there is no 'last missing gear'/'one wrong number' dynamic, AND each insight makes the AI a little better), then you can't specialize in "intelligence". | + | if there are many insights till AGI (and there is no 'last [[missing gear]]'/'one wrong number' dynamic, AND each insight makes the AI a little better), then you can't specialize in "intelligence". |
[[Category:AI safety]] | [[Category:AI safety]] | ||
Latest revision as of 23:01, 6 July 2020
what is the relationship between the "you can't specialize in 'intelligence'" argument and "there are a small number of insights for AGI" (a.k.a. secret sauce for intelligence) argument? is it ok to conflate these two?
if there are only a few insights till AGI (and assuming that compute isn't a bottleneck), then you can specialize in intelligence.
if there are many insights till AGI (and there is no 'last missing gear'/'one wrong number' dynamic, AND each insight makes the AI a little better), then you can't specialize in "intelligence".
