diff --git a/good_ideas.myco b/good_ideas.myco index 8d3c53e..a67be1c 100644 --- a/good_ideas.myco +++ b/good_ideas.myco @@ -2,8 +2,11 @@ * Meme sparse autoencoding (I think a CLIP SAE already exists though) * Overengineered LLM-based autocomplete to spite "it's just autocomplete" people. * Expanded [[vengeance]] policy. -* Combine new CRL (https://arxiv.org/abs/2408.05804) with offline pretraining. -* Similarly, contrastive RL for computer algebra (specifically, proving that expressions equal other expressions via making substitutions repeatedly). Try and contrastively learn a "how close is this expression to this other one" function (I think with an action input?). Bootstrap to progressively harder problems. +* Combine new CRL (https://arxiv.org/abs/2408.05804) with offline pretraining. Might be redundant in some sense. +* { +Similarly, contrastive RL for computer algebra (specifically, proving that expressions equal other expressions via making substitutions repeatedly). Try and contrastively learn a "how close is this expression to this other one" function (I think with an action input?). Bootstrap to progressively harder problems. +* What does Gyges mean by "anyone who wants it: you should be able to train contrastive models way faster if you use lsh to determine pairs to contrast"? This might contain alpha. +} * { Startup ideas: * Automated reminders to make spontaneous gestures to maintain friendships. @@ -26,5 +29,8 @@ Next-action-predictor editors/UI. * Possibly just for prefetching/preloading. } * Do WiFi sensing but good (with more data (https://b.osmarks.net/o/a0bddc9b742b4efbba18600ae6d51d98)). -* Comparisons rather than scalar ratings for things (https://b.osmarks.net/o/60b26f1735134d628164217be52ca2d3) +* { +Comparisons rather than scalar ratings for things (https://b.osmarks.net/o/60b26f1735134d628164217be52ca2d3). +* It shouldn't be that hard to aggregate all the things I describe positively or negatively on my website and build a thing to allow me to rate pairs. +} * Constrained "LLM agent" (how I dislike this terminology) which chooses things to buy for you (https://thezvi.substack.com/p/choices-are-bad etc). \ No newline at end of file