4 ms·
Humans often do not have a clear definition of success and instead create a post-hoc narrative to describe whatever happened as success.
by mnky9800n 2y ago
Humans often do not have a clear definition of success and instead create a post-hoc narrative to describe whatever happened as success.
- cadamsdotcom 2y agoContinuous RL in a sense. There maybe an undiscovered additional scaling law around models doing what you describe; continuous LLM-as-self-judge, if you will. Provided it can be determined why a user ended the chat, which may turn out to be possible in some subset of conversations.
- ahartmetz 2y agoI'm not following. Do you have an example?
- mnky9800n 2y agoThe milliken oil drop experiment, “winning “ the space race, mostly anything C levels will tell the board and shareholders at a shareholder meeting, the American wars in Iraq and Afghanistan, most of what Sam Altman or Elon musk has to say, this list continues.
- Yiin 2y agoI think you're approaching it form very high level, when you should think about it from much lower level, i.e. success is being determined by stress/dopamine hormones or similar
- d0mine 2y agoThe lower level seems to work eg, “Dopamine regulates decision thresholds in human reinforcement learning” https://www.nature.com/articles/s41467-023-41130-y https://www.nature.com/articles/s41467-023-41130-y
- Splinelinus 2y agoThis article is kind of vague on that tbf: To conclude, we observed no credible evidence for a beneficial effect of L-dopa (vs. Haloperidol) on reinforcement learning in a reward context, as well as the proposed mechanistic account of an enhanced striatal prediction error response mediating this effect.
- mnky9800n 2y agoSo your claim is that success stories are some kind of emergent phenomenon that comes from people chasing dopamine highs?
- echoangle 2y agoIs that controversial? I would say everything a human does is to feel better, and everything someone does that doesn’t make them feel better immediately is just done in the expectation of even greater pleasure later.
- mnky9800n 2y agoI don’t think the human experience can be reduced down to “if feels good is true then continue else try something else”. But I could be wrong.
- echoangle 2y agoWell mine can, with some tactics and strategy layered on top. If I do something I don’t like, I only do it because the payoff later makes it worth it (or at least I think it will from my current knowledge).
- earnestinger 2y agoWhile humans seek “profit”. It is important that “profit”, comes in various forms, which exchange rates are problematic to calculate (or maybe there can’t be any): not hungry, not thirsty, tastes good, not cold, feel safe, feel excited, feel righteous, feel powerful, listen to music, watch a movie, get curious, satisfy curiosity, laugh, love, sex, rock n roll.
- dutchbookmaker 2y agoI think it is both and this is very obvious if you read Pareto and Sapolsky.
- dutchbookmaker 2y agoIt is more than just success too. Most behavior we believe is some kind of rational action when it is really blind actions based on fiction or just completely random with rationalizations for the behavior after the fact.
- mnky9800n 2y agosomeone once told me > there are the facts we know and the stories we tell ourselves and that really changed my outlook on life, the way i think about things, and basically everything else.
- ben_w 2y agoAesop's fables, "sour grapes".
- ahartmetz 2y agoI did think of sour grapes (only thing that came to my mind) and was hoping for something better. Sour grapes doesn't seem too interesting. I think most people can tell you, if you actually ask them, what the differences between their actual successes and their wildest dreams are. But any improvement is still a success and I think that's valid.
- emsign 2y agoThis!
- ben_w 2y agoAnd also sometimes write down the conclusion and work backwards, without considering that the reason most likely for the conclusion isn't necessarily going to have the conclusion as the most likely conclusion — I hope I phrased that broken symmetry correctly.
- _heimdall 2y agoCompanies are bad about doing this on purpose. If they set out to build AGI and accomplish something novel, just call that AI and go on fund raising from people who don't know better (or more likely don't care and just want to gamble with others' money).
- idiotsecant 2y agoYes, I've always thought that LLMs need the equivalent of a limbic system. This is how we solved this problem in organic computers. There is no static 'reward function'. Instead, we have a dynamic reward function computer. It decides from day to day and hour to hour what our basic objectives are. It also crucially handles emotional 'tagging' of memory. Memories that we store are proportionally more likely to be retrieved under similar emotional conditions. It helps to filter relevant memories, which is something LLMs definitely could use. I think the equivalent of an LLM limbic system is more or less the missing piece for AGI. Now, how you'd go about making one of those I have no idea. How does one construct an emotional state space?