from Diary 2023-03-27
Title: Differences between the true value of “trying to use it” and observers’ perceptions.
Summary: The When the goal is to “use” something new, what it is used for is not so important, but sending it out is a signal to find others who are doing the same thing. People of the same kind will receive this signal correctly, while others will try to find meaning in an “unimportant” object.
As they repeat “unimportant things,” they accumulate know-how, which they then use to begin working on “unimportant things.” At this stage, it appears to observers to stop because it is transmitted in a different way from “unimportant things. It emerges after a while, when it has reached a level of perfection. To the observer, it appears that a leap forward has occurred.
- Here, in the original, “For those who haven’t accumulated the know-how, it looks like a leap forward has occurred. 」だったのだけどステップが短縮されすぎててAIには理解できなかったようだ
Those who found meaning in the “unimportant” object repeat it, whereas those who found meaning in the “use it” object are not interested in the repeated “unimportant” object and think, “I’ve already seen it.” When the reward probability decreases, they hit the button repeatedly because they only see the short-term reward.
While people say that ChatGPT’s use on Twitter is all about grandstanding, those who are a step ahead are running LLMs locally, using vector DBs, and talking about American Equity.
input
nishio It doesn’t matter what you use it for if your goal is to “use it”. Sending it out is a signal to find people who are doing similar things. If the people who see what is transmitted are of the same kind, they will receive the signal normally, but if not, they will try to find meaning in the “unimportant” object.
nishio When you accumulate know-how by doing “unimportant things”, you begin to use it to do “unimportant things”. This is not transmitted in the same way as “unimportant things. It appears to observers to stop. It emerges after a while, when it has reached a level of perfection. To those who have not “accumulated,” it appears as if a leap has been made.
nishio People who have found meaning in the “unimportant” object repeat it, but those who were “trying to use” in that “unimportant” object had found meaning in the “I don’t care” object, so they are not interested in the repeated “I don’t care” object and have “seen it already”. They hit the button repeatedly as the probability of reward gradually decreases.
nishio If you think you have to keep pushing the button to get paid, you are only seeing short-term rewards.
nishio Oh, I started the conversation by saying that the ChatGPT talk on Twitter is full of big happy talk, and I was wondering what kind of signals people are giving off now that I feel they’re a step ahead of me. I was thinking about what kind of signals people who are giving signals that I feel are a step ahead of me are giving now, such as running LLM locally, using vector DBs, and talking about American Equity (I’m way behind on the last one…).
This page is auto-translated from /nishio/「使ってみること」の真の価値と観測者の認識の差異 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.