Wiki Contributions

Comments

Nice! Last weekend I expanded https://www.gptrim.com/ to allow the user to a) see savings in both characters and tokens; b) determine their own combination of word processing functions. Then I saw, like you said, that to save tokens you only want to remove stopwords. I will next add the option to remove punctuation. I also want to give users two general recipes: optimize for saving tokens vs. optimize for saving characters. Always happy to take more ideas.

I will probably write again on this, on my new personal Substack or other websites, reporting what I've learned. Would you like me to cite you and link to your profile? My DMs are open! 

P.S: Due to my speedrunning the coding, the website now has an issue where it sometimes adds spaces to the text. I am aware of this and will fix it latest next weekend. The Python package that runs all this is accepting PRs: https://github.com/vlad-ds/gptrim. 

I see your point. I think the existing tokenizer is designed to keep all parts of text, while the idea here is to sacrifice some information in favor of compression. But writing this, I also realized that this approach is more effective at saving characters than tokens.

This is what I was hoping for when I wrote this post. Thank you for your insight. 

New position: sometimes when using ChatGPT, you only care about the number of characters, because of the character limit in the chat message. In that case, you want to get rid of spaces. But if you want to save on tokens, you probably should keep spaces. I think the solution is: a) allow the user to choose the mix of transformations for their use case; b) show them how much they are saving in characters and tokens so they can optimize for their use case. 

Free, universal financial tracker.

I wrote an article on this subject (i.e. why do we play zero-sum games while praising positive-sum games?)

https://native-wonder.blogspot.com/2020/12/things-people-want.html

Thank you, this is very useful. Lately I've been interested in programs that are fully online and could be completed in a year. Would you have any recommendations for that?

Strongly upvoted. As a Kindle-dependent newcomer who's delving into the classics, this is precious.

I have read RAZ. Does this file include it? I would actually need only the posts that are not there.

Do you plan to do this for other authors?

I had trouble understanding how the different facts and judgments in your post are connected between each other and with the concept of upside decay.

But I want to say that I really appreciate the concept, because something very similar occurred to me once, though at the time I didn't give it a name. I was studying the careers of creative artists, and there is a lot of discrimination in these fields. Against women, against people who start out in less prestigious institutions, and so on.

My idea was that because many people were excluded and diversity was stifled, this reduced the probability of "hitting the jackpot" with an extremely brilliant artist that would be the far right of the "artistic potential" curve and end up being the next Picasso. I wanted to model this intuition and verify it in the data, but eventually my project changed and I moved on. The idea, anyway, is that you reduce the chance of getting outliers (or even black swans) in the tails, but you only care about positive outliers.

Thank you for your questions, they're proving very useful.

But it is interesting to understand, what's happening to other children, who actually do math. Suddenly you realize, that "solving problems" for them is less energy demanding, which is awkward!

I'm not sure this is the case. We're humans, maths is hard for everyone. I imagine it's more about developing an ethics of work early on and being willing to delay gratification and experience unpleasant sensations for the purpose of learning something valuable. Though of course it takes a basic level of intelligence to find motivation in intellectual work. And there needs to be some specific motivation as well, i.e. math is beautiful, or math is useful.

As for the other questions... You may be getting closer than me at hitting the target here. I think the comparison between GPT-3 talk, where nothing is wrong, and "manipulation", is central.

But "manipulation" isn't like pattern-after-pattern, it is something different. What is it?

I think the whole thing revolves around mental models. Programming "clicks" when the stuff that you do with the code suddenly turns into a coherent mental model, so that you can even predict the result of an operation that you haven't tried before. I became better at programming after watching a few theoretical computer science classes, because I was more proficient at building mental models of how the different systems worked. Likewise, maths clicks when you move from applying syntactical rules to building mental models of mathematical objects.

It's easier to build mental models with programming, because the models that you're working with are instantiated on a physical support that you can interact with. And because it's harder to fool yourself and easier to get feedback. If you screw up, the computer will stop working and tell you. If you screw up with pen and paper, you might not even realize it.

This is not the whole story, but it's a bit closer to what I meant to say.

Load More