• 0 Posts
  • 4 Comments
Joined 1 year ago
cake
Cake day: November 17th, 2024

help-circle
  • In a free country, you can tell the government to go fuck itself without a mask

    Political infrastructure works well until it is not. U.S. used to have okay political infrastructure in protecting democracy, then patriot’s act happened and many of its loophole identified, now president can just kidnap a foreign president as “law enforcement”.

    I would love a system where people don’t have any need to be anynomized, it would make many things much simpler, but that seems hard to imagine for me. And I am not from the U.S. and I have lived in both U.S., U.K., and outside of the west, so it is likely not caused by “U.S. brainwashing”.

    I am not entirely sure what is the “EU secret sauce” to prevent Politician in utilizing these loopholes or strong centiments to gradually regulate speech. One day, they might be able to make use of these data. People in U.S. protested, they shot protester, and no one can protest forever, unfortunately. I am curious what would prevent EU to replay what US have now, except with much much more targeted data at the government’s disposal.



  • I like how Yaron Minsky from Jane Street characterized LLM: “It is smarter than we expected it to be, but dumber than we needed it to be… It feels like something really dumb, but somehow memorized the entire internet.”

    This is kind of what I feel: despite all these impressive BAR and IMO achievements, in my work, I feel they do a great job at parapherasing the internet, but fails when you need it to do something mildly intelligent.

    Does it improve my efficiency? yes, but only at some very tedious and specific taskes, once I go slightly out of scope, it comes up with inelegant solution that I will need to rewrite from scratch.


  • The image is taken from Zhihu, a Chinese Quora-like site.

    The prompt is talking about give a design of a certain app, and the response seems to talk about some suggested pages. So it doesn’t seem to reflect the text.

    But this in general aligns with my experience coding with llm. I was trying to upgrade my eslint from 8 to 9, and ask chatgpt to convert my eslint file, and it proceed to spit out complete garbage.

    I thought this would be a good task for llm because eslint config is very common and well-documented, and the transformation is very mechanical, but it just cannot do it. So I proceed to read the documents and finished the migration in a couple hour…