Hacker Newsnew | past | comments | ask | show | jobs | submit | My_Name's commentslogin

Well, #2 is something that historically has regularly been required by the powers that be, so the addition of #1 as another option is actually an improvement.

I'm hopeful that if things get bad enough Option #1 becomes the preferred reality. Not holding my breath (for here in america)...

Oh. I thought they were going to use AI to create an artificial Steve Jobs.

Never mind.


I have the opposite reaction, when it is confident, or says I am right, I accuse it of guessing to see what it says.

I say "I think you are getting me to chase a guess, are you guessing?"

90% of the time it says "Yes, honestly I am. Let me think more carefully."

That was a copypasta from a chat just this morning


it's mostly just agreeing with you (that yes, it was guessing). LLMs have very limited ability to even know whether it was guessing. But it can "cheat" and just say yes it was if it seems like that's what you expect to hear.

The LLM simply agrees with you and you're happy. It is VERY worrying that you don't realize this, even after reading this article.

Hulk: I'm always angry

AI: I'm always guessing


and it doesn’t actually think more carefully

these things are incapable of thinking, no matter what the UI and marketing calls it


Judging from the number of comments, it seems that this follows the universal rule of 'yes/no question in the headline' where the answer is always 'no'.


There seems to be a focus on understanding when talking about LLMs and solving problems. Personally, I do not think understanding is required. I can write a very small program that can calculate Pi to however many digits I like, or calculate any digit in the sequence on demand, without the program or computer having any understanding at all of what Pi is or what it means. I could get Claude to output that same code when prompted to find a solution to generating Pi, also with no understanding of what Pi is, or what it means.

IMO the ability to provide an accurate solution to a problem is not always based on understanding the problem.


You can easily get compressed episodes of a TV show that are 250MB, so it's like watching a TV series at the rate of 2 episodes every 5 minutes. Obviously better quality is in the range 500MB-1.5GB for a 45-minute episode, so even being generous it's 20 minutes of compressed TV or 70 minutes of uncompressed music every 5 minutes.

Just for ads on a website.


And this is why I pi-hole (although I am thinking of changing to technitium)


After working in ad tech for a few years I am fully loaded up on ad blockers


This is why I have a pi-hole and a selection of addons for my browser.

I am considering moving to technitium though, it seems better featured.


So just get the AI to summarise the codebase giving you more time to design a better buggy whip.


Of course, if you use AI, it's very hard to know the sources used for training that went into the output you just got.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: