justin_dash
justin_dash on Hacker News
Same here - I suppose this article did what it was supposed to do - give a short primer introducing the topic and if you want to see how they computed this look at the paper!
Unless if it was done by Fable!
Unfortunately I'm using Chrome still.
I use option + up arrow or option + down arrow sometimes, works the same as spacebar to page up / page down.
What's taken over since then? Codex or something else?
So at this point I think it's pretty obvious that RLHFing LLMs to follow instructions causes this.
I'm interested in a loop of ["criticize this code harshly" -> "now implement those changes" -> open new chat, repeat]: If we could graph objective code quality versus iterations, what would that graph look like? I tried it out a couple of times but ran out of Claude usage.
Also, how those results would look like depending on how complete of a set of specs you give it.
I tested both Sonnet and Haiku from Claude, which got it right 0/10 times in their original test, and they both passed. Here's the Haiku output:
"You should *drive*!
The trick is that you need to take your car to the car wash to get it washed. If you walked, your car would still be at home, unclean. So while 50 meters is a short distance that you could walk under normal circumstances, in this case you have to drive because your car is what needs to be washed."
For the sunset example then, a natural question (for me) is then why isn't the sky green in the transition from blue sky to red sunset sky?
I thought of this while working on a front-end project with many different files, and it was getting quite annoying to have to go to each file and copy and paste it in. This also works around upload file limits such as the 5 for Claude. Ironically it was also written entirely by an LLM.