I wouldn't call myself a historian, but I have been doing a history podcast since 2014.
I agree that Ben's writings on LLMs and how they impact the humanities/history are great reads. But I am also the perfect target market for that kind of discussion, dev by day amateur historian by night.
I think it appears to have done a good job of summarizing the points that it summarize, at least judging from my quick watch of a few sections and from the YT Transcript (which seems quite accurate).
Almost makes me wonder if it is behind the scenes doing something similar to: rough transcript -> Summaries -> transcript with timecodes (runs out of context) -> throws timestamps that it has on summaries.
I would be very curious to see if it does better on something like an hour long chunk of audio, to see if it is just some sort of context issue. Or if this same audio was fed to it in say 45 minute chunks to see if the timestamps fix themselves.
Seems particularly funny in an article about Emacs, a piece of software that lets you get in situations where some portion of your "just create" time becomes "managing my custom emacs, please don't break".
One way to look it is to approach it as a creative practice. A good part of any practice is devoted to developing technique.
Some are just fine with a standardized but unoptimized tool while others are fascinated by building their own high-flying TUI. The journey is the destination. If all you create is a config file, it still counts.
Reminds me of the issue with bad CGI in movies. The only CGI you notice is the bad CGI, the good stuff just works. Same for AI generated art, you see the bad stuff but do not realize when you see a good one.
care to give me some examples from youtube ? I am talking about videos that ppl on youtube connected to for the content in the video ( not AI demo videos).
What makes draws interesting or not depends on two things, what is happening on the field (were there good chances? A lot of goals? Interesting strategic play?), and also the context of the game. Does the outcome matter for your team, or not?
Some of the best games I've ever watched ended in draws, but that point mattered for whether my team was going to take over first place or not, or whether they made it out of the group stage or not.
The really boring draws are when one team is basically just playing very defensively and also the game doesn't matter to either team.
I was going to mention Warp here as well. It is fantastic when it comes to almost anything in the terminal. It has caused me to use the terminal a lot more on all of my computers because I don't have to spend a bunch of time poking around on Google to find the command to run.
I have used it for ffmpeg and then a lot of other slightly more complex commands. A recent one from the other day was gathering up all of the .epub documents in a directory tree, renaming them to the name of the directory they were in, and then placing them all in one single directory. That would have been a whole project for me, and Warp gave me the command with just that description. Any LLM interface would have done the same, but Warp just let me hit "Enter" and run it, no need to copy and paste.
Same - I also read a study saying 22% muscle synthesis increase with 40 grams of casein before bed and one saying protein before bed improved sleep quality.
Yes - unless you have to wake up to pee I guess. Casein is one of the slowest absorbed proteins so it releases over 6 hours in most cases. I think soy beans is probably the same since its mixed in with fiber. If you compare it to whey isolate thats absorbed after 1. This doesn’t have the same effect.
Given the pace of development in this space, it is probably worth noting in the title that this is from November 2024 so the results might be a bit dated.
Author here. It's a fair point and it would be worth revisiting, if not now then within the year.
That said, I wouldn't expect things to change too drastically. TFA goes into details, but in short LLMs are already quite good at whiteboarding (where you interactively describe the diagram you want). They're also really bad at generating a diagram from an existing system. In either case, small, incremental improvements won't really help; you'd need a large change to move the needle.
I agree that Ben's writings on LLMs and how they impact the humanities/history are great reads. But I am also the perfect target market for that kind of discussion, dev by day amateur historian by night.
reply