They are suggesting that Ilya Sutskever played in this coup.
787 karma · joined October 13, 2022
They are suggesting that Ilya Sutskever played in this coup.
I believe that if someone with a patreon wants their identity to stay private, making an LLC (depending on the name of it) may backfire?
That much isn't necessary, GPT4 api costs $0.03-$0.06 for what's more than an essay of text.
It's affordable and accessible to anyone with a credit card. Ethically this is more fair than having to pay for over-priced textbooks.
I didn't see [1] Game Programming Patterns.
Not disagreeing but Free Download Manager stopped being OSS in version 3.9.7, that was 2017 and this article's infected package was released in 2020.
The bright side is that it's easy to move to a new server.
Ironically it's easier to get into deep work at home because you don't have to worry about coworkers bouncing ideas on you.
Why did you decide to use an embedded JavaScript engine instead of pure go?
The rate limit for Deepgram's Nova speech-to-text model is 100 concurrent requests on the pay as you go plan. They claim it's more accurate than their Whisper model but I'm trying out both.
> Possible answer to the Fermi paradox: the intelligent life self destructs.
Although if robotic AI girlfriends become more real than real by using various kinds of supernormal stimuli [1], to the point where actual humans look fake, I can see this happening.
[1] https://en.wikipedia.org/wiki/Supernormal_stimulus
> It is possible to create artificial bird eggs which certain birds will prefer over their own eggs
Teachers don't have the time for that for a fleet of chromebooks.
A better comparison is this vs artstation.
It has a Subreddit List page and sorts subreddits by niche. I like to know what people of a niche or fandom are actually reading instead of what books amazon or advertisements promote.
I've found hidden gems in hobbies I didn't know existed.
The first link uses karma to rank the books, the second uses total comment count.
Unfortunately the reddit site does not show the comments, unlike https://hackernewsbooks.com/
Could be because they haven't actually tried it. There are a surprising amount of applications something that tells wrong answers part of the time can do.
> "I'm a language model, so I can't..."
You won't catch the clever students who programmically remove these (e.g. using Langchain).
I find that prompting chatgpt with book excerpts keeps it grounded in reality.
> For a couple of years, I taught PIM in a lecture at the St. Pölten University of Applied Sciences. My students are surprised, how inefficient they were doing even «common» things like handling email, managing files like photographs on your computer, or searching for information. It is so rewarding to see students realizing how much improvement is possible with some background-knowledge and a few hints here or there. Starting with 2021, that lecture started at Graz University of Technology.
> I hereby claim that I am able to save you at least twenty minutes of your computer time a day just by looking over your shoulder and suggesting more advanced tools, better methods, and efficient workflows.