Mastodawn

Arto Vuori Jan 10, 2023

This is a great list. https://lexfridman.com/reading-list/

Reading List - Lex Fridman

Reading List: The following is an ever-evolving list of books I’m hoping to read (or in many cases re-read) in 2023. Classics, sci-fi, nonfiction, or anything people highly recommend. Some are short stories, some are long books. I try to alternate very short books with long ones. Motivation for Selection: I noticed that in the past few years, I started reading a lot more nonfiction, and that it’s been a while since I revisited many of the books that had a big impact on my life (by Dostoevsky, Camus, Hesse, etc.) Also, an obvious giant gap in my reading has

Lex Fridman

Arto Vuori Dec 27, 2022

No better investment for the free world and the law-based world order than Ukrainian military at the moment.

Arto Vuori Dec 9, 2022

Tom Walker Nov 13, 2022

There was once a dream of a decentralised web.

As recently as a decade ago we had a still very active blogosphere, connected via blogrolls and RSS. Specialised web forums were still mainstream and messenger apps could largely interoperate.

Centralised social media slowly ate that dream. It had plenty of positives, but it pulled more and more people away from the open web and into corporate walled gardens.

Some people kept the dream of decentralisation alive. And now you are here.

Arto Vuori Dec 6, 2022

#ChatGPT can be persuaded: it first expressed concern for murdered Russian opposition leaders (when we discussed Russia), but twice refused tell who it was referring to. When I said that it can't just claim something without concrete examples, it finally listed Nemtsov and a couple of more valid examples.

Show thread

Arto Vuori Dec 6, 2022

@Setok it's also cool that it can refactor code based on pretty vague feedback.

Arto Vuori Dec 6, 2022

A snippet from #AndrejKarpathy chat with #LexFridman. Andrej suggests (or makes a hypothesis) that language models may not need special architecture for having long-term memory. It may be enough to explain the idea for them in plain English, and they just may get it.

Arto Vuori Dec 6, 2022

A funny detail on #ChatGPT, if you prompt it not in English to write code, it uses the same language to comment the code as I prompted. I have never seen anyone seriously writing code to comment in any other language than English. Funny that this has gone unnoticed by the neural network. Or maybe I am the one in a minority bubble, not ChatGPT.

Arto Vuori Dec 6, 2022

Had a feeling I have arrived in future when I pasted #ChatGPT generated #Python program (based on a prompt I gave in *Finnish*, write a movement detector using the webcam), and it just worked, and the visualizations were looking cool.

Show thread

Arto Vuori Dec 6, 2022

@kah at least programmatically using openai APIs. You can probably also ask #ChatGPT for a reference implementation. I didn't check though if this version is actually provided yet. According to ChatGPT it is, but it has a tendency to lie.

Arto Vuori Dec 5, 2022

Funny #ChatGPT fact: it makes off-by-one mistakes when doing simple arithmetics with integers.