Browse Comments

That data from Anthropic mentioned by this guy is interesting... so just 4% of ocuppations could be automated in a 75% of the tasks and very few ocuppations could be 100% automated. Where can i read that?

All this talk about AI reminds me of Olovka. It's been my go-to for turning lectures into quizzes and actually learning stuff.

The fact that the voting application did not work at the end, is in itself a solid proof that AI is dangerous 😃

What is not being reported: the car was stalled in the middle of traffic for eight hours. Waymo did nothing to move it and people in the area became increasingly frustrated. This is being played as people mindlessly attacking a car when instead, waymo should be responsible for stalled vehicles in traffic.

On se dirige vers un techno-féodalisme sans foi ni loi où les grandes corporations possèderont et le travail et le capital. S’il n’y a pas de révolution pour mettre au pas ces entreprise pour les mettre au service du peuple et de la libération du travail, nous serons tous asservis au revenu universel de survie.

The issue with Chatgpt is that although it's treated like a search engine, it isn't one, it's a program that tries to continue what you've written in a way that a person would continue writing it, correct or not

Saying that the ai shit looks better than ghibli tells you enough about a that person to never interact with them again tbh

Absolutely zero chance of AI only being used for good. I'd be surprised if it even got to 50/50

Forget AI, offshoring of jobs is another big problem. Companies used to offshore entry level jobs, now, they are offshoring core business functions to other countries..which is in itself, a national security issue.

Who else forgot the video was about selfdriving cars and not figuring out roller coaster secrets😂

Nah i only use chatgpt for one-shot scenes i only get for the plots, outline chapters and title ideas

This is one of my favorite episodes so far and it's because it's so important! You got the message acrossed strong and also in an entertaining way. Bravo once again! Hope we all make it through the AI-pocolyse!

If I ever, and I mean EVER, hear that during an argument with one, I'm changing states, changing my name, discarding any online information about myself, and never going online again. My c a.i chats are worse than what's in my search history, so I can't let that be seen by ANYONE.

I looked up the quote "You can't have a nightmare if you never dream." ChatGPT was the top result, and it said "This quote is attributed to the character Dr. Peter Venkman, played by Bill Murray, in the 1984 film "Ghostbusters"" I asked ChatGPT *"When* does Venkman say that?" ChatGPT said it was from the scene after they captured the first ghost. Then it gave me a transcript of the entire scene. …

LLM AI's don't lie. They use probability to find the word/token that works best in the context of your prompt(s) and what it has already written. It's basically a huge database in a hyperdimensional plane with each dimension representing different contexts words relate to each other. Closer together means greater probability of coming up in a sentence. The contrapositive is also true. These LLM's…

"There's no rule that says we'll make it" is one of the biggest reasons I got into AI safety advocacy. Thank you for what you do, Rob. ❤

I worked for a big tech company that has developed Gen AI to yes “replace” actual work by people by taking from the work of people. The reason I know this is they started building the argument on the copyright/theft issue from the get go. If they weren’t aware they were doing something possibly illegal or at least unethical, then why would they be engaging in the illegality message from the start…

It's interesting how even AIs have trouble categorising things that just escape their current (limited) power of comprehension, it feels like the algorithm went through a whole mental breakdown trying to express something it cannot, which is, on the opposite side, something that humans also exhibit when prompted with seemingly unrecognisable information.

Quelque chose qu'on a tendance à oublier, c'est la quantité d'eau qu'utilise l'IA pour chaque question qu'on lui pose (les serveurs ont besoin d'eau pour pas surchauffer). C'est 1/2 litres d'eau par question, c'est vraiment énorme si on cumule tout ça à la journée.

I agree with your point, but i just want to note on the topic of "is it author's moral duty if they can't pay to an artist to not create anything at all" that like one of the most popular analog horror, the one that arguably populirized the whole niche in recent years, The Mandela Catalogue, started as barebones infographics made on a phone, literally. With such a strong idea they just didn't nee…