"1984, but with LLM's"

#AI #LLM #dystopia
#manipulation

https://garymarcus.substack.com/p/1984-but-with-llms

#fediverse
https://youtu.be/YRJHIJy5Nno?si=UjbWzrj-8Wc7laCR

@TheFox21 juniors or seniors alike, in my opinion. Vibecoding is definitely wrong, but it doesn't take long to realize that. "Honest" AI-use, however...
I think you could compare it to importing an library. Do you want to rewrite JSON parsing for every project? Do you need to understand every single line of the gson library?
But at the same time, if a new vulnerability is found for GSON, what do you do then?

My take is that AI-use is alright as long as you understand most of what the AI wrote

@cjkarr.bsky.social did you feel that you had a choice before AI? Playing devils advocate here, but come to think of it, I dont think anyone ever got to decide if a new tech was becoming mainstream or not.
- You see this wheel? It's the future !
- Noooo I want to carry my stones by hand, that's my choice
To be fair, there weren't any tech bros when the wheel was invented and capitalism was probably not so dominant and deciding everything for everyone

If you want to drive #MAGA asshats mad, claim that #Trump #tariffs are actually his #4DChess strategy to fight #climateChange —no more #China / #USA shipping lanes will lead to drastic cuts in #pollution after all!

Ironically, before #LLM s were a thing, I already expected #Twitter administration to go rogue and posted thousands of randomly generated tweets on my account using a Marklov chain. I hope every #AI trained on my tweets start outputting non-sense.

@AndrewRadev you can always self host your AI. I do, it works great, props to ollama. But to be fair, they had your proprietary code decades ago, dont you think?

@AndrewRadev copy pasting from SO makes you a bad developer if you don't (fully?) understand and can explain to someone else what you're copy pasting. Copy pasting from chatgpt makes you a bad developer if you dont (fully?) understand and can explain to someone else what you're copy pasting.
The same goes with a programming book, an obscure hacking forum from 2001 auto translated from Russian, a YouTube video, you name it. It's true that knowing the source can help, but that's about it.

@lauren Is there a stream?

@PrivacyDigest So the Chinese are using chatgpt for mass surveillance ? How do we know ? With American mass surveillance ROFL

@paco While mine is still too young for #docker and #selfhosting, I utilise containers any chance I get.

Spread the knowledge Paco! 🤘🏻

Chères et chers camarades socialistes

Chères et chers camarades #socialistes, où allez-vous? Quel est votre cap? Je vous interroge parce que je redoute que, à nouveau, votre boussole se brouille, et n’indique plus clairement notre nord: du changement pour les gens, et contre l’argent.

✒️ Par Ruffin François | En accès libre › https://blogs.mediapart.fr/ruffin-francois/blog/240125/cheres-et-chers-camarades-socialistes?at_medium=rs-cm&at_campaign=mastodon&at_account=mediapartblogs

@BenjaminHCCarr https://indieweb.social/@azzale/113593316899119630
Generative ai and chatgpt are two different things. Chatgpt does generative ai. You can do generative ai without chatgpt.
If you generate content on your own PC at home, how much water is used in comparison to a google search ?
Is it even possible technically to have a computer at home that would run a google search ? Nope. Yay for generative ai at home without internet, less energy/water than a google search!
Check out ollama.com

@schymans Googling to get an answer without having to read the whole book 15 years ago made us already quite lazy. But AI really is next level lazy. I find it scary.
And yes, 15 years ago, people were already complaining that one google search was burning so many trees and using so many resources. AI is nothing new in that regard, just more of the same. Except... Except that you can run it yourself. No trees burnt (not more than playing a game). And... Can you run a google search on your PC?

@schymans it's just an example. I see many people critic #AI but their argument is too often "this or that big tech company will own us", "energy wasted", or "so much power in the hands of one lunatic". While those critics are all true and should be addressed, AI can also be installed locally on your own PC. And then all of those claims become invalid. I think #AI is dangerous, but not for the reasons I see on mastodon. Mostly because it makes us lazier, I think?

@schymans what if you install AI on your own computer and use your own GPU? Definitely using less resources than a google search. You won't get good results if you ask a question either. But then again, Google is also biased (search for information bubble) and also gives wrong results (maybe less often). Check out ollama.com

"Can't find a good job?
Blame bosses not immigrants"

Poster spotted in Toronto, Ontario

@superpixel Imagine what world we would be in if 20 years ago if those fine prints were also on google search results

[Article] Julian Assange parle, le Conseil de l’Europe écoute et approuve

Interrompant exceptionnellement sa convalescence en Australie, le fondateur de WikiLeaks est venu à Strasbourg le 1er octobre pour témoigner devant le Conseil de l’Europe. C’était sa première intervention publique depuis son emprisonnement en 2019.
https://www.blast-info.fr/articles/2024/julian-assange-parle-le-conseil-de-leurope-ecoute-et-approuve-kaMea8vcQneFyPAlLdNI-A #Europe #Journalisme #Justice #Liberte