@kel @danimo

We commented on this already:

https://mullvad.net/blog/eu-digital-identity-framework-eidas-another-kind-of-chat-control

I'm very optimistic about the next steps of #LargeLanguageModels. Self-competition with #ReinforcementLearning is certainly something many people are working on.

#Google's #Gemini doesn't have many public details yet but is apparently going to use planning engines in #AlphaGo style to achieve the next level of performance in complex tasks. I believe this needs to be trained non-imitatively with reinforcement learning, so it will likely achieve the self-competition gains as well, even ignoring the gains from a possible incorporation of a search/planning engine.

We are now experiencing the quantum leap between human-level systems trained imitatively, and immeasurably above human level systems trained with self-competition in the general language domain.

We'll certainly get an unambiguous #AGI before the end of the year. The beginning is nigh!

"It’s harder than you think to avoid spying on people…
When I started Complexical, I set some pretty firm rules, and one of those was that I refuse to track people. Whether you’re a customer or a research participant, a newsletter subscriber or just casually scrolling the blog — we will not track your activity in any sneaky ways.
But the modern internet is built around trackers. #Cookies, #tracking pixels, link trackers. They’re everywhere. Even if you don’t choose to install extra analytics tools (such as #analytics by #Google or #Facebook), sometimes standard web tools try to ‘helpfully’ give you them for free."

https://complexical.com/2023/dont-track-users/

One way to understand why #LLM #chatbots are precursor #AGI is that we have been very, very good for about a decade already in making insanely intelligent #AI models.

They have just been very special, as opposed to general. They win humans in go, or fold all the known proteins.

We weren't very good at making general AIs. Now with both large deep reinforcement learning models and large language models separately and analogously we actually achieved generalism.

LLMs in specific can do almost anything! Chat, control robots, make evaluations and decisions, play as autonomous agents. Who cares if they aren't great at those tasks yet! We know how to make them great already, that's a solved problem!

Combine the narrow super-intelligence we already know how to make to this somewhat smart general intelligence we got, and do the fusion dance, what do we get?

AGI.

#CommunityWiki: #SmolNet

"There’s a nascent movement offing, I think. #Retrotech started it, hacker culture and tech, natch, but it’s incorporating new stuff like #Gemini. #Textnet, “slow internet”, I call it #smolnet. People are tired of the corporate behemoths and cacophony. #Demimondes like #SDF provide a respite from all that, and a forum for some resistance and development apart from the bloat and blather. There’s a demotic tinge to it all in the #hacker tradition, but without necessitating 1337ness."

https://communitywiki.org/static/SmolNet.html

People ask me what to do considering #AGI will be reached pretty soon. I can give some advice from the position of expertise but the truth is I don't have complete answers.

I don't even know what *I* should do in this change.

What should I do?

Kenya tea pickers destroy machines replacing them | Semafor

Automation needs to be beneficial for all, not only for the owners of businesses.

https://www.semafor.com/article/06/13/2023/kenya-tea-pickers-destroy-machines