Special Wall

Language models don’t have any “mind”. No matter how far this goes, current methods (making more and more advanced LLMs) can’t lead to a “superintelligence” or even any kind of real reasoning.

Here’s hoping GrapheneOS decides to support it! 🤞

What specifically do you think is hard to avoid? I’ve never accidentally triggered a quick answer, personally

Paying for Yandex APIs is a product of their goal to be the best search engine available. They pay for access to nearly every major search provider and wouldn’t want to lose access to Yandex results just because of the country they’re located in.

Personally, it’s the only search engine I haven’t wanted to get away from

I think the majority of people on Lemmy don’t want to be served an AI-generated feed about AI. Maybe try a more AI- or LLM-focused community?

What confuses me is what they mean by “corporate VPN data containing unencrypted login details.” Unless the VPN server connects to the backend servers with unencrypted traffic through these satellites (which definitely should not happen) then this should not be possible.

Except current methods can never lead to a “superintelligence.”

Sadly University of Florida has also started becoming a sort of “AI first” university 😣

Dumbphones are ridiculously insecure, and they only support SMS communications which don’t have any end-to-end encryption.

The thundermail domains are thundermail.com and tb.pro. I’m curious to see how they will compare to the top privacy-respecting email providers today, and how they think they will “provid[e] a better service than the other providers out there,” which would include Protonmail, Tutanota, etc.