Michał "rysiek" Woźniak · 🇺🇦
Hacker, activist, free-softie ◈ techie luddite ◈ formerly information security and infrastructure at https://isnic.is/ and https://occrp.org/ ◈ my opinions are my own etc.
(he/him)
profile image: drawing of a head and shoulders of a cat-person, in a space suit.
banner image: long-exposure photo of a large tent, brightly illuminated from inside, looking as if it is made of lava
#foss #libre #privacy #infosec #fedi22
(public toots CC By-SA 4.0 if applicable)
🇪🇺 🇵🇱 · 🇧🇦 🇮🇸 · 🇺🇦
@lloram239 great. ChatGPT and other LLMs demonstrably lack the ability to model the world and make predictions based on such models:
https://www.fastcompany.com/90877523/chatgpt-doesnt-know-what-its-saying
Glad we agree they’re not intelligent, then!
> Circular reasoning. “LLMs are different from human brains because they are different”.
LLMs are different than human brains because human brains are biological organs and LLMs are probability distributions over sequences of words. These are two completely different classes of entities. Like, I don’t know how much more different two things *can* even be.
Are you claiming they are literally the same? Are you saying they are functionally the same? What *are* you claiming here, exactly?
> But human sensory inputs aren’t special
It’s not about sensory inputs, it’s about having a model of the world and objects in it and ability to make predictions.
> The important part is that the AI can figure out the pattern in the data it does get and so far AI systems are doing very well.
GPT cannot “figure” anything out. That’s the point. It only probabilistically generates text. That’s what it does, there is no model of the world behind it, no predictions, no"figuring out".
@lloram239 ah, so you’re down to throwing epithets like “idiotic” around. Clearly a mark of thoughtful and well-reasoned argument.
> Predictions about the world are probabilistic by nature, since the future hasn’t happened yet.
Thing is: GPT doesn’t make predictions about the world, it makes predictions about what the next word, phrase, sentence should be in a text, based on the prompt and the corpus it got “trained” on.
@lloram239 that’s really akin to claiming that a mannequin is a human being because it really really looks alike.
The “predictions about the world” you refer to here are instead predictions about the text. They are not based on a model of the world, they are based on loads and loads of text the model was trained on.
I don’t have to prove ChatGPT is not intelligent. That would be proving a negative. The burden of proof is on those claiming that it is intelligent.
@Flax_vert or “ex-Twitter”
@Natanael enshittification is about power, and ATproto is designed to look decentralized but enable secondary centralization where it matters for power dynamics in the network, in a way that the Fediverse very much doesn’t:
https://rys.io/en/167.html
(shameless plug, I wrote that, but it dives somewhat deep into the “why” of what I said above)
tl;dr it doesn’t matter which PDS you use if everyone is still beholden to the same entity that controls the “reach” layer in BS.
> The Mastodon fediverse have stronger network effects because big servers can enforce policies on other servers to stay federated. It’s complicated for users to move servers.
Well, I wrote about this as well, so I think I might not be missing these details:
https://rys.io/en/168.html
@Natanael you seem to continue to focus on PDSes even though I explicitly said it doesn’t matter which PDS you’re on, the secondary centralization (and thus control) happens in the “reach” layer, outside of what PDSes do in ATproto.
In other words, changing a PDS gives you way, way less agency in BS, compared to agency you get with changing an instance on Fedi.
BS is designed to make that secondary centralization happen, and to be where the real power in the system is.
@xenspidey @DolphinMath one note though, BitWarden requires MSSQL (you read that right, Microsoft SQL Server).