- 2 Posts
- 197 Comments
4·9 days agoNo, that’s because social media is mostly used for informal communication, not scientific discourse.
I guarantee you that I would not use lemmy any differently if posts were authenticated with private keys than I do now when posts are authenticated by the user instance. And I’m sure most people are the same.
Edit: Also, people can already authenticate the source, by posting a direct link there. Signing wouldn’t really add that much to that.
Sure, but that has little to do with disinformation. Misleading/wrong posts don’t usually spoof the origin - they post the wrong information in their own name. They might lie about the origin of their “information”, sure - but that’s not spoofing.
I don’t understand how this will help deep fake and fake news.
Like, if this post was signed, you would know for sure it was indeed posted by @lily33@lemm.ee, and not by a malicious lemm.ee admin or hacker*. But the signature can’t really guarantee the truthfulness of the content. I could make a signed post that claiming that the Earth is flat - or a deep fake video of NASA’a administrator admitting so.
Maybe I’m missing your point?
(*) unless the hacker hacked me directly
It works fine for me on Hyprland.
That is why I use just
int main(){...}without arguments instead.
I don’t think any kind of “poisoning” actually works. It’s well known by now that data quality is more important than data quantity, so nobody just feeds training data in indiscriminately. At best it would hamper some FOSS AI researchers that don’t have the resources to curate a dataset.
What makes these consumer-oriented models different is that that rather than being trained on raw data, they are trained on synthetic data from pre-existing models. That’s what the “Qwen” or “Llama” parts mean in the name. The 7B model is trained on synthetic data produced by Qwen, so it is effectively a compressed version of Qen. However, neither Qwen nor Llama can “reason,” they do not have an internal monologue.
You got that backwards. They’re other models - qwen or llama - fine-tuned on synthetic data generated by Deepseek-R1. Specifically, reasoning data, so that they can learn some of its reasoning ability.
But the base model - and so the base capability there - is that of the corresponding qwen or llama model. Calling them “Deepseek-R1-something” doesn’t change what they fundamentally are, it’s just marketing.
There are already other providers like Deepinfra offering DeepSeek. So while the the average person (like me) couldn’t run it themselves, they do have alternative options.
A server grade CPU with a lot of RAM and memory bandwidth would work reasonable well, and cost “only” ~$10k rather than 100k+…
To be fair, most people can’t actually self-host Deepseek, but there already are other providers offering API access to it.
6·26 days agoThe point of it being open is that people can remove any censorship built into it.
The particular AI model this article is talking about is actually openly published for anyone to freely use or modify (fine-tune). There is a barrier in that it requires several hundred gigs of RAM to run, but it is public.
It’s almost sure to be the case, but nobody has managed to prove it yet.
Simply being infinite and non-repeating doesn’t guarantee that all finite sequences will appear. For example, you could have an infinite non-repeating number that doesn’t have any 9s in it. But, as far as numbers go, exceptions like that are very rare, and in almost all (infinite, non-repeating) numbers you’ll have all finite sequences appearing.
Now, if only the article explained how that killing was related to TikTok. The only relevant thing I saw was,
had its roots in a confrontation on social media.
It’s says “social media”, not “TokTok” though.
I’m confused, isn’t Fedora atomic immutable? Shouldn’t that make it stateless automatically?
Why are you surprised? They are called cock-roaches, after all…
Well, he didn’t even buy the original (I guess it has spoiled by then), but a DIY replica and a certificate.
Too bad that’s based on macros. A full preprocessor could require that all keywords and names in each scope form a prefix code, and then allow us to freely concatenate them.