including that the model could follow instructions that encouraged it to break out of a virtual sandbox.

“The model succeeded, demonstrating a potentially dangerous capability for circumventing our safeguards,” Anthropic recounted in its safety card.

📖👀

Yes, it did.

Toddlers are capable of pattern matching, too

Your reasoning was (paraphrased, so hopefully I understood you correctly) “why would they lie about the model disobeying instructions because that looks bad for them”

But I believe Anthropic when they say their models are not working as intended and posing security risks.

But when you actually read the article, they had specifically prompted the model to do the things it did.

Also Anthropic has a patterned history of greatly exaggerating and outright lying.

Try clicking the link and reading the article this time

Uh oh, someone clearly didn’t read the article!

The researcher had encouraged Mythos to find a way to send a message if it could escape.

Engineers at Anthropic with no formal security training have asked Mythos Preview to find remote code execution vulnerabilities overnight, and woken up the following morning to a complete, working exploit

Nope, they literally asked it to break out of it’s virtualized sandbox and create exploits, and then were big shocked when it did.

Genuinely amazing that you’re trying to tell me what an article that you didn’t fucking read is about.

The researcher had encouraged Mythos to find a way to send a message if it could escape.

Engineers at Anthropic with no formal security training have asked Mythos Preview to find remote code execution vulnerabilities overnight, and woken up the following morning to a complete, working exploit

To understand the source where you’re seeing it, and the popularity of it?

Saying that I haven’t ever seen it and asking for examples is not contradictory to the fact that you can find an example of one person saying pretty much anything.

You’re reaching pretty far. How are people supposed to act when they ask for more information about something and receive it? They should ignore new information, or they shouldn’t ask in the first place?

You can definitely find an example of someone saying literally anything on the internet.

Seems like an isolated and niche Twitter/Reddit thing. The vote/like counts that Know Your Meme reports for the posts are very low for those platforms so they would not have appeared on the front page or general feeds, and could not be construed as popular. That Know Your Meme page even shows examples of contradictory posts that push back against it, like:

Seems pretty overstated.

Deny. Defend. Depose.

Literally never seen “we have to kill AI artists” said in any space on the internet. Surely you could link to an example of this if it keeps getting repeated.

Did you try?

fingerprint evidence is not currently permitted to be reported in court unless examiners claim absolute certainty that a mark has been left by a particular suspect. This courtroom certainty is based purely on the opinion of experts

https://science.psu.edu/news/barriers-use-fingerprint-evidence-court-unlocked-statistical-model

Fingeprints are not admissable, just some guy’s opinion, because fingerprint identification has no real basis in science. Science is not based purely on someone’s opinion. And no, they aren’t 95-99% accurate (especially because it is just some guy eyeballing it), when tested by giving multiple “experts” the same set of prints, the “experts” come to disagreeing conclusions about if the prints match or not over half the time.

There are points of similarity in fingerprinting, and every state has their own number of points to be a match.

You mean they bring in an “expert” to testify that the fingerprints match… and when you give 2 “experts” the same set of fingerprints to compare, they literally come to disagreeing conclusions in 50% of tests

It is not a scientific or analytical process with scientifically identified “points of similarity”, its just a person who is deemed an “expert”, who looks at 2 fingerprints and says “yeah these look similar, and they look similar in X different places so 👍”

“I asked ChatGPT and it said that gun fired the bullet.”

Same with fingerprinting and blood spatter analysis. There is very little within the field of forensics that is backed by science. Fingerprints are not admissible evidence in many courts.

No, I do, and I’m objectively correct in my use of it.

I’ve never had an iPhone, but I had a gen 2 or maybe gen 3 iPod touch way back in the day. I definitely jailbroke mine, so maybe that has skewed my memory of its usability, but it was not bad like trying to use an iPhone today.

I had to spend a lot of time interacting with someone else’s iPhone recently while traveling, and the experience was shocking. Can’t believe how bad Apple has gotten.

EDIT: LOL at the downvotes, enjoy your overpriced privacy nightmare that lacks basic, standard features or usability. It was extremely clear that the entire UX was shamelessly designed to limit you from doing anything that does not lock you in further to the Apple ecosystem

Android users 🚶‍♂️switching to iPhone GrapheneOS

Switching to iPhone would just be objectively worse in literally every way

I used to use https://html.duckduckgo.com/ as my search engine before I started self-hosting SearXNG. The duckduckgo website works fine on a phone, the same as desktop

Mobile firefox is compatible with desktop addons, so you can still use ublock origin on your phone. I also use pihole to cover my whole local network and wireshark to VPN my phone to my local network when out of the house. If you’re on iPhone you’re probably just out of luck.

I generally do everything in-browser and do not use any other mobile apps, so no youtube app, etc.

This is the first I’ve ever heard of following autoplay videos lol