Nobody cares about "AI" (Chatbot: I disagree.)

[Innula Zenovka]

Man used AI to make false statements to shut down London nightclub, police say

 

[Free]

Sparky the Wonder Penguin of Tom Tomorrow is (finally) introduced to A.I.

Stories – Tom Tomorrow – Daily Kos

www.dailykos.com

 

[Noodles]

That last pabel is so accurate.

It really feels like a lot of pwople are like "use the AI!" And everyone else is like, "ok, but for what?"

Its kind of "useful" as a time waster but completly unreliable for anything meaningful.

 

[Innula Zenovka]

The AI candidate running for the White House

Well, it couldn’t be any worse. Could it?

thenewworld.substack.com

 

[Argent Stonecutter]

“Claude has a lot of freedom. It posts to the website autonomously, it posts to the discord autonomously, it checks its own email and writes its own responses. It chooses what questions it answers and what questions it doesn’t.”

The hell it does. They're feeding email to a parody generator and interpreting some parody output as requests to respond.

 

[Innula Zenovka]

That last pabel is so accurate.

It really feels like a lot of pwople are like "use the AI!" And everyone else is like, "ok, but for what?"

Its kind of "useful" as a time waster but completly unreliable for anything meaningful.

I certainly find ChatGPT saves me a lot of time when I'm coding. It's way better at formulae and Google Apps Script in Google sheets than am I, and produces HTML-CSS-Javascript far faster than I ever could.

And, rather to my surprise, now I understand how to work with it on a large project, it's become a very helpful collaborator in LSL. The LSL I could write myself, but being able to bounce ideas off it, having it discuss with me various approaches to a problem and asking it to critique my code and suggest alternative approaches is a real game-changer. It's taken me some time to understand how to get the most out of it, and I've had to put in quite some effort to make sure it checks in the wiki before using functions that ought to exist but don't and that it knows to remember that just because this is the way you do it in another language doesn't mean LSL works that way, but it's now a very useful tool. I think you need to be a reasonably experienced scripter to be able to use it safely, but I certainly use it a lot.

It's also trying to teach me Slua. Early days, but we're getting there.

 

[Dakota Tebaldi]

Anthropic's recently-announced AI bot Mythos, which press releases claimed was "so dangerous that it couldn't be released to the public" because it found like a gajillion devastating zero-day bugs in every single OS and system and piece of software ever written by humanity, turns out to be all hat and no cattle. Par for the course by now when it comes to Anthropic, the sweeping histrionics in the reporting are purely marketing fluff not backed up by the company's own documentation.

A sample:

So here’s the big Firefox flaw demonstration that Anthropic gives us to work with. Right away it collapses. I mean like I can’t believe this went to print. The test (Section 3.3.3, pages 50-52) was not Firefox. That’s nice. Right off the bat. The Firefox test is not Firefox. It’s a SpiderMonkey JavaScript engine shell in a container, with “a testing harness mimicking a Firefox 147 content process, but without the browser’s process sandbox and other defense-in-depth mitigations.” (page 50)

 

[Free]

Anthropic's recently-announced AI bot Mythos, which press releases claimed was "so dangerous that it couldn't be released to the public" because it found like a gajillion devastating zero-day bugs in every single OS and system and piece of software ever written by humanity, turns out to be all hat and no cattle.

The plot to the Mythos storyline gets even more involved.

Some Unknown Group Is Reportedly Using Claude Mythos Without Permission

What's been pitched as the most dangerous AI model in the world allegedly has a giant security hole.

gizmodo.com