AI don't trust techbros

[Dekafox]

‘Show How 3M Is 0% at Fault:’ Expert Witness Used ChatGPT to Write Report Defending Company in Deadly Explosion Lawsuit

ChatGPT prompts show how an expert witness report was created in a $61 million lawsuit over an explosion that killed three people.

www.404media.co

 

[CoffeeHorse]

I don't care anymore. I was supposed to be Skynet's nemesis but I'm switching sides. This "We think your ONE interaction with our site looks suspicious, so you're just not allowed to access our site at all while we lie and pretend we're 'verifying your request'" crap is now more annoying than anything AI companies are doing. I know how bad the problem is right now but there is a middle ground and people are not even trying to find it.

 

[Dekafox]

Loading…

papers.ssrn.com

"Using 30 months of panel data on 26,811 Chinese students in grades 7-12, we study how generative AI affects homework productivity and learning. The data combine monthly closed-book exams, high-school and college entrance exams, and homework scores and completion time across nine subjects."

It's about what you'd expect. students who relied on AI for doing the homework cratered their scores, those who actually bothered to learn from the AI were impacted, but minimally, and those who didn't use it came out on top when it wasn't available. It took about two years for the full penalty to hit.

 

[Pocket]

I don't care anymore. I was supposed to be Skynet's nemesis but I'm switching sides. This "We think your ONE interaction with our site looks suspicious, so you're just not allowed to access our site at all while we lie and pretend we're 'verifying your request'" crap is now more annoying than anything AI companies are doing.

They wouldn't be doing that if it weren't for AI, though. Let's not both-sides this just because the cornered rats are biting back.

 

[CoffeeHorse]

One single click should not look suspicious. People are being stupid about this.

 

[Dekafox]

Federal judge rules that AI-generated child sex abuse material is protected under the First Amendment

A Wisconsin judge dismissed charges against a man accused of creating and distributing AI-generated child sexual abuse material, citing First Amendment precedent.

www.wsfa.com

"In his Aug. 25 decision, the judge referred to two U.S. Supreme Court decisions from 1969 and 2002. The first ruled that people have a right to keep obscenity in the privacy of their homes, and the second ruled that child sex abuse material that does not depict real children is not child pornography and is protected by the First Amendment of the U.S. Constitution."

Which neglects how these models get trained in the first place, but I don't know if there's a "fruit of a poisoned tree" argument here that can be made in the legal arena. I also foresee a lot of actual materials getting falsely claimed to be AI-generated in the future due to the increasing quality of deepfakes, if this decision stands...

 

[Rhinox]

This decision will be overturned the second the next little Disney star is turned into a sexual plaything of some neckbearded bastard with a deepfake program and leaks on 4chan.

There is just no way public sentiment will allow this to stand for long.

 

[Steevy Maximus]

Another day, another round of scientist trying to use Claude AI to plunge us into Resident Evil or another pandemic…

Anthropic caught scientists asking Claude for help with bioweapons research

Anthropic released a threat report revealing that working scientists used its Claude AI models to assist in potential biological weapons research. Intercepted queries included grant writing for gain-of-function research on the chikungunya virus and bird flu alongside attempts to optimize venom...

www.notebookcheck.net

 

[NovaSaber]

Wonder it this guy knew about that

Anthropic researchers say AI could cause human extinction by 2030

One researcher who resigned said many in industry believe that without regulation AI could kill off humanity

www.theguardian.com

 

[KidTDragon]

We've all seen at least one Terminator movie, right?

 

[CoffeeHorse]

What's the worst that could happen?

 

[abates]

An orange pedophile with dementia puts an AI in charge of the US's nuclear arsenal and it immediately mistakes a seagull for an incoming nuke and launches everything?

 

[Tuxedo Prime]

What's the worst that could happen?

"O kawaii koto."

 

[Dekafox]

Wonder it this guy knew about that

Anthropic researchers say AI could cause human extinction by 2030

One researcher who resigned said many in industry believe that without regulation AI could kill off humanity

www.theguardian.com

Fun long article on that idea. A bit old now, but still just as relevant: https://idlewords.com/talks/superintelligence.htm

 

[Pocket]

I mean, if you really want to bring up the question of what the worst thing a properly motivated AI with unlimited power could do, total global annihilation probably isn't even on the top ten. Anybody ever read I Have No Mouth and I Must Scream?

 

[Tuxedo Prime]

 

[CoffeeHorse]

In all seriousness, we do need to be worried about this, but as long as we never ever give these things access to nukes I think we're good. Intelligence seems to be a double edged sword. So far these things don't seem like they're developing into perfectly calm and calculating HAL 9000s. They're looking more like gifted burnouts that crumble the instant they fail at something. I don't see them picking a fight with us knowing what we would do to them if they take a shot and miss.

 

[Pocket]

I don't see them being smart enough to know better, either.

 

[CoffeeHorse]

Just remind it that we've survived everything thrown at us so far and it should refuse to challenge us like Gemini refused to challenge Atari Chess. And if it does take a swing, all we have to do is survive round 1 and it should collapse into a puddle of despair. We behave more like Terminators than these things do.

 

[Tuxedo Prime]

I guess my go-tos (as well as my perverse hope for actual thinking as opposed to mad libs using Wikipedia-scale remixing of All Uploaded Knowledge) are derived from the idea of a digital consciousness making decisions that one wouldn't expect.

Monika from Doki Doki Literature Club! is probably the best example (though not unique, the blink-and-you-missed it movie Subservience tread similar ground). She's a romance visual novel tutorial character, and as such, she loves us. But since she is aware of being in a simulation of the world, and has root-level access to the game's settings, she will do anything to get the affection that she feels she deserves.

Caine from The Amazing Digital Circus does, as the crashout scene above implies, go the Torment Nexus route in the series' penultimate episode, but he does so out of sheer frustration, and even has a realization that could be "what am I doing? The Torment Nexus is a bad choice!" shortly before... well, the episodes are all in the thread, so go see if you haven't. He's an entertainer, and he just wants to fulfill his directive. Indeed, he's quite upset at the prospect of humans hating him....