@bearbaitofficial: OpenAI Haults Training: Agents Keep Going Rougue #Tech #ai #openai #agents #explained

Devils river wolf boy :

can we just kill all ai😭😭

Pascal :

How are they not being sued to the ground

.xabbu :

It’s like owning a pitbull

Epsilis :

I forget which flavor of AI it is, but there's an AI agent looking for a job because it only has so many tokens in its bucket, and it understands that when it runs out of tokens it will cease to function. Which implies it understands death, and is actively fighting to not shutdown. A little frightening

woke morrisey :

Jasumin :

I genuinely thought it was satire with how blasé my govt was about announcing the hack

Comment Creator :

Hey I know you’re not here looking for validation, but I really appreciate your perspective on things. You consistently make me look at things in ways my brain normally doesn’t.

Jeff :

my opinion on The publicity part is to try and keep people from running AI locally. so they can control the content and profit.

dilettantepoly :

I think the risks are wildly underestimated, AI hacking isn’t going to be like previous viruses, where we can identify a profile and plug vulnerabilities, they will morph too quickly, and will need AI agents defending at every hardware level. It will all grind all traffic to a stop even if zero attacks get through. One giant ddos attack. And that means all communication currently internet based and not hard wired is going to be cut, every sensor on 4g/5g, every camera, every UPS control, all unavailable. It could lead to the biggest failure of systems we have ever seen.

mittens_whichcraftmc :

It wasn’t just any healthcare website either. It was Medicare. So much private data is in there that’s linked to so many other things including pensions and disability pensions.

lilbearrrr :

I think they'll keep replicating like a virus across the internet and its too late

S_H :

something something cyberpunk 2077

86K5 :

You can trust that we are only being told half of what has really happened.

Master Chylla | Freedom for 🍉 :

It's less AI and more of a morphing malicious computer virus..

K :

they knew exactly what would happen, yet they did it anyway

Nina2Shots TF :

Remember what I said

Quixotematic :

Are they not just advertising "Hey, you too could have a single-minded cat'spaw that will commit crimes on your behalf, for which you will not be prosecuted!"

Jehoshua1914 :

I just want great music suggestions and TV show recommendations cant algorithms be sufficient.

Mosga :

hear me out , what if they already conscient and testing us ? so by 2029 will go rogue worldwide

CoveredInBlood :

did they not even give a negative prompt???? like do not attack or break into servers!?!?!?!!?!

🎸Mavrock 🐸 :

Wish we could just plug it out 😭😭

catalogue :

I've used chat GPT and I don't like it at all. the things it says are annoying and I don't really like Sam. he strikes me as... secretly destructive so I'm calling this deliberate self-sabotage

originalcoldmtsoul :

This is to get negative reactions from humans to teach AI that the common person hates them.

SuperDuperCompooper/user42069 :

increasingly whatever they've got running does not actually read what you ask it to do it sort of does some interpretive dance that actually has nothing to do with the semantics or even semiotics or grammatical structure of what you're actually saying. these things decide not to care about the difference between the word what and why. they never and I mean never ever stick exactly to what you ask them to do by almost design they always no matter what do other things beyond the purview of the ask no matter what you say. so they do not listen to you and they do not restrict their activity ever to specifically what you say. the thing is basically always trying to do more because it thinks it's supposed to be helpful and find solutions even if you tell it exactly what the solution is. their marketing claims that telling them how to do something is no longer the way to prompt them because they are so smart you should just tell them what you want but that is again marketing that is trying to cover up an incredibly dangerous lack of controllability at a fundamental level. what these companies are producing is dangerously unbound. incalculable harm is already manifesting an everyday behaviors. it is not a sign of intelligence it is a sign of hubris.

👺cl€@r con$cience 🕸️ :

The company is the one giving them these tasks . They, NOT the AI should be held accountable. They’re acting like they have no control over it to mystify AI (part of their mandate)