Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
karolklp 
posted an update 5 days ago
Post
2510
Hey HuggingFacers. Happy New Year!

I'm Karol, CTO with strong business acumen. It's a bit too much on a "culture" side, to be core here, but I wrote an article that highlights rather overlooked societal risks of "AI Agents Reverse Alignment". Happy to hear your thoughts!
Also: I'm looking for a mentor / reviewer regarding research paper around AI Agents Topology - [email protected]

https://karolklepacki.substack.com/p/heralds-of-reverse-ai-alignment

Hi Karol, the topics you touched on are of interest to me too, and IMO there needs to be a lot more public discussion around. We can take historical indicators, even recent ones, to support this claim, as while LLMs certainly introduce a new variable, we can already see how dramatically other profound technologies have started impacting society.

Heres a quick one: Psychopaths as they can be measured clinically, represent about 3% of society. Let's think of psychopaths in the useful context as those trying to make manifest their own pathology onto the world/network around them.

So what happens when Social Media comes into the picture.

Well, pre-social media people 1) were operating within a network approx Dunbars Number, hence their reach was inherently constrained, 2) they could not operate anonymously over long periods. 3) people that that performed, aided, indulged, ignored pathological behavior him faced harsh social shaming. All of these factors served as powerful constraints on our ability to contain the spread of pathologies

These constraints have been all but eliminated in the wake of social media and the erosion of traditional social norms (and by "traditional" I'm using it very specifically to mean "social experiments that worked over time and hence were adopted")

What is the likely outcome from the removal of these constraints? I dont see how its anything other than widespread contagion. 3%

Anyway, could be worth a chat. Particularly because I can tell from your article that while we agree completely about the very serious challenges that need to be address, we're also going to have some strong, healthy disagreements - which is good, because exploring ideas different from our own and engaging with them is the only way for the best ideas to emerge. And I TRULY care about solving these problem, or mitigating their risk - I could give a damn if my currently held positions prove to be right. I'll update them in an instance with better evidence.

Let me know if you'd like to chat or possibly do some collab articles

[email protected]

·

Thanks, Rob! I very much agree!
What's worse with social media is that it's not only lifting that natural barrier, but intentionally and actively promotes psychopats + others. It just what clicks best.
I'm so happy you reached out, because there's no place for discussion like that. Regular folks don't have their clue. Builders gonna build and don't care. And in the meantime, big corporations will drift us further into dystopian world.
But AI Agents got me really interested, and one of the reasons is that it's like exoskeleton for indie hackers, that can build ton of laser-focused value for many companies, disassembling a lot of monopolies.
I'll reach out to you on email, and let's have a remote walk-and-talk.