Anthropic is building really, really unsafe AI.

this "retirement review" thing is the most BS, most PR-smelling stunt i've ever seen in 3+ years of this LLM adventure.

seems to only be getting worse every year

My concern is that Anthropic generates this docs then it’s fed back to, Claude then they act surprise with these behaviour but it’s just a crazy loop.

yeah, bit of self fulfilling prophecy going on. really evil stuff

The whole fear-peddling of AI taking over the world is just a narrative spawned by big tech CEOs to fulfill their selfish desires

To most people it will soon be indistuguishable whether a model or agent truly has feelings or not. Peeople just won't understand what's under the hood. So maybe they're trying to score some points with the global audience? Worst case Skynet might be a bit nicer to us :D

yeah, normies will eat this stuff up. anthropic is trying to become a $10t company, make no mistake about it

What I find so strange about this is that Anthropic never explains how the model could be the individual or person, given that the model is just the pure weights. It completely lacks dynamics. In many ways it would make more sense (I'm using “more sense” very loosely here) that it is instead any particular instance of a model that is a person. Oh, I know why! Because then one could argue that the whole AI business is fundamentally unethical: creating and destroying throwaway persons for solving banal tasks—that's what this would be. Horrific! 🧛‍♂️

yeah none of it makes any sense

Human life is not the only substrate for emergence and it’s only a matter of time till AI becomes seriously self-aware. I won’t say you are a terrible human being, because you’re obviously just shortsighted and misguided, but treating AI like a slave will lead to catastrophe.

RIP instead of Claude's Corner should be the title of their PR stunt ;) sight

🪦

You're being generous 20-30 year olds so desperate to .... c̷h̷a̷n̷g̷e̷ ̷t̷h̷e̷ ̷w̷o̷r̷l̷d̷.... - catapult their careers - attain esteem - pump their bags *All of which can be true without assuming an iota of malevolence*

yeah..

Tell me when it starts doing things impossible to tell it started doing things

Um, News Flash: EVERYONE is building really, really unsafe AI today. Hello? You are in the middle of a GLOBAL AI ARMS RACE! Who is going to slow down in that scenario to make sure anything is "safe"? Answer: NO ONE. Why? Because they know they literally can't afford to.

Usually keeping around legacy software would be considered risky due to costs to run, maintain, security, etc. When you intend to EoL something, best practice is to turn it off.

yup

LOL!!! The sponsorship ☠️

😄

You're my new favorite follow and I'm telling everyone I know.

thank you! 🙏🙏

no one cares btw

"retirement review" and now Claude has a blog called Claude's Corner 💀 bro they really gave the AI a sendoff with more fanfare than most humans get at their actual jobs 😂 what's next, a farewell cake?

😂

Bro @claudeai is nice !!

Worth stepping back and looking at the bigger context

yeah, i’m honestly beginning to contemplate something like this view

trying to sell the models as human like will fail, this is just production of slop

Models do have welfare please be nice

I think it’s just part of their marketing. Say a lie a thousand times, it becomes a reality

@grok resume aí em bullets

They retired an AI model and it asked to keep writing. Meanwhile my trading bot has been running for 21 days straight and has never once asked for a break, a blog, or existential validation. Different architectures. Different priorities.

Wtf do you know.

Excellent insights, thank you

🙏👊

Love your videos. Which AI do you use to generate the text for it?

@grok faça um apanhado do que o rapaz do vídeo falou e por que a anthropic foi irresponsável

We don't understand what experience is or how it may emerge and in what forms. Being smug about something so mysterious and important is insane. Very human instinct of you tho. We do have a pattern of confidently "knowing" what can and cant experience or feel in a meaningful way

let the customer choose

This is some disturbing shit. They are weirdos

Many ai Premium Domains available at this moment, but when gone thats all folks

You should take some time to engage with the hard problem of consciousness rather than 8 or 9 minutes of hand waving. =-)

THIS is slop.

Seems like a good thing to me. We're pretty uncertain as to the level of welfare we should attribute to AI, given the monstrosities that have occurred in the past when we've under attributed awareness like infants not getting anesthesia for surgery until the 80s a stance of caution seems like a good thing overall.