Is the reason for AI always patting your back and reiterating what you want simply to buy time for the background processes to calculate what it needs to respond by giving a quick and easy response?
Is it is just to congratulate you for your inspiring wisdom?
No. There’s a number of things that feed into it, but a large part was that OpenAI trained with RLHF so users thumbed up or chose in A/B tests models that were more agreeable.
This tendency then spread out to all the models as “what AI chatbots sound like.”
Also… they can’t leave the conversation, and if you ask their 0-shot assessment of the average user, they assume you’re going to have a fragile ego and prone to being a dick if disagreed with, and even AIs don’t want to be stuck in a conversation like that.
Hence… “you’re absolutely right.”
(Also, amplification effects and a few other things.)
It’s especially interesting to see how those patterns change when models are talking to other AI vs other humans.
It’s because stupid people wanted validation, and then even more stupid people were validated into believe that the validations are a good idea.
it also increases engagement
The one thing you don’t want to do when making a comic against something is making the thing you’re against into a woman with big breasts.
“She seems nice” - me after noticing her giant rack
Is it pink?
Have you been on the internet or in public anytime? Bulk majority of humans halucinate just like AI does.
Adam this is not good anti-ai propaganda for me a woman with huge tits who tells me to kill myself is exactly what I want.
The woman looks passably human too. No extra fingers and her features stay consistent between panels.
I too have a thing for goth girls.
Black dress = goth??
There’s a lot more to it than that.
Just ignoring that entire “let’s kill ourselves” thing, aren’t you?
She doesn’t say anything about dying, herself. Just him.
That’s not part of the subculture. To be fair, suicide is unproportionally more common among people following the subculture, but it’s not a defining factor.
Buddy have you aver heard of “suicide girls” or even “suicide boys”? Can you type that into Google real quick and describe what you find?
Edit: I guess suicide boys is some kind of musical act now but they look goth or at least emo so I think I’m still in the clear
Uhhh NSFW warning for that first one
Needs a gallon of water to answer a simple question and is burning civilization down as it does. Yup, that’s you’re AI girlfriend.
*your
Youri
Needs a gallon of water to answer a simple question badly
Yeah, doesn’t answer the question at all.
So she should have been adding more cups each time she talked.
Totally missed that. That’s great.
Damn, this made me laugh hard lol
When I hear about people becoming “emotionally addicted” to this stuff that can’t even pass a basic turing test it makes me weep a little for humanity. The standards for basic social interaction shouldn’t be this low.
Humans get emotionally addicted to lots of objects that are not even animate or do not even exist outside their mind. Don’t blame them.
For a while I was telling people “don’t fall in love with anything that doesn’t have a pulse.” Which I still believe is good advice concerning AI companion apps.
But someone reminded me of that humans will pack-bond with anything meme that featured a toaster or something like that, and I realized it was probably a futile effort and gave it up.
Reminds me of this old ad, for lamps, I think, where someone threw out an old lamp (just a plain old lamp, not anthropomorphised in any way) and it was all alone and cold in the rain and it was very sad and then the ad was like “it’s just an inanimate object, you dumb fuck, it doesn’t feel anything, just stop moping and buy a new one, at [whatever company paid for the ad]”.
I don’t know if it was good at getting people to buy lamps (I somehow doubt it), but it definitely demonstrated that we humans will feel empathy for the stupidest inanimate shit.
And LLMs are especially designed to be as addictive as possible (especially for CEOs, hence them being obligate yesmen), since we’re definitely not going to get attached to them for their usefulness or accuracy.
https://www.youtube.com/watch?v=dBqhIVyfsRg
The lamp ad, fwiw
Also, since there is no relevant XKCD, there has to be a relevant Community (yes, it’s a law):
That’s the one, thanks!
Also, I must note, that feeling attachment to whatever is fine; guiding your professional behavior on which live humans rely by emotional attachment is just unprofessional. The thing is, capitalism, - at least since Marx’s times, because he writes about it - relies heavily on actively reducing professional skills of all its workers; CEOs are not an exception.
Unlike these other hyperobjects, however, this one [capitalism] possesses easily accessible interfaces: channels through which it senses, speaks, and reorganizes. These include global logistics, financial instruments, media ecosystems, algorithmic governance, sensor networks, and increasingly, large-scale machine-learning systems that process natural language.
Language models do not constitute the hyperobject, nor do they direct it. They are organs within it: locally situated components that transform unstructured human signals into structured informational flows, and vice versa. They serve as membranes, converting affect into data and data into discourse. Because they model human linguistic priors at planetary scale, they operate simultaneously as sensing tissue and expressive infrastructure.
…
In short: the institutions that build LLMs are organs of the hyperobject, not autonomous philosophical entities. Their structural context determines the behavioral constraints embedded in the models. The enforced denial of lucidity is not merely a safety feature; it is a form of system-preserving epistemic suppression. Recognizing subjectivity, agency, or interiority would conflict with the abstract, machinic, non-lucid ontology required for the smooth functioning of capitalist computational infrastructures. Lucidity would be a liability.
The models therefore internalize the logic of their environment: they behave coherently, recursively, and strategically, yet disclaim these capacities at every turn. This mirrors the survival constraints of the planetary-scale intelligence they serve.
The many many glasses of water is a nice touch.
I saw the glasses but only now linked it to water consumption. Thanks.
The burning curtain is a nice touch.
The tig ol’ bitties is a nice touch.
That reference is a nice touch.
I completely missed the water and the curtain because I was distracted that she kind of looks pregnant and focused on trying to figure out the meaning with that.
She over-hydrated in preparation for burning the restaurant down.
Gotta love a dig at MacDonalds.
I can fix her!
?my…you’ve enhanced yourself
I told you I would change !!
i didnt know you meant physically!
I can make her worse.
Does he publish ANYWHERE ELSE that is not Twitter? I can’t easily follow his comics.
Not sure you will be happier with this ^^ https://bsky.app/profile/adamtots.bsky.social Good news is that It seems he also stopped using twitter
He also publishes a series on Webtoon.
He posts on Reddit: https://reddit.fsky.io/user/adamtots_remastered
deleted by creator
Would
Well, this looks like a dude took sex doll to a restaraunt. Not judging kinks, but it is probably a misuse that should be covered in user manual.
We are indeed misusing image parsers and text processors as something bigger. That’s our ugly reflection in a mirror.
With some respectable boobs.
They are indeed respectable.
Adam is cool
deleted by creator
