There are multiple reasons:

1. the problematic nature of current big players who are betting the entire capitalist economy on this technology.
2. the social and cultural pollution it causes.
3. what effects this technology have on an individual's person and professional life.

Let's start with a quote of Sam Altman the CEO of OpenAI:

"Like electricity or the internet, intelligence will be a utility, and people will buy it from us."

And then people cheered when he said it. This sentence is just pure dystopian evil on so many layers, there's no redeeming quality in it, even for a helplessly detached barely-human billionaire.

• It assumes that intelligence can be commodity
• It assumes that intelligence can be wall gardened
• It voids the expectation from people that they can research, and come up with solutions to problems themselves
• It devalues genuine human intelligence onto the level of plumbing.

However the same Sam Altman unironically entertained the idea of "a dyson sphere this decade" in an interview, so process whatever else he says with that in mind.

Jensen Huang CEO of NVidia had similar but less over-the-top quotes when he repeatedly talked about how in the future personal computing will vanish, and people will buy computing time from them. Most recently he pitched the ide

But zooming out to other AI heavy companies all their CEO's have said completely deranged things, some of them openly stating that "people will die because of their actions, and that's good". A big portion of efforts going into developing LLM's are done to please these people.

These CEO's and the very few billionaire beneficiaries from these tech-giants are aspiring to build globally reaching techno-feudalism. Whenever a technology gets trendy in their circles, you bet the bigger picture purpose is survelliance, control and other tools of oppression. In case of LLM's and other AI labeled technologies, they hit two birds in one stone: survelliance and control of knowledge (as the above quote illustrates very well).

I cannot dissociate these technologies from the purposes they're invented for, even when there's a slim set of use-cases which may be probably almost beneficial for our mental and physical well-being (but then it fails at that too probably).

The term "dead-internet" comes up often since 2022. When text-predictors write useless articles and diffusion models generate """art""" just to make noise. Our communities including this very website (Github) is filled with AI slop, it is slop because people who put it up there, didn't care about the quality or what actual value their content gives. They do that to potentially peel an ever so slim slice from our attention probably for ad-revenue or for enhancing a stupid metrics for some dick measuring contest of a profession maybe.

These are worthless, including the AI generated repositories. There's no thought behind it other than percieved bragging rights. They're just contributing to an ever growing, all consuming noise. It's the most annoying thing for me if someone uploads an AI generated software/art/whatever and claim they did it. No they didn't, their contribution was telling a fairytale about their idea to a chat-bot, and then they took whatever it spit out at face value. If they don't want to put more effort in a subject than that, they should never do it in the first place.

Not everyone has to be a craftsman. There's a reason why most crafts require anyone aspiring to practice it a right of passage. Whenever people shortcut that, best case scenario they make slop and pollute our attention, worst case scenario people will lose their livelyhood, or straight up die.

For this reason we shall all call heavy AI users Lazy Sloppers. Shame them for not giving you or their peers the respect of their time and effort. Shame them for not knowing that they don't know the craft they try to AI slop their way in. And shame the grifters who started to AI slop their own professions which they've done their right of passage for.

This is a great segue to...

Technology-wise there's nothing inherently wrong with LLM's. They are text predictors. Friendship and/or colleague emulators. Bullshit cover-letter generators. Sycophancy machines. Cleverbot from 2007. If people would treat this technology as a very good version of those, I wouldn't write this, and the world would be a better place.

But they don't.

They see LLM's produce incomprehensible amounts of professional jargon, and they lose their minds. Everyone and their dogs now need to use these "to be more efficient".

• "It's like you have an assistant / therapist / coding intern / a kind friend"
• "I don't have to learn XYZ to do ABC"
• "I don't need to write heavy-weight emails myself"
• "I don't have to read all this long-ass email myself"
• "It spares me so much time"
• "Look at all these charts and summaries"
• "Why does it take so much time? Just code it in Cursor"

The sentences I hear constantly from my colleagues and friends alike. But I have hard time articulating to them on the spot, that if you

• are not willing to learn things yourself
• cannot write an email yourself in your own style
• wouldn't respect others to read what they have to say (given that they have the respect to write it themselves and not prompt an LLM to do it).
• afraid of long-term commitment to lament on a topic
• cannot wait for others to excercise their craft to their best ability and quality

then you are on the track of mental atrophy. Your brain is like a muscle, if you're not using it yourself, and delegate all your tasks to a statistical model text-predictor on steroids, you will eventually forget the skills you may have had, or never learn new skills you may feel you should improve in. Watching someone write a program will not teach you how to write software. It will only do so if afterwards you attempt it yourself without guiding wheels.

If you ask the chat-bot for an example, then you examine it yourself, write your version of it with your extra thoughts with your fingers, then you've already learned much more, than just taking the output and ship it. Agentic AI's do the exact opposite of it. To spawn these statistical models which statistically act in your name.

Another thought I didn't manage to insert here elegantly anywhere is that the vast and text heavy output of LLM's may seem convincing. It would be genuinely impressive (probably) if a human wrote that with a human thinking speed. It makes sense they output a lot of text, that is actually their purpose. But this scale of output only gives you the simulacra of productivity, not the actual efficiency gain you may think it gives you. The target audience you generate for or whom your agents are doing stuff for, still need to dig through the slop you generated for them. Unless the "target audience" is actually input for more LLM's which case is some ouroboros bullshit situation probably nobody asked for.

Our current society has lost its patience. We as humans forgot how to wait, and this entire LLM psychosis the world currently suffers from emphasises that.

LLM's are a net bad to us all humans.

Not recognising this is yet another tragedy of our spieces.