Alexraptor wrote on 2026-01-08, 07:57:

I've become so disgusted by AI/LLM's, the ethics of both the models themselves and the companies that make the hardware as well as the impact on consumer products and the environment, that I'm weaning myself off them and cutting them out of my workflow, as a programmer, entirely.

It should all be destroyed.

The cost/benefit ratio of AI can be drawn either way so wanting the opposite destroyed can't automatically be the ethical position, if you're interested in that. In fact if you want a common service destroyed then might as well drop the ethics unless you're willing to be convinced otherwise by other viewpoints. But in any case you have to cut off a lot of things because AI is already in hardware and software design and development. You could still limit yourself to let's say 2017 hardware max, but the software is going to be an issue if you want to function in society. Also you can't contribute on the internet because it's all going into AI.

"AI Integration. What are you thoughts on it all?"

The new favorite scam from the rich and powerful.

I hate that every goddamn laptop out there now "has AI". I hate how it tries to summarize Facebook discussion threads. I reaaaalllly hate when it suggests questions related to the subject matter, and those questions are stupid AF (like "is the Creative PCI64 a PCI or ISA sound card?").

I hate that it makes up PowerShell commands whole cloth (not every Get-* command has an associated Set-* counterpart). I hate that I've corrected it on PnP PowerShell and it's all smarmy with "you're right! I told you the wrong command to use! Here's the correct command" and then I find something wrong with that one.

There are legitimate uses for AI, but what we're getting is generally not legitimate.

And don't get me started on AI 'art'. (Thank you!)

Ive been using Chat GPT to write a book... Its helpful in that I can feed in a load of information i would like to use and it takes those ideas sprinkles in some psychology and creates a timeline to work to.
Its good at merging different writing styles of famous writers too...

I like the auto translation intergrated into my phone, I can talk to the people in Italy now and have the conversation translated in real time as were talking.

The image generators are good, I think thats an up coming minefield that will see a lot of censorship in the future in the UK, and it wont be long either as people are using it to attack without realising they are attacking people, so it needs to be controlled.
It will have unfortunate effects on other generators though.

AI is good to Linux commands, you tell it what you would like to do and it tells you, which is nice. Linux is no longer the preserve of those who know. Its become a useable landscape now that isnt anywhere near as elite user as it was.

Its interesting thats for sure.

ElectroSoldier wrote on 2026-01-08, 23:12:

AI is good to Linux commands, you tell it what you would like to do and it tells you, which is nice. Linux is no longer the preserve of those who know. Its become a useable landscape now that isnt anywhere near as elite user as it was.

Its interesting thats for sure.

Sigh... Give a man a fish, or teach him how to fish.

The people who use "AI" (let's be fair; it's predictive text on a large scale for LLMs) don't end up learning anything, just regurgitate the slop that they are presented with as fact.

No one is making any money from this, and it's a dead end... what is going to happen when all of the existing "free" technical documentation, code snippets and similar are hoovered up? Do we simply stop trying to produce any new algorithms, because the LLM doesn't know about what it cannot hoover up?

There are some avenues that are worth continuing; machine learning, computer vision and digital twins of real world counterparts (be it protein folding, materials science or whatever)... but what the general public are calling "AI"? No. And the quicker it all implodes, the better.

It should be a choice. I get more hallucinations than average, but there might have been some potential in it being a tool with more time and research (as long as your brain isn't broken in the same way as mine). Unfortunately, it fell into the hands of big tech executives, who forced it into everything before it was ready and demanded that it was implemented in the absolute dumbest way possible.

AlaricD wrote on 2026-01-08, 21:17:

I hate that every goddamn laptop out there now "has AI". I hate how it tries to summarize Facebook discussion threads. I reaaaalllly hate when it suggests questions related to the subject matter, and those questions are stupid AF (like "is the Creative PCI64 a PCI or ISA sound card?").

I hate that it makes up PowerShell commands whole cloth (not every Get-* command has an associated Set-* counterpart). I hate that I've corrected it on PnP PowerShell and it's all smarmy with "you're right! I told you the wrong command to use! Here's the correct command" and then I find something wrong with that one.

I'm also really fucking tired of those AI summaries and suggested questions. Then you have Copilot in fucking Notepad trying to summarize code, log files, and my own meeting notes. At the same time, Windows 11 Notepad doesn't register keystrokes reliably, which is the entire job of a text editor.

I've never let the "summarize" thing finish running, but based on how long-winded and obnoxious the LLM writing style is, I would expect the summary to be longer than what I originally wrote with two pages of bulleted lists that use emojis (mostly the green check mark, but sometimes other emojis loosely based on the content) instead of bullet points. Who thought that was a good idea? Emojis for bullet points looks horribly unprofessional.

To be honest, the way LLMs write reminds me of a high school student trying to pad out a paper to fit a minimum page count five minutes before it's due while also not understanding the content. I think I said that last spring. Hell, maybe those high school students are the ones who made this stuff.

An LLM that outputs emojis is clearly not intended for professional summarization, like you don't take your notes with a crayon either. If it's offered for your professional use case then that's an integration problem. If there are no style controls then you can use paid APIs that give you that control, as a professional you probably should be using them anyway.

At least with a ChatGPT custom GPT, no matter how many different ways I try to put it in the prompt for it to not use emojis, or bulleted points altogether for that matter, in addition to directly telling it in the conversation not to use emojis, it still ends up using them. Not to mention that certain unavoidable triggers (such as real world problem solving) cause it to lose its prompt entirely and revert back to a generic chatbot.

I don't use ChatGPT but also I don't remember the last time I saw an emoji from AI, I've seen it but not very often. OpenAI has Platform for more serious use.

leileilol wrote on 2026-01-07, 18:36:

fuck it

This.

lti wrote on 2026-01-09, 05:12:

To be honest, the way LLMs write reminds me of a high school student trying to pad out a paper to fit a minimum page count five minutes before it's due while also not understanding the content.

That's exactly my impression, most of the time 😀

lti wrote on 2026-01-09, 05:12:

hallucinations

I would not be surprised if this term was coined and promoted by the industry themselves to forward the notion that they really have created an artificial mind. There's totally a ghost in the machine, it just needs some tuning. Confabulation seems more appropriate.

Big Pink wrote on 2026-01-09, 23:24:

lti wrote on 2026-01-09, 05:12:

hallucinations

I would not be surprised if this term was coined and promoted by the industry themselves to forward the notion that they really have created an artificial mind.

They've created a type of mind. If you don't accept this then consider what happens when you fail to A/B human and AI.

vvbee wrote on 2026-01-10, 00:06:

Big Pink wrote on 2026-01-09, 23:24:

lti wrote on 2026-01-09, 05:12:

hallucinations

I would not be surprised if this term was coined and promoted by the industry themselves to forward the notion that they really have created an artificial mind.

They've created a type of mind. If you don't accept this then consider what happens when you fail to A/B human and AI.

Human intelligence is not increasing over time I’m afraid…

The so called AI is just manipulating symbols, something some humans seem to have a hard time doing these days.

I have found that some LLM’s can perform well specified tasks that would have been a drag to do yourself quite well

However, where I draw a solid red line is *asking* “AI” ANYTHING.

As has been pointed out before, it then becomes a shit-up-maker that generates output towards your specification, whether it is correct or not.

It does this shit-up-making with extreme confidence in phase 1, performs condescending gaslighting when contradicted in phase 2, and apologizes profusely when cornered in phase 3. I believe its the phase 3 behaviour that trips some people up to think that the thing is sentient.

I laughed about the writing style matching a school child padding up his essay to reach minimum page count on a subject he understands absolutely nothing about, I couldn’t have said it better myself and will henceforth shamelessly plagiarize it, thank you 😀

When it comes to negative AI behavior, people don't like to apply the same standard as when they say that something novel AI did must've been in the training data. In other words the comparison isn't to the pool of human behaviors but to some ideal proper human. You end up drawing circles that overlap and saying one excludes the other.

vvbee wrote on 2026-01-09, 15:54:

An LLM that outputs emojis is clearly not intended for professional summarization, like you don't take your notes with a crayon either. If it's offered for your professional use case then that's an integration problem. If there are no style controls then you can use paid APIs that give you that control, as a professional you probably should be using them anyway.

The Copilot instance at work lets you turn that off, which is good. I just don't like that it's the default option, even in the free LLMs. I've never known anybody who wrote like that.

sunkindly wrote on 2026-01-09, 17:01:

Not to mention that certain unavoidable triggers (such as real world problem solving) cause it to lose its prompt entirely and revert back to a generic chatbot.

I've had that in Copilot (the only LLM we're allowed to use where I work, and that's only because Microsoft refused to let even IT admins disable it - that option was recently added to insider builds of Windows). If I'm trying to solve a problem, it might take a few minutes for me to have a reply, and then it acts like I'm starting over with a new chat, even though my chat history is still on the screen.

On the subject of Copilot, I don't trust that it's really a private instance. I don't trust Microsoft to not train its regular public Copilot on any trade secrets that someone put in a "private" prompt. I feel alone in that opinion, but maybe someone here will agree with me.

vvbee wrote on 2026-01-09, 19:32:

I don't use ChatGPT but also I don't remember the last time I saw an emoji from AI, I've seen it but not very often.

I see it every day. Those check marks it uses in place of bullet points and the other symbols it throws into responses are technically emoji. They're in your OS's emoji selector.

Every single time I saw it used for work tasks, it proved to be more trouble than it's worth

I do use it for hobby stuff, to do things that a) I couldn't do otherwise b) can't affect anything if they turn out to be bullshit

Companies, however, will use it an excuse to fire people, because "we're using AI now, much future, such advanced" sounds better than "we want to cut costs".

This is not going away anywhere anytime soon, even when the bubble bursts (okay, maybe a little once AI vendors stop offering their services at a loss). Don't hear about a lot of companies that used AI to generate more revenue, it's always about costs.

xcomcmdr wrote on 2026-01-07, 18:18:

I wish it did not exist.

Me too,its not good and its a shame ppl will not wake up until its too late 🙁

i'll be glad when it's gone

sunkindly wrote on 2026-01-09, 17:01:

At least with a ChatGPT custom GPT, no matter how many different ways I try to put it in the prompt for it to not use emojis, or bulleted points altogether for that matter, in addition to directly telling it in the conversation not to use emojis, it still ends up using them. Not to mention that certain unavoidable triggers (such as real world problem solving) cause it to lose its prompt entirely and revert back to a generic chatbot.

It is easy to do so.

When you first start a chat, you give it a list of rules to abide by.

You can even add rules after starting the chat.