On common criticisms of AI

[bulletttproof]

I'm not asking you to like AI. I don't, I have mixed feelings. But many common arguments against AI are weak and fall apart when you think about them. And, the biggest threat of AI isn't talked about enough.

AI is reducing thinking skills
There are no studies that prove this. However, a commonly cited study shows that when performing certain tasks, relying on AI will stimulate areas of your brain less than solely using your brain. The implication is that overreliance on AI will lead to a decrease in learning skills. The study isn't comprehensive enough and lacks the correct methodology to solidify such a claim. And, I'd argue that if you rely on AI for every cognitive task you were retarded before so there's no loss.

You should use your brain instead of AI
Many people saying this lack intellectual humility. Chances are, you don't "know" more about economics, or history, or chemistry, than an LLM. I'm sure an expert would, but the vast majority of us aren't experts of any academic disciplines. Unless you are, you may learn something from prompting an LLM, or be presented with a different perspective. Would it better better if you read a textbook, or completed a degree? Yes, but that doesn't mean it isn't beneficial.

Also, to "use your brain", i.e. think, you need information to work with. LLMs can be a useful source of such information, or at the very least a starting point. They can provide incorrect information, but ChatGPT hallucinates far less than it used to and now provides sources.

There are caveats. Because the information can be incorrect, you should verify it. And, because LLMs tend to be agreeable, you should ask it to push back on your argument.

AI is destroying the environment
You may have heard various figures surrounding AI's environmental impact. A ChatGPT prompt uses _ mL of water, or a data centre's yearly energy use is equivalent to ____. These figures are useless with context.

Let's compare cumulative water use to other streams. In 2023, U.S data centres consumed 17 billion gallons of water, according to Lawrence Berkeley National Laboratory. Including indirect, water use, total data centre water consumption is estimated at approximately 228 billion gallons per year. I'll now compare that with irrigation. The latest USGS national irrigation consumption estimate was 27.6 trillion gallons per year. That's 121x data centres. Mind you, AI use currently represents a subset of total data centre computation.

Data centre water use is an issue for certain regions, but blown way out of proportion. There are also other methods of cooling that are being experimented with, such as underwater data centres.

Note: Data centre water use has increased since 2023, and the figure for agriculture water consumption is from 2015-2020. These figures are from the latest available credible datasets. Indirect water use of AI and irrigation is excluded for a better comparison.

AI slop will ruin the internet
Every corner of the internet has become worse over time. Porn ads, porn bots, OF promo, "X", short form content, search engine decline... Everything listed precedes modern AI. It would've been nice for AI to be the final nail in the coffin but it's just a different flavour of slop.

AI is going to take all our jobs
This is my biggest concern. Full AI automation would alter humanity as we know it. The implications are huge. You can talk about class divides, corporate power, and more. Yet my biggest concern is philosophical. How will man reconcile with the fact that he is useless, and always will be?

 

[Sneaky lurker]

This is not reddit nigga. We all hail hitler over here.

AI is going to take all our jobs
This is my biggest concern. Full AI automation would alter humanity as we know it. The implications are huge. You can talk about class divides, corporate power, and more. Yet my biggest concern is philosophical. How will man reconcile with the fact that he is useless, and always will be?

>Society actively rejects and replaces you and millions more.
>Become part of a displaced demographic.

>|
|
_______________
|..............................|
v............................. v
Exit society: start a new one. Revolt and take back society.​

>profit.
Not deep.

 

[batterybee]

ummmm.... achsully there are no peer reviewed academic studies that affirm what everyone can clearly perceive so it isnt happening chud.

 

[Goose the Redeemer]

AI is reducing thinking skills
There are no studies that prove this.

Yes. Let the machine think for you, just as you defer to the authority of science-man.

Chances are, you don't "know" more about economics, or history, or chemistry, than an LLM.

I probably know at least as much as the average redditor, wiki editor, or stackoverflow commenter that went into training the LLM.

Also, to "use your brain", i.e. think, you need information to work with. [...] Because the information can be incorrect, you should verify it.

I need information, but, the LLM might give me wrong information so I need to do my own research anyway. What exactly does this give me over just skimming a reddit thread or wiki page?

I'll now compare that with irrigation.

Food is a necessity, you can live without an LLM.

It is interesting that you didn't mention the obscene electricty demand. How many of these datacentres aren't even online yet because the grid can't actually support them?

AI is going to take all our jobs
This is my biggest concern.

I'm sure it is.

 

[Chuck McGill]

Anyone earnestly worried about AI taking le heckin' jobs has outed themselves as someone that hasn't been in a white collar environment recently. 90% of the jobs are already fake and gay. They could have easily been automated with technology we had access to in 2019. If Google Calendar didn't make them obsolete, why the fuck would I be worried now?

 

[teriyakiburns]

90% of the jobs are already fake and gay

They're UBI in disguise.

 

[Lemmingwiser]

Yes. Let the machine think for you, just as you defer to the authority of science-man.

Don's use the slide rule to calculate either.

 

[Bioniclelover]

I don't need a fucking peer reviewed study to notice that AI is reducing thinking skills, I can just look at what Karl Jobst and SmashJT are doing.

 

[Penis Drager 2.0]

Talking to an LLM is a bit like talking to the most autistic internet know-it-all you can imagine. Except at the same time it will break its back to concede and agree with you if it can.
It can vomit some some pretty decent information at you if you ask your questions in the right way and can't be fucked to do "legitimate research." But it will sometimes hallucinate shit out of thin air based on something it half "remembers" from a completely unrelated concept.
It serves a lot of the same functions as a Wikipedia rabbit hole but much faster and slightly less reliable.

Pros:
Ask a question; get an answer. It's easier than sifting the library that is The Internet looking for the right place to answer the question you have.
Infinite patience. Your professor, if you're in school, doesn't have the time to answer all your dumb questions that are likely only tangential to his field of expertise anyway.
Cons:
It's too eager to validate. The company running the machine wants you to keep using it. So it is designed to keep you engaging with it. It will almost never say "you're fucking retarded, what you're trying to get at is so wrong that it would need an hour long 7th grade lecture to even get you on the right track." It is designed to say something more like "that is a brilliant insight, but here's how it can be made stronger..." with the following statements refuting the core premise and the user not bothering to read it.
Easy != good. You already elaborated on the neurological pathways that appear to be suppressed by extensive AI use. So I'll leave it at that.
Hallucinations need to be on your mind constantly when working with these things. You have to engage your brain and think about what the machine is vomiting at you. I personally enjoy feeding 4chan threads at LLM's and watching it try to make sense of what's going on. I'd say it's about 70% solid but the other 30% is attributing claims to the wrong user and inventing arguments that were not made.

As far as "AI taking the jobs" goes, those are typically application-specific AI's and it's frustrating that people conflate this with toys like ChatGPT.

The bottom line is that LLM's are a perfectly valid tool. But when used as a crutch they actively make you dumber. This whole post has gotten drunk and rambling so I'll leave it with "don't be the 'let's ask ChatGPT' guy."

 

[Goose the Redeemer]

Don's use the slide rule to calculate either.

A slide rule will not lie to you, nor will it insist with absolute confidance that it can give you an answer with 20 significant figures. No-one would accept a calculator that sometimes gives you the wrong answer, yet "hallucinations" are largely accepted as an inherent part of the LLM.

 

[Lemmingwiser]

A slide rule will not lie to you, nor will it insist with absolute confidance that it can give you an answer with 20 significant figures. No-one would accept a calculator that sometimes gives you the wrong answer, yet "hallucinations" are largely accepted as an inherent part of the LLM.

Believing the confidence of an answer is a general human flaw.

It's not just about the quality of the tool, but also the quality of the tool usage. A slide rule is useless in most people's hands too.

 

[bulletttproof]

Yes. Let the machine think for you, just as you defer to the authority of science-man.

That's exactly what I meant. Thank you for understanding me.

I probably know at least as much as the average redditor, wiki editor, or stackoverflow commenter that went into training the LLM.

Really? About everything? About economics, politics, sociology, physics, chemistry, etc. At best you're proficient in a handful of these fields.

LLMs aren't just trained off Reddit, Wikipedia, and Stack Overflow. They also train off textbooks, articles, research papers, encyclopaedias, and more.

I need information, but, the LLM might give me wrong information so I need to do my own research anyway. What exactly does this give me over just skimming a reddit thread or wiki page?

It might give you incorrect information, anecdotally the chances are very low. You can check though, ChatGPT gives you sources. You can't ask a Reddit thread follow up questions. And an LLM can extract information from tables and reports way faster than you can. You're missing out on a valuable tool.

Food is a necessity, you can live without an LLM.

I don't see how that's relevant buddy. My point was to contextualise LLM water consumption. It's dwarfed by agriculture, which itself is a fraction of total US water consumption. It's not as significant as most people say.

It is interesting that you didn't mention the obscene electricty demand. How many of these datacentres aren't even online yet because the grid can't actually support them?

Sometimes it's due to regional grid constraints. Other times its due to public backlash, difficulty sourcing equipment, and legislation.

 

[Maurice Maine]

You're not gonna find pro-AI arguments anywhere on the internet let alone Deep Thoughts on kiwifarms.net. I think the tech's solid. It's gonna fuck a lot of people I hate like even now influencers now have to compete with AI and they're feeling the heat. I wanna see them thoroughly fucked. It's good for coding too I don't see the downsides.

 

[browserbowser]

Let me solve the problem for you.

Every complaint about AI is a complaint about the effects of immigration.

The problem is these retards don't accept that the hot new thing that it's in vogue to hate isn't the problem, it's immigration, legal or otherwise. Hell, especially legal immigration as the consent of our government(s) to this bullshit makes it that much harder to fix the problem at its source, which is having way more workers in our domestic industries than there should be, for the benefit of the worst people in our economies.

Literally take any complaint about AI, compare it to the endless immigration done to prevent employers from having to deal with a job market wherein there's more jobs than people, and thus having to compete for workers which increases average pay, benefits, etc, and you have the answer.

It lets people bitch about the problems they have without being an icky racist. Because they're turbofaggots who need to grow the fuck up.

 

[Smart Turtle]

AI is reducing thinking skills
There are no studies that prove this

You don't have to access sources for common-sense stuff like that.

 

[Nick Wilde]

They could have easily been automated with technology we had access to in 2019. If Google Calendar didn't make them obsolete, why the fuck would I be worried now?

The perception shift in the elites is that previously we had to maintain these fake jobs to maintain social stability, and keep a semblance of economic circulation via those fake job people spending their fake job incomes; but lately they are actually starting to flirt with the idea that they can drop these people. The paradigm appears to be shifting.

Now, don't get me wrong, as a real job haver, I won't feel particularly sorry for the HR ladies and middle managers who end up on the short end of the stick here, but I do think those people have a reason to be worried. They should have thought about the implications of their position within the hierarchy a lot sooner.

 

[nycifer]

I've spoken about AI here and there. I quite frankly find the Anti-AI crowd very intolerant, because they spread extremist ideals and that collides with me being anti-extremist so they all can go fuck themselves at the end of the day.

I find that a lot of the criticisms against AI is entirely blown out of proportion, exaggerated and leans heavily on fearing the unknown. There are some valid concerns somewhere in the bullshit, but leave it to a lot of dumbasses online to screech to the heavens about the sky is falling and projecting doomer outlooks on everything.

There's a net positive to using AI, but leave it to no-life people to mess with something they don't understand. I would rather speak to an AI, than some idiot who is burning what little of their braincells that there are away, just to give me a simple fucking answer to a curiosity I had.

Being mislead by an LLM AI is no different than a human, but the key difference is that AI isn't literally doing it on purpose, it is doing it based on what you try to talk about it with. With a human, anything is on the table from how they mislead you on purpose to them actually not knowing or perhaps not giving a shit.

I just find it all silly, how so many people get upset about AI but there aren't enough people upset about how we took for word, about some unproven invisible deity who dropped one of the most boring books imaginable out from their ass and we chose to live by it. I'm more concerned about those nutjobs running anything and being of any help, than anything regarding AI.

 

[shatterspinel]

I’m still not sure what the huge deal is, there’s a few nifty applications but literally nothing that boomers didn’t think that bots and photoshop was already capable of 20 years ago. If anything, I spent more time imagining an artificial intelligent companion that would actually help me figure out how to create really fucking cool stuff, like I did back then when both sides where still aware they were talking about science fiction

 

[GOY MENACE]

Why niggas Chimping out?

If im interested in a topic (for example safed at its golden age) i ask ai for book recommendations on the topic and its excellent at that.

I ask ai to find thinks i vaguely remember and its find them.

When i ask it questions it gives me the desired outcome most of the time.

So what's ths problem?

Some people abuse it and are addicted to it but obviously that doesn't mean its useless.

 

[Do Not Touch Me Pls]

AI is reducing thinking skills
There are no studies that prove this. However, a commonly cited study shows that when performing certain tasks, relying on AI will stimulate areas of your brain less than solely using your brain. The implication is that overreliance on AI will lead to a decrease in learning skills. The study isn't comprehensive enough and lacks the correct methodology to solidify such a claim. And, I'd argue that if you rely on AI for every cognitive task you were retarded before so there's no loss.

How about don't rely on AI? For now it's still avoidable somewhat.

AI slop will ruin the internet
Every corner of the internet has become worse over time. Porn ads, porn bots, OF promo, "X", short form content, search engine decline... Everything listed precedes modern AI. It would've been nice for AI to be the final nail in the coffin but it's just a different flavour of slop.

This is my biggest gripe. It's not natural to say the least, it is a human-engineered mass extinction event rather than any natural happening. While everything goes to shit after a while, AI just shits everything up to the point where actual video games, content creation (save for drawings which are much harder to replicate) and viral internet memes outright go extinct. The last viral internet meme I saw is the Axel in Harlem one and the last indie music hit is Tetoris by Hiiragi Magnetite.

At one point people will lose the will to even pursue anything but war because of AI ruining everything.

AI is going to take all our jobs
This is my biggest concern. Full AI automation would alter humanity as we know it. The implications are huge. You can talk about class divides, corporate power, and more. Yet my biggest concern is philosophical. How will man reconcile with the fact that he is useless, and always will be?

AI is way too inaccurate in a lot of aspects in terms of jobs. Drawing cannot be faked with AI at least.