Uses of AI You're (Not) Okay With?

[Moonface]

Inspired by @Maniakkid25's recent response to me in their Ask Me thread where they talked about a use of AI that they agree with. I remember a time where the focus on AI discussion was everything bad about it and I admit I couldn't ever really think of applications for it that were agreeable to me, but I've noticed that as the initial knee jerk reactions to AI have settled down now that it's been around for a little bit that not everything about it is met with negativity.

So, what are your thoughts on different use case scenarios for AI?

[Maniakkid25]

I think where most people knee-jerk react is with generative AI. That's the main thing that everyone talks about, and with good reason: how are you getting the file data to feed into the AI without breaking massive amounts of copyright? It's a legal dilemma that our system currently isn't set up to handle, so it's gonna be a while for the law to catch up.

Another big one is AI making important decisions that should really be filtered through a human. While an AI can be good at organizing data and saying "hey, maybe someone should see this", it shouldn't REPLACE that person doing the checking. A rather famous example of why was brought up by John Oliver, with a medical team using an AI to detect cancer in people. The problem was the AI found a shortcut that they didn't anticipate. See, in the training data the AI was fed, the cancers were accompanied by rulers for scale. The AI thus determined that rulers meant cancer, and looked for any rulers in the image. This rendered the entire AI completely worthless, because now they have to figure out a way to "untrain" that from the AI. It's a classic cautionary tale that a computer should never MAKE the decision, because computers can't be held accountable. They can AID, but not OVERRULE.

[Mr EliteL]

Artwork in general, I hate looking at most AI art, but that's just me.

Also to a much bigger extent, agreeing with Maniak. AI needs someone to overlook it's process and results, not to be left alone to make the decisions that should be made by a qualified professional.

[Moonface]

I think where most people knee-jerk react is with generative AI. That's the main thing that everyone talks about, and with good reason: how are you getting the file data to feed into the AI without breaking massive amounts of copyright? It's a legal dilemma that our system currently isn't set up to handle, so it's gonna be a while for the law to catch up.

Generative AI should have only been allowed to collect data with permission, but that was never going to happen and it's too late to do that now. Even then though, most uses of generative AI I can think of are ones I'd hate no matter how the AI collected its data.

Another big one is AI making important decisions that should really be filtered through a human. While an AI can be good at organizing data and saying "hey, maybe someone should see this", it shouldn't REPLACE that person doing the checking. A rather famous example of why was brought up by John Oliver, with a medical team using an AI to detect cancer in people. The problem was the AI found a shortcut that they didn't anticipate. See, in the training data the AI was fed, the cancers were accompanied by rulers for scale. The AI thus determined that rulers meant cancer, and looked for any rulers in the image. This rendered the entire AI completely worthless, because now they have to figure out a way to "untrain" that from the AI. It's a classic cautionary tale that a computer should never MAKE the decision, because computers can't be held accountable. They can AID, but not OVERRULE.

I don't know what applications it would have in a medical sense, but using AI to automate menial repetitive tasks and generate an alert when it picks something up out of the ordinary is the only thing I can think of for a possibly good use for it. However, I feel like if that was something we could do that it would've been automated before AI became this big thing because in my mind when you boil such a task like that down it's the equivalent of Word being able to automatically check for spelling errors which doesn't need AI to do it.

Artwork in general, I hate looking at most AI art, but that's just me.

Also to a much bigger extent, agreeing with Maniak. AI needs someone to overlook it's process and results, not to be left alone to make the decisions that should be made by a qualified professional.

AI art is really just ugly to me and most because it all has the same look. Everything looks weirdly smooth and airbrushed, with this sorta uncanny valley look to it like it was trying for realism but at the last minute a Pixar filter got poorly applied over it.



For me, I don't mind AI that isn't taking a job away from someone else. Art for example, if you want to make it for personal use then whatever, but if you're making AI art to gain from it then get bent. If you're using AI to help automate a simple task so you or others can spend more time and resources on more important aspects (i.e. a game using AI for random fodder NPC's so the actual artists can spend more time working on the notable characters) then that I'm fine with because you didn't take a job away to avoid paying someone, you just allowed people to focus on more worthwhile tasks that can benefit from more time.
Some AI is pure garbage no matter how it gets used though. Google Search AI is fucking awful, and all it does is scrape the content from a website and display it at the top of the results. Ignoring that it isn't always the right info because the AI is stupid, the internet is already too condensed as it is and AI search results are just going to make it worse if it causes sites to close down or go behind paywalls to block the AI, or just pushes people away from using search engines period and just sticking to trying to find answers from a singular site like Reddit or whatever.

[Moonface]

I'm watching Trihex right now and he's got this video on about how AI is going to affect content creation, and some of the stuff in it is really messed up to me like how it'll become hard if not impossible to know if someone is real or not because the pictures and videos will be so convincing:



There's also AI channels on Twitch in the top 1000 on the platform, along with some YouTube channels pulling in huge numbers too (all at around the 8 minute mark I think). Not sure how I feel about that stuff until I process it for a bit but it's just wild to me that unless we already know a creator from before AI became a thing that we could end up getting into someone we think is real but they're actually not. If the channels are open about being AI then whatever but I bet a lot will try to pretend they're actually real people but aren't.

[ShiraNoMai]

Right now it's fairly easy to tell the difference, at least for roughly 60% of the Twitch viewership, maybe less so for YouTube viewership; but yeah as time goes on, and depending on what is being shown, it could absolutely be real freaky to witness. I still think there will always be a human element missing from particularly specific content, particularly live content, that I think could never be replicated, in the same vein of artwork.

[Karo]

For me one use of AI that I think is just despicable is this one he goes in the video about it:


Basically AI that maximizes profit in the rental industry and it just awful leaving house vacant to drive up the rent and the landlord can pull a "It not us doing it we are just following the order of the algorithm"

Now of course I have a problem with the usual generative art stuff and the content farms. It all really just messed up.

[Moonface]

Ugh, as if I didn't already have enough reasons to dislike landlords in general. They always come off to me like a condensed version of the worst of capitalism; taking advantage of people and putting them through the worst they can get away with while at the same time maximizing their own pocket filling. I'm sure they'd rent out a dumpster to homeless people if they could.