Discussion
Serious question: If you willingly use generative "AI", do you believe you're immune to or sufficiently protected from de-skilling, cognitive decline, and what's commonly known as "AI psychosis"?
If yes: What makes you believe that? How are you special?
If no: Why do you still use it, given that you're aware of the risks to your cognition? Do you just accept it?
@aesthr The number of people in the replies who don't qualify as "willing" users is flat out terrifying. I've never seen a "tool" forced on people like this
@aesthr I don't willingly use it, but the mental rot is still there. A lot of it stems from abandoning any sense of ownership over code and projects, and checking out from the corporate employer. None of my previous employers really valued genuine technical expertise, either. Anyone who ever held the potential to make *great* software was always directed to dull their mind's edge by repeatedly striking it against hard rocks. They either gave up or moved on. But now, nowhere to move on to.
@aesthr I’m using AI and I work very closely with LLMs - I’m building engineering system around them, I’m aware of limitations and I’m very critical when it comes to their output. At the same time I see how llms can be useful when used responsibly. As always, critical thinking is key. I'm not worried about psychosis I'm more worried about losing skills, but then again, who can do quick calculations without a calculator these days? We lose some skills and gain others
@aesthr willingly? Sort of. We are forced to use Copilot at work, and I do find it useful for data entry type tasks, pulling information from Acrobat files into spreadsheets. Never ever let it write my emails or anything like that. Loss of skill? Maybe? I used to get more creative trying to work around the insufficient reporting we get from various sources but on the other hand, now I have time to review the content of those files so I'm not sure. I try to do the work humans are better at.
@aesthr Does using it to generate job interview questions to help you prepare count? I've had a really hard time lately finding relevant questions the traditional way and you can prompt a chatbot to be as punishingly challenging as real interviewers. Besides - they're probably getting those questions from a chatbot as well.
@aesthr I personally am not too concerned about using AI because i don't use it to solve the problems of the world but like any other tool i'd use to solve an issue i have. I'm more concerned about AI usage in the news and the fact that felt 90% of Youtube videos are AI generated now and nobody made even an effort to make them interesting, just hourlong text fillers to tell a 10 minute story. Saying AI steals our time mostly.
@aesthr no, on the contrary, I learn new things. It is a tool. I think people said the same when other technologies were invented (eg. internet, smartphone).
@Thales_Curiosities except it's not "just a tool" https://estherweidauer.com/articles/why-resist-generative-ai.html#its-just-a-tool
@aesthr 1)So by asking you actually want to impose your opinions on others.2) the text you linked has grammar issues ("it's" instead of "its"). Thanks anyway.
@aesthr No, I don't believe I'm immune, and I've seen people I used to respect fall for it. I use it very sparingly, mostly to check what my students who use it are exposed to, and it's very scary. The other thing is I am very exposed to second-hand LLMing (reading my colleagues' slides or my students' work), and this is also affecting me because of the energy it takes to push back.
@aesthr just about the only thing I use is the YouTube AI summary on videos which are too fucking long and could have been a text document, and even then it’s just to get timestamps for salient moments. And that’s just a shitty fix for a different shitty problem (the proliferation of podcast-style “edutainment” as the only source of information about a bunch of things). And I hate using it even for that.
@aesthr the other thing is if I’m super desperate to learn or find an API that has shitty documentation, sometimes Gemini at least points me in the right direction. And I hate that too. I wish people were better at writing docs, or making APIs that actually make sense.
@aesthr I was a CTO 25 years ago, Been an architect since. I have trained dozens of young programmers. Handing work off has caused me no issues that I am aware of. I don't write bash scripts, have no interest in learning, and Deepseek is pretty good at it.
I am quite sure that using LLMs in every part of my life would do me real damage. So I don't.
Driving a car will reduce your ability to walk long distances. I'm not giving up my car, but I do go to the gym every morning.
@fatsam so, what's the gym equivalent for your brain?
@aesthr If yes that's already an early symptom of being susceptible to it
@aesthr no because I effectively have a gun to my head in my career. If I don’t embrace AI and figure out how to work with it, I will be fired.
It’s like being demanded to use cocaine or something. It’s awful. I know it’s bad but I don’t know what else to do right now except keep my eyes open for an opportunity to leave software engineering.
@pandora_parrot hence the "willingly" in my post
@aesthr yeah. I forgot to mention that I’m specifically referring to my software work outside of my job. That’s closer to “willingly,” but in part I use my personal projects to build experience with this stuff so I can work more effectively, so it’s borderline.
@aesthr
I hadn’t really thought about it before. I’ll admit that I delegate quite a few things and have been using AI daily for about four years. But, thinking it over, I realized that this fits what I want: I like orchestrating the process, and there are some things I’m simply not interested in learning how to make. I’m happy with the final result.
@willianf are you not concerned about your very thinking becoming dependent on a service you have to pay for?
@aesthr
Your question made me reflect. I’ll be more mindful from now on. The idea is to use AI as a reference or supplement, while building the habit of thinking things through and trying to solve them myself before asking AI. Thank you.
@aesthr my answer is "no" to all.
But I'm an unwilling user, working for cultists
@aesthr I have a disability that prevents me from doing heavy research. I will *occasionally* use an LLM to get a gist of something whose understanding is not critical and would be dangerous for me to reach on my own. (Once a week, max) E.G. I recently asked ChatGPT why asbestos is dangerous but fiberglass and carbon fiber are less so. I don't work with any of them, it was just a nagging curiosity.
As for skill loss.... You shouldn't be using it in areas where your skills matter.
@aesthr I think nobody is immune, and I'd be more worried about someone who says they are.
If we learned anything from the ongoing COVID pandemic, it's that people are singularly incapable of assessing risk, and extremely skilled at rationalizing and minimizing injury post-hoc.
@aesthr In the months when I started using LLMs on the daily I went through periods of psychosis where things felt one or more of the following: (a) what I'm doing is meaningless, (b) high's like you can approach any problem and solve it quickly (reality check, of course you can't, bit this is what power or empowerment feels like, (c) depression like perhaps I'm not really that good or good enough anymore, (d) elevated anxiety due to constant context switching.
I am concerned about what you mention, and I do wonder how best it should be approached without losing the the technology. One thing I'm exploring is developing and learning more problem solving thinking tools, but right now I'm more focussed on the tool than using it, so that's a bit of a moot point right noe, but perhaps not in the future when I start using it.
I think we will remain able to construct challenges in the small that translate to changes in the large, like if you play with lego blocks it might translate to other things in life. I expect a lot more higher level programming languages to be developed if we care about understanding. And I'm quite sure that many folks still want to understand things, I know I still do despite it being hard and not always possible. I'm humbled by the issues and bugs I've created that an LLM picked up and pointed out. I believe there are enough folks out there that still want understanding.
@aesthr The number of people who seem otherwise sensible who are saying “I was talking about this with Chat GPT and it said…” blows my mind. They are using it as a sounding board on personal issues or to ask if there’s scientific literature to support something they were thinking about. 🤦🏻♀️
It’s like those creepy toys in the 80’s and 90’s (maybe they’re still around) that you could push a button or speak a phrase and it would tell you that it loves you. But these are *adults*. 🙃
@aesthr I try to avoid deskilling by not using AI to generate anything that I actually end up using as-is. Meaning:
1) I mostly use it to review/search things that I then analyse myself
2) In the rare cases I do ask it for some reference code (e.g. something I don't want to necessarily be proficient in but need to use that one time, etc) I always re-type and reorganize the code myself and search the documentation of every construct used that I'm not familiar with to see if it fits, and learn
@aesthr yes, because I think those risks are overstated. 🙂
I organically time-limit my use and have reconfigured the personality because I don't want to be friends with a chat bot, for the same reason I don't want to be friends with my electric tea kettle. My smart home assistant is called "Computer" for similar reasons.
As for atrophying skills, I feel that's like saying that using a microwave for a few meals a week means that I'll forget how to cook. My lived reality is that I don't need to cook a roast dinner every day to remember how to cook a roast dinner. 🤔
Finally re: cognitive decline, I guess that depends on what people do with the time they're saving over a sustained period of time. Some people will use it to improve their mind and body. Others will indulge in vices and pleasures of the flesh. Whatever they were doing before AI is what they'll probably continue doing unless they decide otherwise. 🤷🏻♂️
> because I think those risks are overstated
these effects are pretty well studied by now. You might be seriously underestimating the risk there
@aesthr possibly. I tend to be sceptical about people taking hard-line positions for or against AI because I've seen people do this with several technological revolutions. I also grew up during peak hysteria about kids playing video games, so my instinctual reaction is to eye-roll until I see the evidence and verify the source.
Another thing to perhaps point out is that I don't live in the US, so I'm not living with the constant butt-clenching anxiety of what my country's government is doing. So maybe I'd be more sceptical of AI under those circumstances? 😅
@trechnex the evidence is there though, I linked a whole bunch of studies (even from places like Microsoft themselves) here:
https://estherweidauer.com/articles/why-resist-generative-ai.html#psychological
I also don't live in the US, and have been around for the rise of the Internet in the 90s and the moral panic about video games. This is very different.
@aesthr de-skilling no, It's probably just a matter of lazyness and speed.
I also know that I should spend less time on social media and more reading books or exercising against cognitive decline, but that too is hard.
AI psychosis: not really worried. I am fully aware that LLMs do not think in a traditional sense of the word. It's a powerful, though potentially harmful to society, tool, nothing more.
@aesthr The risks to my employment rank higher.
@aesthr (I guess that's not really "willingly". But I do employ cognitive prophylaxis at work to avoid skill rot and psychosis.)
@afeinman if you want to share, what are those prophylactic things you do specifically?
@aesthr I could write a book.
(Actually, I did, but we're waiting on the cover art, and it's a novel, not a list of prescriptions.)
Tl;dr: agency, locus of control, and consent. Will think about how to extricate the intertwingling of the actual presentation I have made on this from $work.
@aesthr I think I'm safe(ish) from AI psychosis, but definitely suffer from de-skilling and cognitive decline and general I-hate-my-job-now-ness. I use it exclusively at work because I'm forced to do it. I don't think I'm special, but I hope that knowing the basics of how AI works, and using it exclusively on my job and completely avoiding it elsewhere helps me not to suffer from that specific problem.
Install Fedizen
Get the full app experience
@aesthr I am in the position where I don't have to use it (due do currently doing else).
However, I used it for some projects. Setup opencode with the free tier of a chinese engine. Did some simple stuff and a font previewer in Vala for linux/gtk.
I stopped using because the previewer kept on leaking memory like hell and, due having done it in a language I'm not familiar with, I wasn't skilled enough to fix it.
On that day, I felt my brain going mush. I was very, very scared.