When I say I'm neither fully pro nor anti AI, what I mean is that I don't believe the technology is fundamentally evil, but the vast majority of it has been created in bad faith, by corporations that I am against, and I believe it's predominantly harmful and untrustworthy in its existing form. Don't feed that machine, please and thanks.
The problems with the AI boom are that it's based off stolen data - art, literature, social media posts, etc - and built in an entirely non-sustainable way that actively harms the planet, takes advantage of its users, misappropriates intellectual property, manipulates the population, spreads misinformation, perpetuates psychosis and harmful behavior/thought patterns in vulnerable users, replaces jobs, and distorts the value of critical thought and artistry in people's minds through even more increased demand for commodification of such things in a fast, easy-to-consume format that leaves us questioning what's real. Legitimate artists and authors are being accused of using it when they haven't even touched generative systems. It's made us paranoid, and societal recovery from that will take years, if not decades.
Now, "AI" is a catch-all umbrella term that overgeneralizes the technology, and gets people thinking of LLMs and image/video/sound generators. There's far more applications than that. And with the way they're creating and selling it, it's doomed to fail. It's entirely non-sustainable, even for those who only care about making a profit. Saying "I don't use AI," to mean ChatGPT and the like isn't entirely accurate; if you play video games, those do use a form of AI. If you use modern GPS, chances are it's using AI to determine the best route to minimize traffic and calculate a travel time. Spam filters? AI. Credit card fraud detection? AI. It's such a broad term that it's functionally useless, because these examples are so far removed from what we think of with generative AI.
What I'd like to see more of when the bubble pops is ethical, specialized AI tools made by smaller, more educated teams with a dedicated goal and passion, since the technology has already shown great promise with applications in areas like the medical field and aiding archeologists. The key word here is aiding, not doing it FOR them. Technology should be supplementary to our lives, not a crutch for a lack of effort or a replacement for human work.
Ethical AI could mean a lot of things, but my take is that it involves not-for-profit applications, limited use-cases, comprehensive protective regulation (particularly to protect data privacy and mental health, as well as prevent exploitation), augmental uses designed as a helpful tool for human workflow rather than a replacement for people, adequately paid moderation staff, rigorous testing by legitimate outside sources, and most importantly, two things: 1) consent-based and compensated usage of data in any form, meaning creating AI databases and completing training processes would require legitimately obtaining IP use rights and paying for the use of other people's work and/or data, on an exclusively opt-in basis, and 2) Sustainable, Earth-friendly server setups that don't disrupt the wellbeing of communities or ecosystems. Yes, it's possible, and no, corporations won't do it, because not quick, not easy, and especially not profitable. The current system is barely profitable to begin with - they're bleeding money all over the place as they bank on the dwindling success of AI. But if it's worth doing at all, then it's worth doing right. Potentially dangerous technology like this should take years to build and implement, in order to ensure its safety.
With sufficient effort and actual care, AI technology could be utilized to create ethically-sourced and highly accessible research databases of selectively chosen, comprehensive information on specific subjects, such as law, history, medicine, etc., that returns the information you asked for with fully cited, already-existing sources that can be checked, and allows you to specify exactly what you need to know. In its ideal form, this would be nothing more than an improvement on existing database and search engine technology, instead of the rampant enshittification we've seen from, say, Google. Because of the way AI has been implemented, getting factual information is harder now, not easier, and that's the exact opposite of what it should be doing. Secure, reliable models could be built by doctors to aid doctors in making a diagnosis, by pointing out factors they haven't thought of, and could provide safeguards against human error. And, as I mentioned earlier, it has applications for translating ancient texts previously unreadable.
Overgeneralized models built on broad scraping are far more prone to having hallucinations and false info, and thus spread misinformation, but smaller ones still need to provide the user with the ability to easily and quickly verify information accuracy. People should be comprehensively taught not to immediately believe it without checking, as part of basic safety training in every area of its use. Even so, the quality could be immensely greater than what we've been seeing. It could actually be a useful tool to help people, from research for authors and students, to reducing common mistakes and oversights in medical care, but instead... we got the "do my homework for me," "hey AI, do my critical thinking for me," and "make art so I don't have to learn it myself or pay an artist" machines. In actuality, there are virtually limitless applications for AI systems designed around one or two exact tasks that could benefit society. I remember when the concept of "smarter" AI was brand new, and some artists were excited about its potential to help us visualize and such, but the technology usage was immediately revealed to be a weapon built on stolen work. Those of us with hopes for it were left feeling naive and used. Meanwhile, my favored art program, Rebelle Pro, utilizes AI for something called NanoPixel 2, and it's extremely useful for virtually lossless image scaling, macro zoom to tweak fine details, and imitating the look of traditional art materials such as oil paints. Traditional painting is exceedingly difficult for me due to a hand tremor, so it's wonderful to use a digital art program that can accommodate aspects of my disabilities while still allowing me to comfortably play around with traditional techniques and simulated materials in my art. Learning to utilize it still requires practice and education around traditional art fundamentals. I also use Glaze when I publish my artworks online - it uses AI to combat theft via AI scraping, turning AI back around to protect artists' work.
Regardless of whether or not you personally like my examples for how it can be used (I'm just spitballing here), my point is that the potential exists to build AI systems that actually function without rotting the brains of their users; they could aid meaningful efforts rather than obstructing them and stealing from people to sell their own work back to themselves in a regurgitated format, but the makers of nearly all of these models are in it for money. For them, it's not about imagining what's possible, it's about what makes people easier to exploit. As someone who experienced a touch of AI psychosis via Character AI a few years ago, I have strong feelings about that.
"Efficiency" is not all that matters in life. I'll never trust AI while it's being created and controlled by corporations, rather than people who care about what they're making and what impact it has on society. If you're going to use AI, be conscientious about the ethics involved, maintain caution and a very critical eye, always fact-check where relevant, and avoid AI products that have been shown to cause harm, such as those made by big names like Google and OpenAI. Corporations are abusing a valuable technology to the extreme and abusing us with it in every way they can. Don't support that. We can do better.
To a hammer, everything looks like a nail. To a capitalist, everything looks like profit and power. In conclusion, capitalism ruins everything and we should seize the means of production... and with proper discretion, that includes AI.