The Real World of Generative AI

I had a call with my best friend from college the other day. While I went into software development for the web, that guy is a scientist playing with plasma physics. He’s also a world-class Ultimate Frisbee player, but that’s probably best for another blog. Anyways, he asked me my thoughts on AI. I gave it to him straight: I believe Generative AI to be an unforgiving and unforgivable technology.

I stole that phrase from Ursula Franklin, who was an amazing woman. In her 1989 Massey Lecture, The Real World of Technology, she says that “nuclear power is an unforgiving and unforgivable technology.” She says this as a research physicist that refused to work with nuclear power. She said that nuclear power left an indelible mark on both the planet and mankind and that no use of it could excuse that unethical side effect. I believe the same thing with Generative AI. While there is nothing as dramatic as Hiroshima, Chernobyl, or Three Mile Island to point to as an example of the unforgiving impact of Generative AI, the evidence continues to mount.

Is Generative AI itself going to destroy the environment? No. Its data centers are. For the last decade, cries from climate scientists that we are nearing the point of no return have become a louder and louder chorus. While some nations, like China, have decided to try to pivot to greener options for power demands, the good ole’ US of A continues burning coal and other fossil fuels and even looking to open up ANWR to get more. And so as Generative AI calls for more computing power through its data centers, that increased power demand has caused at least fifteen coal plants to reopen instead of retire. Or the data centers themselves, like Musk’s Grok data center in Memphis, TN, are emitting 35 methane gas turbines with no Clean Air Act turbines.

This also doesn’t even touch on the fact that much of the data that goes into generative AI is either populated by exploitation of workers or just straight up stolen.

At the same time, Generative AI is also causing irreparable damage on those who use it. Psychologists have gone as far as to coin a new term, “AI psychosis”. Psychologists, which are saying this is not yet a clinical diagnosis, have stated that “AI models have amplified, validated, or even co-created psychotic symptoms with individuals.” We see examples of this with individuals either murdering others at the prompting of ChatGPT or even encouraging them to kill themselves.

I want to cap this off by talking about a colleague of mine. Let’s call him Fred. Fred is a developer that almost exclusively uses Claude. Fred only writes what Claude tells him to. When I challenge him on code (even good code), he quickly says “oh I don’t know how that got there.” I don’t know if Fred ever had curiosity about code – but its absolutely gone now. He has no interest in improving the codebase, making it maintainable, improving in his craft. This is the future employers want. I’ve heard execs talk about how “sometimes Claude surprises them” and, to me, this sounds like someone being surprised that a Magic 8-Ball was correct.

I have both family and coworkers that talk as if generative AI is inevitable – it’s just like the internet, right? And I know I’m not the only one that feels like AI is absolutely anathema to me. The only good side effect is that I had a mental block in my own art forever – it had to be perfect. Now, imperfect art feels like a telltale sign that something was made by hand. I cherish the imperfections in my art now. I lean into them. Its made me grow as an artist.

My "art." Dumb enough that only a human could make it.