by Doc – Owner, Founder, Not Sure Why He Ever Prompted an AI to “Make it Sexy” But Oh Well, It Was Funny Anyway
A very significant portion of my day-to-day workload revolves around words, and most of that workload is solved with a handful of AI models. Chief among them is the Rolls-Royce of ChatGPT models, currently the o1 Pro model (though I maintain that Sonnet 3.7 surpasses it in every practical way). I likely have exponentially more practical, hands-on experience with a variety of AI models than everyone who will read this post, due in no small part to the fact that my small audience are primarily younger people who haven’t been alive long enough to get that experience.
I’ve used AI to write press releases. I’ve used AI to write Python programs that optimize several administrative functions. I’ve used AI to write statutes that are now on the books (not without putting on my attorney hat and revising them, of course). It’s safe to say I know my way around most of the consumer-centric forms of AI available on the market.
I am not worried about some sort of impending AI catastrophe like most, though I’ll go on the record as saying that the most likely form of an AI catastrophe will be caused by cybersecurity breaches where AI is used as the tool, not acting independently. I’m not worried about AI for three reasons:
- The AI models’ growth is obtained by its consumption of data, which 0can’t outpace its starvation for quality, human-generated content, resulting in the ‘Ouroboros Effect’. (This is an interesting article on that topic, by the way.) There will come a point where there simply isn’t enough human-generated data for AI to continue to increase at the rate that it has, at which point the models can only be tweaked to improve their skills in more specific applications like programming.
- All forms of AI on the market either run on consumer-level hardware and don’t have the quality and level of inference comparable to models like ChatGPT and Claude, or run in expensive, energy-hungry compute centers that are subsidized at a massive loss by venture capital… and venture capital will want a return on its money or shut down the program.
- I am only able to achieve the high-quality results out of AI that I get because I am highly trained in use of the English language; the best AI results come from using niche, specific words (such as “epigrammatic” instead of “slogan” when writing copy), and an AI will not typically suggest the right words for you to prompt it with –as a result, most users of AI will not prompt at a level sufficient to create high-quality properties that will actually surpass human creations. In other words, people have taste, and AI can’t beat that.
Hopefully this puts your mind at rest somewhat, though I’ll remind you that while I am an attorney, I am not a computer scientist or an AI engineer. And I have never and will never use AI to write my Amiibo Doctor content.
See you next time,
Doc
Amiibo Doctor 