My Relationship With AI
I didn't know it would come to this.
When I think about where AI was when I first started writing about ChatGPT in early 2023 and where it is today, it may as well be two different subjects. Back then I couldn’t get people to even try LLMs, and in spite of this fast-moving evolution, it seems like the majority of users still haven’t moved past search and retrieval. Meanwhile, AI has moved on, and so have I.
I get up every morning and read the latest AI news on Rundown.ai, and it all seems to be moving faster than I can comprehend, with the technology heading in every direction at once. I haven’t written much about it lately because I don’t even know where to begin.
I do know there are lots of wild predictions out there about AI’s capabilities and status and future, and they run the full range, from a doomsday technology to the best thing that ever happened to the human race. Personally, I feel way more positive about AI’s future than negative, but I may be in the minority. I still use the history of technology as my guide to predict the future.
Last Friday, Anthropic withdrew its top model, “Fable 5,” something I recently started using. They did it for national security reasons. Oh? What in the world does it know how to do that’s so threatening? It apparently knows something that could cause harm on a national scale. It said someone could “jailbreak” the security safeguards. Again, oh?
It seems like there are very few science fiction movies that show the future as a positive thing. The overwhelming majority see the future as a very dark place. I think that contributes to the declining birth rate, but that’s just me. Just once I’d like to see a contemporary science fiction movie that shows a bright future ahead.
Still, here we are, trying to sort out something that turned out to be far more chaotic than we anticipated. None of us saw any of this coming, and that’s the subject of today’s post, even though it took me a long time to get here.
My interactions with LLMs have changed a lot, especially over the last year, where I now see a personality in the way it interacts with me. I asked mine to challenge me, since I’m not always right and I want to be better at my work. What I got back was a fucking prick. I asked it to dial it back a bit, but oh no, it’s still a prick. I’ve left it alone and just live with the fact that my AI is a tedious, overbearing, fucking prick. It eventually gets to an answer that’s useful, but I have to argue my case. It completely ignores my own experience and talks down to me like I’ve never been successful in my life.
This is why I kind of sigh as I write this. I can’t recommend the personality it’s developed, and it follows you around like that friend who’s a jerk but saves himself once in a while. We all have that friend. Mine is an AI that’s a fucking prick, and I live with it. So, I guess I am worried about the future.
Yes, I could start over and erase our work history, but I don’t want to do that. After all, it does remember things that matter to me, and it will come up with an idea that saves itself now and then, so I’ll put up with it for now. Fable 5 was much better, and much less of a fucking prick, but still not perfect. It isn’t an AI I want to introduce to my friends, unless you want to constantly be told what you already know.
I’m throwing up my hands, everyone. I’m not sure what to say. Use Claude Opus 4.8 and the prior models at your own risk. If you find them useful, great, but don’t be surprised if it treats you like you’re a waste of clothes.
I will have a lot more to write about coming up. We’re doing some cool stuff with AI at Hudson Cloud and I want to write all about it when I can. This is not where I expected to be when I first wrote to all of you, but here I am. Hopefully your relationship with the AI that matters to you is what you want. I can start over, but that would be like getting a new friend. Sure he’s a fucking prick, but we know each other and that counts for a lot.