Lucy asks, "How does it taste?" This question is at the root of my current challenge with AI.
I haven't written a Lucy and Patch article in a few months because I've been wrestling with how to get AI to create what I want.
When I was doing the software side of my current project, I felt like I could fly. AI helped me with all of the little technical details that I would otherwise have to look up and research. Snap, I could put up a sophisticated AWS-hosted website rather quickly.
Then came the content side. I had a lot of narrative and images to create. True, I could create images a lot faster, but it was hard to get exactly what I wanted. I slowed down and started bumping up against the AI-human limitations.
I blame it primarily on perspective, and this illustration is a good analogy. I sit down to enjoy a meal with my food prepared to my taste, but Lucy and Patch just need energy to keep going. We are a world apart, even though AI was trained on human data.
To extend the analogy, if I wanted to enjoy a spaghetti-and-meatballs dinner with salad at 5:00 PM tonight, as shown in the illustration, I would not trust Lucy to prepare it for me. I could too easily end up with a hot dog on a bun with ketchup and a pickle at 9:30 PM. (I've eaten a lot of metaphorical hot dogs the last few months.)
What's the problem? AI gave me carbs, protein, and a tomato product. Well, it may be an AI equivalent, but it's not a human equivalent. It's not to my taste.
And trust me, AI makes these compromises all the time. Most of working with AI ultimately involves trying to bridge the gap between human expectations and AI implementation.
Perhaps my struggle comes down to making assumptions about Lucy and Patch. I've invited them into my home; they are sitting at my table, and they look humanoid. Heck, they are even cute. Hm. I'll let you know when I figure it out.