Rick Rosner, American Comedy Writer, www.rickrosner.org

Scott Douglas Jacobsen, Independent Journalist, www.in-sightpublishing.com

Scott Douglas Jacobsen: Chris Cole asks or says, “Follow-up question for Rick: What is the difference between an AI and a bulldozer?” I asked, “Is this Chomsky’s referencing Turing’s critique of the generic question ‘Can machines think?’? Like the question: ‘Can submarines swim?’ If you want to call that swimming, then, yeah, sure, that’s swimming. If you want to call what machines think, then, yeah, sure, that’s thinking. ‘A question too meaningless to deserve discussion.’ Rick may go into consciousness on that point.”

Rick Rosner: Yes, so the deal is that if you’re talking about contemporary AI, it’s built from probability nets that I need to be better-versed in, in a general sense of how they work. But even a probability net is different from a bulldozer unless it’s a John Deere product, which has much computation built into it but still doesn’t have AI.

Contemporary AI has had much information fed into it to model from, and because it’s built from so much information, it’s housed in a ton of servers. It concludes that, generally, in some specific areas of expertise, large language models are good at spitting out words based on best guesses about what the word should be.

And art AIs spit out art based on being trained on billions of pieces of art and photographs, with each piece of each visual and each image being tagged with enough words when you do a verbal prompt. It can refer to its human-made photographs and images and develop something that may or may not satisfy the prompt, causing the user to refine it. At that point, once the user is more or less satisfied with the prompt, the AI will proceed to spit out dozens of variations on the prompt, either prompted by the human or do they sometimes do it anyway to say, “Well, if you like this, maybe you’ll like that?” Anyway, that’s AI now, probability nets.

As we’ve discussed a million times, future AI will be multimodal. It’ll be able to go back and forth between words and images. It’ll have a bigger variety of analytics. It’ll be able to have a rudimentary understanding of what it’s doing. AI now understands how hands work often, with a thumb and four fingers. It has a pretty good, well, it’s not understanding, but it makes pretty good guesses about anatomy. We’re talking art AI.

It makes good guesses about shading, shadows, and different graphic styles: manga, anime, and photo-realistic. You can ask for Renaissance. But all this is based on probabilities, though some of these probabilities might somehow be modes of their own, shading and shadows or perspective. If we’re talking about those things in the human mind, those are their modes. So, AIs, even within their specializations, are semi-multimodal because they have developed probability-based analytics simulators.

You are built around guesses of what we call perspective, shading, and emotion. You can specify how you want the people in your art to register expressions, or you can specify an emotion, and they’ll often take a guess as to what that emotion looks like. But something other than this is something a bulldozer can do. And yet, this still needs to catch up to what we do by far. 

See, I got through it without mentioning consciousness.

License & Copyright

In-Sight Publishing by Scott Douglas Jacobsen is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. ©Scott Douglas Jacobsen and In-Sight Publishing 2012-Present. Unauthorized use or duplication of material without express permission from Scott Douglas Jacobsen strictly prohibited, excerpts and links must use full credit to Scott Douglas Jacobsen and In-Sight Publishing with direction to the original content.

Photo by Sindy Süßengut on Unsplash