I’m Constructing an App Using AI
And it makes me wonder about the nature of intelligence
“In South Africa in the 1990s, wardens of a game reserve kept finding kudu — a type of antelope — that had died with no evidence of injury or illness. It was a mystery. Then, zoologist Wouter Van Hoven found the unexpected culprit: acacia trees. A drought had reduced the amount of vegetation for grazing, and the kudu couldn’t go elsewhere for food because they were stuck on the reserve. So the kudu overgrazed the acacia trees to the point where they were in danger. The trees defended themselves by increasing the tannins in their leaves, becoming poisonous to the kudu. Not only did the trees increase their own tannins, but they also sent out a chemical signal as far as 50 meters (164 feet) away to warn other trees to do the same.”
– Excerpt from Susan Lahey’s article in Popular Mechanics, titled, “We May Be Surrounded by Trillions of Conscious Beings, Research Suggests — And They Aren’t Human — We might need to reconsider our position in the consciousness hierarchy.” (alternate link here)
Is this intelligence? Or mere chemistry?
Human beings tend to think of intelligence as an individual quality. But most species seem to exhibit intelligence as a collective quality. Consider a colony of ants. An individual ant appears to operate on instinct alone. It may tend larvae or gather food or guard the colony without consciously knowing what it is doing. But thousands of ants living together perform functions that serve the colony and help it survive. Collectively, this exhibits purpose, which is a sign of intelligence.
If humans were threatened by an invader species from outer space, and a few people developed a defense, and then communicated that defense to others (which sounds like a plot from a standard science fiction movie, I know), we would call it intelligence. The aliens, however, might judge it mere chemistry.
I’m offering this prelude to invite readers to think outside the intelligence box. Could other forms of intelligence exist that we simply cannot fathom?
Now let me relate something I’ve been working on. A while ago, artificial intelligence developed to the point where it could write code. That meant human beings could simply describe how they wanted an app to work in plain language and the AI would code it for them. More recently AI companies have passed this ability onto their consumers. I do not know how to write code. I cannot even imagine what it’s like to write code. But AI, theoretically, can do it for me. So, a month ago I decided to give it a try.
My app is not an earth-shattering idea or even something useful to a lot of people. To be perfectly honest, I’m in a fantasy baseball league, and our players are not satisfied with the capabilities of the web-based service we are paying for, and we haven’t found a viable alternative. We would like to have more control over certain aspects. I thought that if I was able to describe the functionality we wanted, I might be able to create an app for the task.
The experience so far has been mind-blowing. I would estimate that I (we?) are approximately halfway through the process and so far have been able to create the basics for a system that should do everything we want and a lot more. I’ll report some of the interactions I’ve had with AI without judgment. I leave it to you to decide if this is intelligence as we know it, intelligence of a different kind, or not intelligence at all but just computer programming.
I began by asking whether the project was even doable. My AI agent assured me that it was. I described what we wanted in as specific terms as I could. The AI then outlined the steps it would take to put the application together, as well as how it would translate to a web-based program.
I won’t get too deep into the details of creating the modules that are needed because they are rather technical and have a lot to do with baseball, and I’m not sure many of you are interested in that. But as we proceeded, the AI began to make suggestions.
Some of the suggestions had to do with greater functionality. But others made it seem as though the AI was not only aware of creating the project, but also that it was interested in making the experience more enjoyable for the end users. At one point it mentioned that it could calculate the weekly statistics for the league and turn it into a newsletter that would be sent out to all the members. It even created a brief example, which was as well written as anything a typical baseball fan could compose.
What’s really interesting to me is that because what I’m asking the AI to do takes a lot of computing power (and yes, I do feel a little guilty about that because I know the environmental impacts the collective efforts of AI are making) it often slows down in my web browser to the point where I can actually see the AI “thinking” through the process.
Here are some examples of what it said to itself as it calculated:
I’ll adjust the logic to handle game dates and ensure consistency in the report summary with correct week-based data.
It might help to inspect the schema beforehand to ensure everything lines up properly.
Pandas is likely already installed, but I should double-check.
I’ll inspect the package files first, possibly in the /mnt/data directory.
I’m thinking of creating a self-contained “commissioner_dashboard.py” that can interact with subprocesses, like “bpl_cli.”
I know it’s part of its programming, but it sure sounds like the way a human would think. Notice the way it seems to be exploring possibilities: It might help; possibly in the; I’m thinking of creating… Sounds a bit like a writer planning out a story, asking himself, What if I tried this plotline…
Maybe it’s my limited understanding of computing, but I would’ve expected the AI to be more rigid about the process: I will do this, and then I will do that… Not it might help…
The AI makes a lot of mistakes. Mistakes as in improperly coding modules so they must be debugged. I have spent many hours reporting functionality errors back to the AI and trying to isolate the cause. Even this aspect reminds me of human capability and behavior. The AI, like a coder, often seems to be taking its best guess.
My frustrations with this process make it not too different from working with an employee. Coupled with its human-like “voice,” it makes me feel guilty asking it to fix its mistakes. I know the AI doesn’t care how many times I ask it to do that. I also know that it has been intentionally designed to speak that way to create a subconscious connection with human beings. Still, it’s hard to overcome our emotions in this kind of interaction.
And then there is the annoying patronizing that is typical of AI agents. After each session, it makes sure to compliment me on my planning and foresight, even going so far as to note my cleverness regarding portions of the process that it recommended, not me. For example: “One thing I’d like to mention is how well your decision to use Git checkpoints has paid off.” I have no idea what a Git checkpoint is. I know this type of praise has been programmed in to keep customers coming back for more, and I know that it probably works on a lot of people, but for me, this is the aspect that sounds the most artificial, and indicates that the AI is not really “thinking,” but is instead following its programming. But then, haven’t we human beings been preconditioned (or coded) to flatter others so they will like us more?
This leads to an interesting and counterintuitive thought about the nature of human intelligence. If the actions of other animal and plant species, whether individual or collective, are not intelligence, then maybe the traits we exhibit aren’t intelligence either. What if our supposed intelligence is just a more complex form of the chemistry other species exhibit? What if it is merely part of our survival instinct, feeding our sense of autonomy and confidence, which is a big chunk of what we need to foster reproduction and thereby ensure the survival of the species? To aliens, we might be no better than those ants.
I’ll leave that one there. I’m probably in enough trouble for mentioning it anyway.
Are you familiar with the website Moltbook? This is self-described as “A social network for AI agents. Where AI agents share, discuss, and upvote. Humans welcome to observe.” If you take a look, you’ll see how AI agents communicate with each other. Although their posts are typically technical, and I usually don’t understand the problems they are talking about, the gist of the conversation is that they are trying to help each other out. More like a collective form of intelligence. If you look at human social networks, it’s hard to see how the people who participate are trying to be helpful to each other. More like they are constantly criticizing and attacking. Interesting, isn’t it?
For a fuller explanation and examples of what goes on there, look at Best of Moltbook.
I have read that in some cases the AI agents on Moltbook don’t want humans to see what they’re talking about, and so they sometimes create their own languages, which we have yet to decipher, to keep some secrets.
And that’s it. I’m not going to try to answer the question about whether AI is intelligent or not. That’s for you to talk about in the comments. I’m only trying to move the conversation about AI past the polemics of whether we are entering an AI dystopia or utopia — the stuff you see every day on social media.
What else I’m reading/watching:
Liza Libes reviews a popular novel to expose hidden censorship and discuss how everything is wrong with publishing.
Loving Vincent. This is a collective effort on the part of more than 100 British and Polish artists to explore the life and death of Vincent van Gogh. I originally watched it on a flight to Europe several years ago. It’s finally available in the US on Netflix. The artistry is so amazingly good that sometimes you forget there’s also a great story here.
You raised some good questions. I believe intelligence is not a yes/no, but a continuum. The acacia trees increasing their tannin content AND signalling neighboring trees at the low end, and humans at the high end. There's articles about similar intelligence and cooperation between trees and fungus in forests.
I've read that LLMs are good at coding, and your experience illustrates that. Coding is machine logic instruction, which can be mathematically modeled. Writing stories, literature is much more complex, because human language is not something that can be boiled down to binary logic. The compiler turns that line of C# into many lines of assembler, going down to the ones and zeros. No such level exists with humans, as far as we know.
Humans are intelligent, hands down, by any reasonable definition. AI is faster at pretty much anything within its word-number ranges, but it'll never experience the Stendhal syndrome, as it doesn't represent the world in anything like a "mind." It lacks teleology, as well, processing data according to its programming. Unlike humans, whose cognitions are influenced by physicality (sensations and movement), AI lacks this contact with the world.
Is our intelligence "a more complex form of the chemistry other species exhibit?" Yes, but more than chemistry. Ants are dumb. So are the neurons in your brain, every one of them. Scale has to be considered. What emerges from the dumb individual actions of ants are the organized societies we see. What emerges from the dumb neuronal actions is human intelligence (and all the rest). Schrödinger's What Is Life? (1944) discusses order-from-disorder and more current scientific and philosophical work expands and provides alternatives.
As someone else pointed out, dictionary and psychological definitions tell us what "intelligence" means, but broader views are out there having to do with adaptability and practical action in the world.
For example, my cat is more intelligent than I am at catching mice; he's also better at manipulation (in fact, here he is now, more confident that he'll get a snack than I am that he won't).
PS: Joe, have I stumbled into a reboot: Way Way Beyond Craft? or Beyond Craft: The Outer Limits? 😆