My Thoughts on AI

I admit I am still on the fence about where I stand on AI (or any proximity to AI). I also will admit I have not used the AI generators much. However, I do have my opinions. Agree or disagree, this blog is just a compilation of my jumbled thoughts on Artificial Intelligence.

My earliest experience with AI is the Eliza programs on the old PC or Apple machines. I have always been fascinated with how such programs work. I am very curious to know how modern AI tools learn to react. I am curious how machine learning differs from human learning. Could software be taught to learn more like how humans learn? What are the next steps of AI evolution? Do developers create algorithms to be more flawed like humans, where they forget things or mess up, and then adjust their learning based on these mistakes?

When I took the developer bootcamp, our teacher encouraged us to learn how to use the modern AI tools. A good friend has also encouraged me to learn to use AI. The teacher made two points about using AI for the class. First, do not use AI to complete homework assignments or tests. And second, when using AI, you still need to understand the answers the AI provides. I think that is where my hesitation to start using AI lies. If the tool spouts an answer, will it also provide an explanation of the answer? And, will that explanation be simple enough where a child could understand? Or, will it just regurgitate a bunch of jargon and methods that will also need further explanation? I learn better by doing, which is why I wrote my websites from the ground up. I want to know how things work and make them work the way I envision them working. Could an AI tool act more like a teacher by explaining something and then challenging me to solve a problem in my own way, and then provide feedback for my solution? If so, I feel I would get more out of AI.

I have seen some of the results of AI, in which people ask for some random picture and the outcome is based on how it interprets their ask. Some of the results are ridiculous, because people will have nine fingers on one hand. I'm not knocking anyone with more or less than four fingers and one thumb per hand, but that is not what I would consider "intelligent". The pictures seem like mad collaging skills with glorified Photoshopping techniques. It may be able to imitate some of the techniques, but can it produce an original image by imitating how it interprets another image? Think of a kid told to draw a house. Most drawings will have similar starting points when learning to draw. Most often, these houses start with a rectangle house with a triangle roof. Next, they learn to add more details like doors and windows. How would you react if you asked a kid to draw a house for the first time, and they gave you architectural blueprints?

As for writing, some of the earlier attempts at AI writing scripts is a bunch of ridiculous nonsense. It is quite a step up from automated MadLibs, but is far from an number of computer monkeys producing a lost work of Shakespeare. The AI "learns" from pattern matching, but how well does it understand actual sentence structure? Ai writing books reminds me of the controversy of Dan Brown's formulaic books (Is that guys still around?).

On the one hand, I understand how AI could possibly be a useful tool. On the other hand, I see AI as something to approach with caution. It does not exactly seem "intelligent", but more like a mimc. As in the realms of Dungeons and Dragons, mimics should also be approached with caution.

One of my concerns is the source material used for its machine learning. The AI is not really learning how to create new art. Like I said, it is a collage, blending together bits with various filters. If someone could disassociate preexisting images, could it create something resembling an apple purely based on a description?

What about text responses? As humans, some portion of us know not to believe everything we find on the internet. How can the AI distinguish between what is real and what isn't? What is based on fact and what is fiction? What is truth and what is a lie? If an AI response is regurgitating an answer based on anything it might find on the internet, how accurate is the response? Garbage in, garbage out. Or, is it "smart" enough to find or calculate proofs to prove the legitimacy of its results?

The internet is already a confusing landscape. We have difficulties determining what is real and what is made up. AI has traveled into the uncanny valley of being able to produce a piece of writing to where people are willing to accept it as real or the truth. That is a scary thought. What happens when these false texts become printed books? The physicality of that false text becomes more real, because it is tangible. Before AI, books were vetted. The content written, edited, and maintained by humans. What happens when most of the human element is removed from that process? The human checks and balances fall apart. You are left with human readers consuming thoughts provided by computer algorithms based on learning from who knows what on the internet. How scary is that?

Before I share how I envision using AI as a useful, let's consider the moral argument when it comes to AI. To do this, I am not going to think of the scenarios as a human versus a machine. These scenarios will consider me versus various imaginary consumers.

I am an artist. I create art, both written and drawn. In the case of my board games, I produce art as an interactive concept. When it comes to creating art, I usually do not have any particular person in mind to consume it. I suppose I mostly make art for myself, and hope that others enjoy my silly, weird creations. But now, AI has got me thinking about how comfortable I am with the ways others consume my art. How do I feel when I consume other people's art?

My ideal scenario is that a person out there reads one of my books and enjoys it. Maybe, they like it so much, they want to share it with a friend. One sale appreciated by more than one person. Ok. I'm good with that. I equate this to renting a movie to show to the family.

What about libraries? A library purchases one copy, many people have the option to read the book, but only one at a time. I am still good with this, too, because a purchase was made and potentially introduces my work to many people, who may explore my other works. In both scenarios, it starts with a one-to-one ratio. One book. One reader. Both scenarios based on one sale. Both scenarios could result in more than one person discovering my work based on that original, single sale. Potentially, this is a good thing.

Am I okay if a librarian, a teacher, or someone else reads my books to a small audience? In a way, I equate this to a student doing a book report on one of my books, and then telling the teacher and their classmates about the book. This is the scenario above with a review of the book, which can be a good thing. Where I draw the line is someone charging the audience with a presentation of one of my works, in any form. When that happens, someone is making money without permission off my work. In a way, it's like making money writing and selling fan-fiction.

I am cool with fan-fiction. I write fan-fiction because I enjoy playing in other people's creative playgrounds. I think up new scenarios the original author may not have considered. When I share my fan-fiction, I am not looking to make money from it, because I have not asked permission to make money from playing in someone else's world. However, I do write fan-fiction as a carrot to lure people towards my own stories, which I do sell. I am inspired by X. If you also like X, then you might like Y. In a way, this is similar to the library scenario above. I am fine with others creating fan-fiction based on my own works, not to earn money from the fan-fiction, but I am totally okay if they write fan-fiction based on my work to create a trail of breadcrumbs back to their own work. I would be a hypocrite if I thought otherwise.

What about book clubs? Multiple people reading one of my books at a time. It would be fantastic if each bought a copy of my book, but I am okay if a few members seek out a library or a friend to borrow the book. In that case, at some point, each copy of the book being read has been purchased. Even though the club is a closed loop of people discussing the book, there is the chance some members of the group may not have discovered or learned to appreciate my work, so there is potential to generate that appreciation. There is also a chance people will talk about the book club and mention one of their recent reads. Again, this has the potential of advertising. Another form of breadcrumbs.

When someone is done reading my book, a physical copy of the book, am I okay with what happens next? Sure! Pass it on to the next person. Want to sell it for whatever you can get? Sure! I've already made my sale. And, most likely, that person is not earning more off that one book sale than I did in my original sale.

The areas I have the biggest problems are where other people use my work and call it their own. If I have put in a massive effort to create artwork, others should neither claim credit nor payment. If I create a game like Monopoly in almost every way, except for the place names and called it "Property Wars", I would (and should) get into big legal trouble for infringing on another artist's work. Now, given that the original creation for Monopoly was many years ago, and might be in the public domain, I still shouldn't slap new labels on an old idea. I should be inspired by the original, but tweak it to make it something I can call my own. If I am releasing Monopoly and calling it my own creation, that's just wrong. Call it stealing, plagiarism, or copyright infringement. Whatever you call it, it is illegal. That's where AI does not necessarily know the legal rules, nor knows how to abide by the rules. AI may not know the laws, but the developers and the company designing the tool are fully aware of the laws and should be held accountable for improper use of the original content.

To back up a little bit, when I consider AI siting other sources, I think of some of my friends who are more ferocious book worms than I could ever be. I have pondered what the difference is between AI consuming books and my avid book worm friends. The biggest differences are (1) those friends spent more time and money consuming books and got more out of it than a computer algorithm, and (2) those friends are human. Most humans could not regurgitate an entire work from memory, but a computer with enough disk space could recite works word-for-word, because it is just reading a copy of the work from storage. To me, that is a huge difference between people consuming, processing, and using another's written work and a computer.

Borrowing without consequence sums up my biggest beef with AI. When I wrote papers for school, I was told to site references. From what I understand, Ai does not site references. There is no paper trail. It pulls something from a big, ball of mishmash without specifying the sources. What trail leads back to the origin to let the inquiring person know if (A) any part of the response was copied without giving credit to the originator, and (B) how credible or factual is the originating source? What someone does with that information without traceability can, and has, led to problems.

Saying all this does not mean I am totally against AI. Now that a machine can gather and process information from a huge source (the entire internet?) to be able to give a somewhat coherent answer, it would be interesting to know if a machine can do the same with filtered and focused information. For example, could NASA and the other space organizations, provide data from their histories of science discoveries and theories, and have the machine be able to help us better understand the universe? Could a pool of medical data and case studies provide a more realistic diagnosis for patients? This type of scientific usage should be backed up by human expertise.

If AI ever reaches that level of targeted machine learning, I imagine I would be more apt to use AI as a personal tool. Imagine being a proficient author with a huge library of creations, like Terry Pratchett with his dozens of Discworld books, and having an AI that could help keep track of your bible of characters and plot histories. It would be wonderful to be able to have an AI rebuild and summarize a character's history in the story world. Imagine what Marvel could do if it did the same thing with its entire comic and movie history.

My concerns with AI are less with AI becoming sentient and killing off the human race, and more with the ways humans abuse AI to exploit other humans. Do you choose to use a sharpened pencil to stab someone? Or, do you use it to draw a picture, write a poem, or work through an equation? People have a tendency to find new ways to exploit a good thing. My hope is that AI becomes more of a tool to do good.

Related Content:
 Conversations with Characters (blog)
 Game of Chicken, A (blog)


Click the Copy button to share this page's content:   COPY
A collection of several random thoughts on AI and its usage. #introspection - https://www.checkeredscissors.com?cat=blog-180