Part One: AI in 2022
2022 has been an interesting year for artificial intelligence (AI). Whereas AI, in the popular discourse, has heretofore been laser-focused on defeating Russians in chess, something has now changed. Perhaps the pressure of the chess circuit got to being too much, and AI snapped. Because, for whatever reason, AI is now preoccupied with generating all-original, community college art projects.
For example, there have been a number of AI image generators that have made the rounds this year. If you’re on Twitter, you’re probably tired of seeing these bizarre images. A leading culprit has been DALL-E 2, which is a research project developed by a company called OpenAI. The name is a mashup of WALL-E the Pixar robot, and Salvador Dali, the famous mustache model. It’s a program where you type in something you’d like a picture of - literally, anything you can imagine - and it spits out an original image, which is then framed and hung in an independently owned coffee shop, with a delusional price tag of $250.
For example, I wanted to see a bear in a coffee shop, wearing a neck tie and working on a Chromebook, and here are the totally original images that DALL-E 2 came up with:
In the third image, you can actually see terrible art for sale on the wall of the coffee shop, over the disturbing bear’s shoulder. Very meta!
These are pretty terrible pictures, but in fairness, something about the “grizzly bear” prompt I used seems to have led DALL-E 2 down a particularly direct path to nightmare fuel. Other prompts can do better, and be more entertaining. And as goofy as these images are, one gets the impression this is version 1 (well, literally version 2 of DALL-E) of something that’s going to evolve a lot, and quickly. This is the Ford Model T, and one wonders what the Tesla Model S version will be in 100 years.
Part Two: AI Last Week
If 2022 was the year of AI, last week was the peak. The same company (OpenAI) that produced DALL-E has rolled out ChatGPT. This is a program that can have written conversations, and beyond that, write in specific styles, or genres. It can write poems, or movie scripts, programming code, and even “Cool Story, George” Substack articles. On Twitter, speculation abounds as to the implications, and everyone is once again fascinated with AI.
Speculation abounds as to the potential of this technology. There are hopes and fears around what all the implications are. There are moral, ethical, and legal questions about copyrights and the source material. There are other concerns that it is already being secretly, intentionally, manipulated to be politically/morally biased on certain (predictable) issues. And there are concerns on a smaller scale; could it be a game-changer for writing school assignments like term papers and book reports? Imagine a world in which college students can just feed a question into an AI, and get a book report, thus entirely eliminating the need to pay another student to do so.
Whatever the potential, Twitter has been obsessed with ChatGPT. Twitter’s obsessions are not a totally accurate meter of societal relevance. There have been MULTIPLE weeks in the past year when Twitter was dominated by something called “The Island Boys,” so I’m hesitant to give too much weight to a thing, just based on Twitter’s fascination and attention.
Since ChatGPT can write, I think it’s particularly caught the attention of writers. First of all, it has tossed a nice, new, existential crisis into the mind-space of writers who already overthink everything, and especially so during the emotionally intense season of nostalgia-driven introspection, which is Christmas. Secondly, if you’re like me, and have a self-imposed commitment to publish something written on a topic every week, then ChatGPT has come at the perfect moment, when year-end work deadlines merge with Christmas celebrations, and a two-month-old baby. All of that makes a sophisticated chatbot a potential life-hack that is very enticing.
Part Three: What’s It Like to Use ChatGPT?
I’ve done a personal investigation into ChatGPT. As I’ve mentioned before, I learn by doing, and so in order to understand this new tool, I went ahead and opened an account with OpenAI, thus registering myself with what is thus far evolving to be real-world Skynet. When the robot apocalypse comes, I’ll be on their short-list for sure now, so you all better appreciate this article!
That said, ChatGPT, at first glance, was to me kind of like a Google search, but multiple Google searches, woven into complete sentences. With the added kicker that it can write in the style of your choice.
Like many, my first concern was that ChatGPT is already so advanced that it will replace my popular Substack, “Cool Story, George,” and then come after my day job, my family, my dog, and ultimately, purchase my 2005 Honda Civic Special Edition coupe.
To place my mind at ease, I tried to make it recreate my masterpiece series, “What is a Chromebook.”
First I asked it, “What is a Chromebook?”
That’s fine and well, but it clearly lacks my humor, and mentions nothing about minivans. For the first time this week, I took a breath of relief.
Next, I pushed it for some hot takes. “ChatGPT,” I asked, “Which Chromebook should I buy?”
Pathetic. How can ChatGPT call itself AI and not even know about the HP Elite Dragonfly? The dumb thing hasn’t even heard of the Acer Spin 714, and the incredible sales going on at Best Buy during the holidays! Ha!
However, ChatGPT is a tool, and there can be some operator error. To really get results, one needs to feed it questions phrased appropriately to lead to good results. So I offered ChatGPT a little creative nudge, and a chance to shine.
I went with Shakespeare, and the results are impressive:
Now we’re talking! I especially like “They do not suffer from the weight of legacy.” Oh, but were my own writing to shine forth with such deft eloquence!
Finally, I thought wayyy outside the box, and went for total, Chromebook-centric, originality:
And here, ChatGPT has truly embarrassed itself, once again. The guy is not named “Tom.”
Part Four: My Conclusions
In conclusion, ChatGPT is two things, at the same time. One, it is an entertaining party trick/gimmick/toy, capable of wasting a lot of people's time for a week on Twitter, before we all move onto the next thing. Two, I think, it is also impressive and has a strong chance of forming the future in a significant way. Artificial Intelligence can already defeat the best of the best in the human world at chess and no-limit Texas Hold'em. Now, it can mimic us rather well, in the words it chooses to use. As silly as it sounds, we're one big step closer to having the Terminator, or the Battlestar Gallactica Cylon race. For some reason, GPT stands for "Generative Pre-trained Transformer", calling to mind images of enormous, talking, armed, robots. And/or Shia LaBeouf, I guess.
Science-fiction nightmares aside, this is a huge breakthrough; there are so many angles and opinions and hypotheticals to explore. I could, no-kidding, start another weekly Substack, focusing on AI exclusively; however, I won't. Because, what's the point in trying? An AI probably already has launched that Substack, and the AI can do it better, and faster, and in the style of Shakespeare.