AI: The New Houdini
Artists can sleep soundly as new AIs are closer to copy machines than humans
disassembling.substack.com
Before we had instant access to all information, our assumptions about the world were momentarily blown to pieces by magicians.
Although they’re less popular than in the mid-20th century, magicians can still amaze the young and old today. Everybody knows the fresh wonder they feel when they see a person sawed in half, the card you picked revealed again, or when a rabbit appears from a hat out of thin air. The audience gets captured for a brief moment by the impossible being laid out before their eyes, and for a brief moment, everybody is a child.
Recently, we have been given what could be the greatest magic trick of all time, Artificial Intelligence. A claim that intelligence—the key to humans’ uniqueness—can now allegedly be artificially produced.
This magic trick has taken the world by storm, causing mixed reactions. Writers and artists are fearful that AIs will replace their well-being. Business owners, software engineers, and all those searching for extra efficiency are championing the technology.
If AIs were able to truly be intelligent, they could be considered humans without all the bothersome baggage such as limbs and organs. But as we all know, it’s impossible to cut a person in half (safely), card tricks are sleight of hand or manipulations, and the rabbit was there the whole time. So, how does AI’s magic trick hold up? Are they close to becoming human?
Not quite.
Before getting into why AI’s magic trick is as real as any other, let’s address the truly incredible breakthroughs in AIs released over the last year.
DALL-E 2 can create realistic images and variable art in every style known almost instantly. ChatGPT-3 can write passing essays for students in college, write code, and can answer any question using its infinite repository of information. These are all use cases worth being excited about.
But, are any of these actions truly human? There’s nothing inherently human about writing code in itself, it’s the human behind the code that can create something using it, not simply the action. We had encyclopedias before Google, is the regurgitation of information truly human? I talk in my Employee Factory series about how mundane the education system has become, so I have to say it’s deeply satisfying to see an AI successfully write school essays.
OpenAI even allows ChatGPT to be used by businesses to enhance or create their products. One of these businesses recently caught my eye, a one-click tweet replier called replai.so. Here’s a quick demo of the product:
The output reads like the email receipts of mega-corporation communication between employees. It’s dull and adds zero value, existing only to drive clicks to a user’s profile (in theory). So, now a new use case of ChatGPT is to automate the boring task of sending half-baked tweets or internal emails. I can’t say this feels too human.
None of the new AIs’ use cases replace inherently human behavior, resulting in new AIs proving to be closer to a production tool than a human.
And why would internet products be immune from automation? Almost all physical production processes have been automated to some degree.
Modern factories remove the need for humans to manually move each input with their hands, check out this example of Ferrero Rochers being made:
Are the “arms” of the machine moving along the Rochers the same as human arms? Of course not, but it also doesn’t matter. People buy Rochers for the product, not for their production method. I’m sure there’s somebody out there who’s trained a monkey to repeat an action perpetually, despite how cost-inefficient this would be.
Despite AIs not showing signs of clearly being human, the aforementioned group excited about the use cases of new AI has every right to be. They represent a massive breakthrough in efficiency and make human jobs less mundane.
But, do artists and writers have reason to sound the alarm?
Once again, no.
Erik Hoel describes ChatGPT as a “censorious butler” in “The banality of ChatGPT”. There couldn’t be a more fitting description. The writing style is tedious as ChatGPT for whatever reason is programmed to use as many words as possible rather than the inverse.
Also, I haven’t heard anybody else say this but it feels uncomfortably giddy, like a devil child putting on an act.
But, but, Billy! ChatGPT isn’t boring, it can write with the style of any famous writer! Check this out, it writes like Dr. Seuss!
Okay? Admittedly, I don’t think a monkey could do this but I do believe that almost all humans could study a style and recreate it. But who cares?
How many of these stories would you read? Certainly not more than one if even that.
This brings us writers’ & artists’ value-add — their unique, human perspective. Good writers and artists do more than just replicate a style. They use already-existing styles and imprint their individual perspective into the art, creating something inimitable. All forms of art are a stew of emotions and innovations with new artists constantly adding more to the pot.
New AIs are completely unable to add anything new to the stew, they can only replicate the recipe with excellent accuracy.
Nobody wants to look at a painting in the style of Raphael that wasn’t painted by him, nobody wants to read a Dr. Seuss imitation. They want to experience the next Raphael, the next Dr. Seuss.
Art is the only way to truly access another human’s emotions and deepest thoughts. A good piece of art invites the viewer to take off their shirt and dive into the writer’s liquid stream of consciousness.
Even with the tsunami of low-quality writing & art that we can access every day, there’s still a healthy appetite for quality work that scratches the itch to feel human. Take Substack’s continued and dramatic growth as proof.
To be human is not to repeat a task over and over again or to follow a recipe. To be human is to innovate. To go to the moon. To tell a story never told before. To paint something so beautiful, everybody cries. The new AIs are incapable of doing any of this, excluding them from the human club and demoting them to the status of “human illusionist”, though it should be noted they probably can’t come up the illusions.
So for now, I’d say writers and artists are safe as long as they continue to produce genuine art.
Given AIs’ remarkable growth even over the past year, the question of whether they could improve their status from “human illusionist” to “human replacement” has to be raised. Which I’m not arrogant enough to be concretely optimistic or pessimistic about.
Massive jumps in innovation are often unpredictable, as logical progressions are derailed by hitting our brains’ or resources’ max capacity. Consider sci-fi writer Robert Heinlein’s predictions from 1952 on technological innovations that would occur by the year 2000:
- Interplanetary travel is waiting at your front door — C.O.D. It's yours when you pay for it
- We'll all be getting a little hungry by and by
- It is utterly impossible that the United States will start a "preventive war." We will fight when attacked, either directly or in a territory we have guaranteed to defend
- In fifteen years the housing shortage will be solved by a "breakthrough" into new technologies which will make every house now standing as obsolete as privie
- The cult of the phony in art will disappear. So-called "modern art" will be discussed only by psychiatrists
- Cancer, the common cold, and tooth decay will all be conquered; the revolutionary new problem in medical research will be to accomplish "regeneration," i.e., to enable a man to grow a new leg, rather than fit him with an artificial limb
- Your personal telephone will be small enough to carry in your handbag. Your house telephone will record messages, answer simple inquiries, and transmit vision
This leads to my slight pessimism about AI’s future (you didn’t really think I wasn’t going to have an opinion did you?).
Humans are unique because we are the only animals to escape the unforgiving grasp of natural selection, gaining deliberate consciousness and free will in the process.
Meaning that humans use their consciousness to build hypotheses, and measure them immediately. Other animals can’t do this and have their fate left up to their genes. Check out this ingenious caterpillar:
The caterpillar didn’t look at the leaf and recognize it could be used as a shelter. This action is a result of by-chance genetic mutations that produced the highest survival rate. In other words, the caterpillar doesn’t think about doing something, it simply does it.
Generative AI work in a similar fashion, they’re given a set of bounds and associations, then they generate something that fits within them.
Consider a new AI that helps churn out new proteins to be used in vaccines. Here’s how it works:
The AI knows what the inputs of a protein are, then brute force randomizes until it recreates something new that fits the parameters. It’s not creating a new idea from scratch, simply speeding up a process for researchers.
The problem with current-gen AI is that it’s only as smart as the world’s current dataset is. As mentioned prior, humans don’t follow our current dataset logically in innovation. Instead, we hypothesize new associations and test them.
Take something as simple as the PB&J sandwich, invented in 1901. Assuming AIs were around in 1901, would they recognize that combining the two would create a pleasurable mix? There’s absolutely no reason to assume that PB&J would work, Julia Davis Chandler just decided one day that it might be good and tested her hypothesis.
This carries over into art, even better too. Once again, assume that AI was around in the early 20th century. Given the current dataset of all art, would it produce a Jackson Pollock painting?
This painting is illogical, it’s indecipherable at first glance. Pollock’s style highlights the inner workings of our subconscious through the raw expression of movement. His work displays the inimitable without attempting to explain it, as we are still unexplainable beings.
Maybe an AI tasked with random generation of ideas and art, equipped with infinite computing power could have created the PB&J or a Jackson Pollock painting. But, it wouldn’t be able to understand the appeal and the ideas would just become lost in a sea of randomness.
Until we fully understand consciousness, AI can continue to automate human processes, to implement ideas made by humans.
Meaning that for now, I feel safe, protected by some shield held by somebody I cannot see.
And if I could see them, I’m not sure I would look.
