What Do We Want Computers to Do?

June 11, 2024 • Louie Mantia

It seems like only yesterday Apple was crushed under the weight of criticism for its tasteless ad destroying creative instruments into a thin iPad. I thought that may have given Apple some insight into how artists use their products to supplement the creative process rather than replace it.

But here we are, watching Apple unashamedly introduced “AI” features at WWDC that write bedtime stories for you and generate images for your school notes on architecture. They’re handing off requests from Siri to OpenAI when they want ChatGPT’s “expertise,” a word they actually used in the presentation and twice in their press release.

I have reserved the largest of sighs for this moment.

Siiiiigh.


My friend Pat remarked how outsourcing the creation of a bedtime story for your child is just bad parenting. While I don’t have kids myself, I absolutely agree. It’s not just your job to be creative, inventive, and silly with your child. It’s a chance to do a little improv and giggle about your cleverness together. A bedtime story is a moment for you to bond with each other. If you look to “AI” to do this for you, what you get is simply not human. It’s definitely not you. It may sound crass, but you’ve hired a robot nanny because you can’t be bothered. You’ve admitted—very casually—that actually, you just want your kid to shut the fuck up and go to sleep.

And I think that really sucks.

You may have asked or heard others asking how this differs from other computing tools that help us accomplish tasks. Let me break down what computers can do for us.


Assistance

Assistance helps humans do something. I’d consider most functionality on computers to be qualified as “assistance.” This concept is best described metaphorically. Computers are the “bicycle for the mind.” You still power the bicycle, but the bicycle is more efficient. Video chat enables us to talk at great distance, but we still have to do the talking. When writing, spell check lends us a helping hand.

Automation

Automation reduces humans’ time spent on mundane tasks, like renaming multiple files or batch-editing photos. To qualify as automation, you have to construct, design, or perform a task at least once so you can apply those rules or instructions to something else. Siri Shortcuts, Automator, and Photoshop Actions all are examples of automation. When you copy and paste edits in Apple’s Photos app, it’s operating an assembly line you designed.

Generation

Generation fabricates with minimal human involvement. Predictive text may fall into this bucket, but at its present extreme, generation is used to forge (in both senses of the word, forming and counterfeiting) images and sentences from extensive data. Humans have limited influence or control over the output, and these tasks are generally centered around replacing human creativity. You’re not writing, drawing, or even performing an action to be replayed.

Computers can do all of these things, but what do we want computers to do?

What do we want to do?


Birds instinctively build nests. Beavers instinctively build dams. Humans instinctively create too. We make art. We can write books and music. We read those books and listen to that music. We can draw and paint. And we observe the result in sketchbooks, galleries, and museums. We cook food that is only possible because we moved and adapted ingredients from around the planet. And we taste things that are entirely new. We can dance. We can act, and even capture ourselves acting, and then play it back for others to see. We spend our lives creating and absorbing all of these things, and by interpreting them, it shapes who we are. The act of artistic creation is our fundamental unique characteristic. Life imitates art. Art imitates life.

If I never needed to make money anymore, I would see and hear and taste as many things as I could and just draw things forever. The idea of relinquishing that power and capability feels like robbing me of what makes me feel human at all.

Humans making tools to facilitate creation is wonderful. It’s an act of creating to assist in creating. A digital thesaurus is a wonderful, powerful tool. A digital color palette that makes every color instantly accessible is incredible. These things make creating a joy without taking away from my role as a creator.

One of the most meaningful things I’ve found in my life is going from imagination to realization. Taking a spark of an idea and figuring out how to make it. To paraphrase David Bowie, if I can manifest what I feel inside myself, I can understand more about me and the world around me.


I’m not against playing with these generative AI tools. For months, I played with Midjourney, and I still occasionally play with ChatGPT. I say “play” because they’re excellent toys. However, they are also time-wasters, and I can feel my soul drift away as I toil with specific words to see if I could coax them into realizing what I imagine.

And it can’t do it. It will never do it. No matter how good the models get, these tools will never produce what I see. Because what I see is the exclusive culmination of every experience in my life. Everything I’ve ever seen or heard, everything I’ve ever felt. These experiences are directly responsible for what I create and how I create. I can’t supplant that, and I don’t want to, even if I could.

It’s just a toy, a tool to indulge in manufactured daydreams. Every time I use one of these, it feels both comforting and frustrating at the same time. The recipe for addiction.

In contrast to video games—which are often dismissed as time-wasters but are frequently great simulated learning environments—we don’t learn anything from Midjourney because it doesn’t attempt to teach us. And ChatGPT doesn’t know fact from fiction, so it’s incapable of (reliably) teaching.

Tons of moral panic around how you can’t trust Wikipedia because any human can write it, but comparatively less of that for ChatGPT, despite humans writing none of it.


Every artist I know struggles to realize what they imagine. It is through that struggle that we create. If you remove the struggle and simply wish a paragraph or image into existence, you have decided to outsource interpretation. The concept of interpretation is unique for each of us. It comes from our own selective, critical understanding of the world and how we fit into it. When you use these generative AI products, you condemn your own ability to interpret.

That may be a hint as to why these tools are categorically shunned by artists, but popular amongst non-artists. It gives those who cannot create the illusion of creating, while replicating absolutely nothing from the creative process. There’s nothing authentic about it.

Wish-fulfillment products like Midjourney, ChatGPT, or even Apple’s implementations of these kinds of technologies only exist to permit people to pretend. Let’s not forget the A in AI is “artificial.” It’s not an animation without an animator. It’s not a painting without a painter. It’s not a novel without a novelist. Nor is it music without a musician.

Put another way, artificial intelligence is just faking it. And people who use it are comfortable with faking it.

Will you actually send an email you didn’t write? What will that say about how you value the person you send it to? Will you really use a generated image in your school presentation on architecture? How will you defend that to your teacher? To your boss? To your client? To yourself?

Why pretend to do these things? To what end, and for what purpose? It’s not saving us time, and it’s not making us better at our jobs. Looking at it favorably, it’s an illusion. Realistically though, it’s just a lie.


I’d like our computers to work toward increased efficiency, considering not just our human effort but the energy cost. It should feel like computers are helping us. Replacing or destroying any part of us does not help us.

There are tons of tasks in my life I wish I had computational assistance or automation for, mundane tasks that I’m sure no human actually wants to do. We’ve not yet reached the end of that list. So, let me ask again.

What do we want to do? And how can we get computers to do the stuff we don’t?

If you ❤ like this