One of the most insightful commentators on AI is the science fiction author Ted Chiang. (Somehow I have yet to read Chiang! If anyone has recommendations for where to start, let me know.)

Chang’s recent article, Why A.I. Isn’t Going to Make Art, is packed with razor-sharp observations, like this…

Using ChatGPT to complete assignments is like bringing a forklift into the weight room; you will never improve your cognitive fitness that way.

And especially this…

What you create doesn’t have to be utterly unlike every prior piece of art in human history to be valuable; the fact that you’re the one who is saying it, the fact that it derives from your unique life experience and arrives at a particular moment in the life of whoever is seeing your work, is what makes it new.

The entire piece is a great read and I largely agree with Chiang. But we diverge at an important fork in the road and this leads us to entirely different conclusions.

Chiang argues that AI won’t make art. Agreed. But its true potential lies elsewhere, in helping us manage the more mundane aspects of work.

A better definition of intelligence

Here’s how Chiang views intelligence.

The computer scientist François Chollet has proposed the following distinction: skill is how well you perform at a task, while intelligence is how efficiently you gain new skills.

Intelligence is about how easily you learn. By that standard, my two-year-old son is brilliant. He gains and improves skills every moment of every day and sometimes wakes up having received a cognitive upgrade. 

Viewed this way, AI can be seen for what it is: a very low-level form of intelligence. It’s extremely slow to learn, requires a colossal volume of data, and the result is a very fragile sort of knowledge.

Now, to be very clear, low-level intelligence is a historic milestone, very useful, and very promising. But my guess is it’s not a waystation en route to AGI.

This is not the AI you’re looking for

Chang refers to current AI as “turbocharged auto-complete.” This is exactly what it is. LLMs can cleverly predict words that are related to other words. That’s it.

We have an exaggerated idea of what current AI is because we conceive of it as, y’know, artificial intelligence. A more suitable descriptor would be Language Prediction Systems.

My guess is we are not on the path to AGI, but to something far less dramatic and far less threatening. This AI will be for gruntwork, not creative work and other cognitively demanding tasks.

AI’s superpower is boring work

The point where Chang and I diverge is on the topic of art.

Chiang is an artist and his focus is on AI’s application to artistic endeavors. He asks us this question…

…is the world better off with more documents that have had minimal effort expended on them?

Speaking as a small business owner and niche content creator, my answer is yes. There is not enough time in the day to write all the text that could benefit me and my audience. These tasks are not creative, they are mundane. I can’t generate all the promotional copy, support documents, summaries, drafts, and more that I need.

Also, many of these texts are only for internal use. Chang is a published author and assumes we all want to publish.

It’s clear to me that AI is useful for freelancers and small organizations. That’s what current AI excels at: information processing tasks.

AI doesn’t give you excellence. It gives you mediocrity. And most of the time, mediocrity is great. To paraphrase my friend Peter Nilsson, AI is superb at elevating you to the level of mediocrity. And guess what? We are sub-mediocre at almost everything.

This extends the productive reach of those of us who aren’t surrounded by skilled professionals.

AI may never create good art, let alone masterpieces… but that’s what I’d expect from a Language Prediction System.

This article contains an affiliate link to Ted Chiang's Amazon page. If you make a purchase through this link, I may earn a small commission at no additional cost to you.