The Turing Dilemma

In The New Yorker's "Weekend Essay," the fabulous science fiction author Ted Chiang opines on "Why A.I. Isn't Going to Make Art." It's a good essay on a lot of fronts - well written, thoughtful, passionate, with many interesting arguments, and unfortunately I think fundamentally wrong.

Out of the gate Chiang starts with the weary trope that fiction written by large language models is "terrible" without including any definition of what would make fiction "good" vs. why all of what can be produced today through the use of generative AI is therefore terrible. The second paragraph goes on to justify why he thinks a qualitative definition of art (fiction in this case) is unnecessary because in his view what matters is the methodology for the creation of such art. This framing undermines the rest of the arguments in his essay, as it sets up the logical fallacy of circular reasoning. The creation of fiction by a machine cannot be art because it is necessary to have a human create fiction for it to be art:

"When you are writing fiction, you are--consciously or unconsciously--making a choice about almost every word you type... When you give a generative-A.I. program a prompt, you are making very few choices..."

What Chiang has failed to encompass in his critique of computer generated art is the meta question of how we define art -- as a creator he naturally aligns with the creation process as the crucible for defining something as art. But the rest of us as consumers of fiction (or painting or whatever) define the work through what we perceive and bring to it in experiencing it ourselves.

This is important because actually in the vast majority of cases we experience art without knowing the provenance of the artwork. Thus each time we encounter artwork we unconsciously engage in a bit of a Turing Test--has this painting or this article been created by a human or a machine? And what I'd like to call the Turing Dilemma from this is, does it really matter--is it how I am affected by the work that matters or how it was created?

Well beyond artwork this is the dilemma of the next decade as we get used to a world in which more and more of the things we encounter will have been created with some measure of computation. Was this email from a coworker actually written by them? To take the broadly derided Google advertisement in which a fan letter to an Olympian is created using Gemini (which Chiang also cites) -- what are we mad about? That the Olympian received a message created by a machine? Or that the child didn't have the opportunity to express herself?

I would propose that the resolution of this dilemma is to separate out these two questions and focus on the concern that as human beings we are abrogating creative experience and thus losing the values that we derive from engaging in that process. And so instead of focusing on whether or not machines can create art, we should be exploring what humans should be doing to make sure that we retain the benefits of creativity for our own cognitive and emotional development.

A parallel can be found in the "Homework Apocalypse" which Ethan Mollick has described and just this past week further explored on his blog in a piece entitled "Post-Apocalyptic Education." As Mollick points out there is clear value to learning from doing homework, a value that already was being eroded through students' access to the Internet long before generative AI accelerated this decline.

We should ask, what can we do to replace the cognitive value students gain from doing homework? I'll say this, and as soon as I do I think you'll agree that it is an obvious answer: replace homework with 1:1 tutoring. Which by the way, generative AI is very good at as Sal Kahn has explained in his terrific book "Brave New Words" and demonstrated through the applications built by Kahn Academy. Generative AI gives us the means to provide every learner with a personal tutor - tirelessly patient and expansively capable to meet each person at the point in their learning journey with the right approach for that learner. And tutoring is much more valuable to a student's cognitive development than homework which itself developed simply as a cost-effective alternative to providing 1:1 tutoring.

Yes we need our brains to develop through exploration, curiosity, comprehension, and creation. The activities which future generations engage in will be different from those of ours and our ancestors. Better? Worse? This is of course the same question asked of every technological change to our society and civilization. On balance, I'd say things have been getting better... We can navigate this change too. Without declarations that creations which have the power to move us are or are not "art" based on some measure of the technical ingredients which went into their creation.