You’re probably familiar with the same capitalist catechism I grew up with: that differences in wages reflect differences in human capital, and human capital reflects an ability to do things few others can. Lawyers are paid well because law school is hard, and cooks aren’t, since “anyone can cook;” and one day all those “menial” jobs would go away, replaced by automation.
A colleague was enthusing over lunch about how her partner (recently ejected from their job by shareholder imperatives) discovered that ChatGPT does a really great job of customizing cover letters, taking much of the effort out of applying for the kind of jobs that require cover letters – in other words, the kind of jobs that usually require college degrees, and come with benefits, paid vacation, and something more than a living wage.
Being a hawker requires neither a degree nor a cover letter, and we don’t yet have an AI that knows how to make char kway teow. Has the value of the labor that goes into each of these endeavors suddenly changed? Or has it been the case that white collar work was always more fit for machines, and “menial” all along, while char kway teow always required expertise that only a human could acquire.
Now that a new AI is passing the Turing test every week, is the char kway teow test the real benchmark?
But beyond this reductio ad absurdum, beyond the magnificent cruelty of our economy, beyond the question of whether today’s AIs will build other AIs that will build other AIs that will build robots that will finally make an acceptable char kway teow; there’s a real question: why don’t we have machines that can cook really well? Is there, in addition to the disincentives created by our exorbitant exploitation of other human beings, something intrinsic about food and cooking that we still can’t replicate? And if so, what does that tell us about what food is actually worth?
I actually have no doubt that you could create a machine that creates an excellent char kway teow. And if a huge community of people decided they were going to buy one maybe you could even produce it for some vaguely reasonable price. I think the challenge is that it would ONLY produce char kway teow. And you have to be really dedicated to char kway teow to use up counter space and money on a dedicated char kway teow machine. (I'm reminded of a conversation with you years ago about the special form used to make caneles)
I suspect AI is going to be very good at specialize tasks before it gets good at general ones. It seems "easy" to give an AI an understanding of contract law sufficient to create a licensing agreement between megacorps, for instance. This is the opposite of how we value human labor currently. We see a lawyer who has learned all there is to know about contracts but knows little about other areas of law as having very high value. We see a line cook at a diner who can make a wide variety of foods as having low value. Perhaps AI flips this on its head?
Wow. Great argument! And great discussion. Thanks, all.