Friday, March 29, 2024 -
Print Edition

AI and matzah: history and warning

By David Zvi Kalman

NEW YORK — In the last few months the world has been dazzled by an astonishing sequence of AI systems capable of performing all kinds of difficult tasks — writing code, composing poetry, generating artwork, passing exams — with a level of competence that rivals or exceeds what humans can do.

(Collage by Mollie Suss)

The existence of these AIs has prompted all manner of soul-searching about the nature of humanity. It has also made many people wonder which human tasks are about to be taken over by machines.

The capabilities of these AIs are new and revolutionary, but the story of machines taking over human jobs is not. In Jewish history the most important story of that transition has to do with matzah, and it’s a story that carries important lessons for the present day.

Starting 164 years ago, dozens of European rabbis engaged in a furious debate that would not be fully resolved until the beginning of the 20th century. Matzah, which for millennia had been made by human hands in accordance with the narrow constraints of Jewish law, could now be processed with a series of machines that promised huge savings of time and money.

As town after town adopted these machines, opposition began to rise, until it exploded in 1859 with the publication of An Alert for Israel, a collection of letters from prestigious rabbis who adamantly argued that for anyone interested in following the laws of Passover, a matzah made with a machine was no better than a loaf of bread.

The arguments for this position were many but all will sound familiar to anyone following the AI conversation. Like today, some objected to the machines just because they were new and different, but most had more specific concerns.

First, there was the matter of lost jobs. In many parts of Europe matzah was made by the poorest members of society, who were given the job as a way to help them raise money before one of the most cost-intensive holidays of the year. Ceding this job to machines would take work from those who could least afford it.

Beyond economics, there was concern that the machines just weren’t as reliable as people, especially given the rules around matzah-making outlined in Jewish law. What if bits of dough got trapped in the gears, quietly leavening for hours and unknowingly ruining whole batches of matzah in the process? What if the trays warmed the dough too fast? Without proper oversight, how could you trust your own food?

Finally, some objected to the loss of a literal human touch. Jewish law stated that matzah was supposed to be made by people who knew they were baking matzah. A machine, no matter how sophisticated, didn’t “know” anything. How could you eat matzah on Passover knowing that this most important food was made by a mindless machine?

The responses didn’t take long to arrive. A Cancellation of the Alert, a collection published the very same year, forcefully argued that machine matzah was perfectly fine — and possibly even better than the human product.

No, inventions aren’t inherently bad. No, the machines wouldn’t harm the poor, because the machines made matzah less expensive for everyone. No, the machines weren’t prone to error — and they certainly weren’t more error-prone than lazy, careless humans. No, the machines didn’t know what they were doing — but the people who built them did, and wasn’t that enough?

The machines eventually won, but then something happened that I don’t think either side anticipated.

With Manichewitz’s machine matzahs claiming most of the American market by the early 20th century, it was now the handmade matzah makers who were on the back foot; it was they and not the machines who needed to demonstrate that they were up to the difficult task of preparing this food with the efficiency and reliability of the machines.

The result is more than a little tragic. Matzah is the Jewish food with the deepest origins of all — deeper than brisket, deeper than latkes, deeper even than challah — and yet it is the ritual food most likely to be picked up at the supermarket and least likely to be made at home.

While there are still communities today that exclusively eat handmade matzah, even this job is now largely outsourced to just a few companies that resemble their machine-driven counterparts in scale. While teachers will sometimes demonstrate how to make matzah for educational purposes, across the religious spectrum the era of locally made matzah is over.

Despite the fact that it’s hard to imagine a simpler baked good — matzah is just flour and water, and it’s literally illegal to spend more than 18 minutes making it — its production is treated as though it is only slightly less complicated than constructing a jet engine, and people are worried about shortages as though matzah were a natural resource or an advanced microchip. The transition has been so complete that we barely remember there was a transition at all.

Did the rabbis pushing for machine matzah know this was going to happen? Almost certainly not. The economic impact of machine labor is relatively easy to predict, but the psychological and cultural effects are a lot harder.

There was probably no way of knowing how machines would change the way we thought about matzah in the long run, but today it’s clear that automating this ancient task has changed our own relationship to Passover’s central food — and because the change has resulted in a lot of alienation from matzah production.

I’m not so sure it was a change for the better. Making matzah locally could have been a way to feel connected to the ancient Israelites, who left Egypt so fast that they didn’t have time to make anything else.

Instead of emulating this ad-hoc food, we optimized it for cost and efficiency, in the process turning matzah into just another specialty cracker on the grocery store shelf.

Was it really worth it?

It’s probably a bit much to say that OpenAI is just a modern Manischewitz, but the parallels between the debate about machine-generated matzah and the present debate about machine-generated everything are useful for considering how short-term policy choices around AI won’t necessarily capture all of the technology’s long-term effects on how human beings want to spend their time.

When we relinquish an activity to an AI for economic reasons, we may eventually come to believe that humans are no longer qualified to do the task at all.

Then as now we must balance our economic needs against our ideas about what kinds of activities make for a good and fulfilling life.



Avatar photo

Leave a Reply