When the Sukkah Becomes a TARDIS
On AI, Moral Purity, and What We Do With Powerful Tools
It all started when my friend Rabbi David Seidenberg made a post comparing the sukkah to the TARDIS.
Attached to the post he included an image of a TARDIS-sukkah that was almost certainly AI-generated: the familiar blue police box, festooned with greenery, opening onto a warmly lit sukkah within.
I responded by mentioning my book The Enchanted Sukkah. David’s post hit on something I had been playing with there: the sukkah as a place where ordinary boundaries loosen. Inside this temporary hut, time gets strange. Through the custom of ushpizin, we invite ancestors into the sukkah as guests. The past does not stay politely in the past. It comes forward and sits down at the table.
A TARDIS is actually a pretty good metaphor.
Two comments underneath David’s post caught my attention, though, and stuck in my craw.
One person wrote:
“I don’t think using AI is harmonious with environmental, ethical, or spiritual values.”
Another reminded readers that AI had “learned” by copying books whose authors were never compensated or credited, and suggested that the generated response itself illustrated the problem.
I understand both objections.
I also think they point toward a larger problem in the way we are beginning to talk about AI.
We are turning a complicated moral question into a purity test.
There are no clean technologies
AI has environmental costs. Large data centers consume significant amounts of electricity and water. The manufacture of the chips on which AI depends is resource-intensive as well.
That matters.
It also describes the device on which you are reading these words.
Semiconductor manufacturing itself consumes enormous amounts of water and energy and produces hazardous waste. One study of 28 semiconductor companies found global semiconductor manufacturing consumed roughly 789 million cubic meters of water and 149 billion kilowatt-hours of energy in 2021. ScienceDirect A more recent review describes semiconductor manufacturing as highly water- and energy-intensive and dependent on materials that can produce toxic and persistent wastewater. ScienceDirect
Our phones contain minerals extracted through supply chains with their own histories of exploitation. Our cars, clothes, food, batteries, cloud storage, air travel, heating systems and streaming services all participate in systems that impose costs somewhere.
This does not mean that the costs do not matter.
It means that “this technology causes harm” cannot, by itself, settle the ethical question.
Otherwise the morally consistent response would be to stop using almost everything.
Most of us do something more complicated.
We try to consume less.
We demand cleaner supply chains.
We regulate pollution.
We pressure companies.
We compensate people whose work has been exploited.
We change laws.
We try to repair what can be repaired.
And then we ask what the technology is for.
That question matters too.
AI is already doing good
It is easy to talk about the potential benefits of AI as though they belong to some speculative future.
They do not.
In January 2025, Nature Medicine published the results of a real-world breast-cancer screening study involving 463,094 women. Radiologists using AI-supported mammography detected 17.6 percent more breast cancers than those using standard double reading—about one additional cancer for every thousand women screened—without increasing the recall rate. Nature
Think about what that means.
Somewhere, there is a woman whose cancer may have been caught earlier because a radiologist had another set of eyes.
They happened to be silicon eyes.
AI is also beginning to change drug development.
In 2025 researchers published the results of a randomized Phase 2a trial of rentosertib, a drug candidate for idiopathic pulmonary fibrosis whose molecular target and compound were discovered with the help of generative AI. It was one of the first concrete demonstrations of an AI-discovered drug reaching this stage of human clinical testing. Nature
None of this proves that AI is good.
Technology is rarely “good” in that simple sense.
It proves something more important:
The moral ledger has entries on both sides.
If a technology consumes resources, that belongs in the ledger.
If artists and writers were treated unjustly in its development, that belongs there too.
If it makes oppressive surveillance easier, put that in.
If corporations use it primarily to eliminate employees and transfer more wealth upward, put that in enormous letters.
But if it catches cancers, accelerates drug discovery, makes information accessible to people with disabilities, helps scientists analyze quantities of data no human team could realistically process, or allows a small organization to accomplish things once reserved for institutions with enormous budgets, those consequences belong in the ledger too.
Ethics requires counting all of them.
What about the writers?
The copyright question is real.
Authors, artists, musicians and other creators have legitimate questions about how their work was used to train commercial models, whether consent should have been required, what constitutes fair use, and whether compensation is owed.
Those questions are being litigated right now.
They should be.
But saying that AI “copied all the books in the world” quietly settles several disputed questions before the argument even begins.
Training on a work and reproducing that work are not identical acts.
A model does not ordinarily contain a searchable library of all of the books on which it was trained. Training alters billions of numerical parameters according to statistical relationships learned from data.
That fact does not resolve the copyright question either.
It simply means that the interesting question is harder than “the computer stole the book.”
We need to work out what creators are owed in a world where machines can learn patterns from human culture.
And we should.
Perhaps licensing regimes will change. Perhaps collective compensation systems will emerge. Perhaps courts will define clearer boundaries. Perhaps some uses of copyrighted material will turn out to have been unlawful.
If people were wronged, restitution matters.
But I am deeply skeptical of the idea that an unethical element in the origin of a technology permanently contaminates every downstream use of it.
Human civilization could barely survive that principle.
American music alone would collapse under its weight.
Black musicians were systematically exploited by record companies, publishers and white performers. Songs were copied. Royalties disappeared. Entire musical forms were commercialized while many of their creators remained poor.
We should tell that history.
We should credit the people who were erased.
Where restitution remains possible, we should pursue it.
And I am still going to listen to Elvis.
The danger of moral purity
There is a form of ethical reasoning that asks:
“Is my hand clean?”
There is another that asks:
“What produces the most repair?”
Those questions overlap.
They are not identical.
Imagine a radiologist ten years from now confronting an AI system that has been repeatedly demonstrated to detect a particular cancer earlier and more reliably than unaided human reading.
The physician refuses to use it.
The training data were ethically compromised, she says.
The energy costs are too high.
The companies developing the technology behaved badly.
Then she misses the tumor.
At that point, abstention is no longer morally neutral.
It has consequences too.
That is the part of the discussion about AI that I think gets lost.
We talk about the ethical cost of using a technology.
We rarely talk about the ethical cost of not using it.
Suppose AI eventually helps discover a treatment for pancreatic cancer five years sooner than researchers otherwise would have.
How many deaths does five years contain?
Suppose it helps us design better batteries, model climate systems more accurately, identify new antibiotics, or make high-level tutoring available to children whose parents cannot afford private teachers.
At what point does refusing the tool become its own moral choice?
Judaism has another model
Jewish tradition knows something about living inside compromised structures (see Gittin 55a–I’ve discussed this in previous blog posts.)
The rabbis discuss the case of a stolen beam that has been built into a house.
Strict justice might seem to demand that the thief dismantle the entire building in order to return the original beam.
Instead, the rabbis institute takkanat hashavim—an ordinance designed to make repentance possible.
The thief must make restitution.
The theft does not become permissible.
The victim’s claim does not disappear.
But the entire house does not have to be demolished.
There is something profoundly wise in that.
The goal of ethics is not simply to identify contamination.
It is to create the conditions for teshuvah.
For return.
For repair.
For taking something damaged and asking what responsibility now requires.
AI has serious ethical problems.
So regulate it.
Make its environmental costs transparent.
Build cleaner data centers.
Protect workers.
Protect privacy.
Create meaningful rights for artists and writers.
Demand disclosure from companies.
Prevent discriminatory systems from making consequential decisions without accountability.
Break up concentrations of power when they become dangerous.
And use the damn thing to cure cancer.
Those impulses are not contradictory.
They belong to the same moral project.
What kind of sukkah are we building?
A sukkah is a strange structure.
Its walls are flimsy.
Its roof leaks light.
It reminds us that the things we build are temporary and imperfect.
And then Judaism tells us to sit inside anyway.
Eat there.
Sing there.
Invite guests.
Make it holy.
That may be a useful spiritual posture toward technology as well.
We do not live in a pure world.
We inherit compromised materials.
We participate in systems whose harms exceed our ability to disentangle ourselves from them completely.
The spiritual question cannot simply be whether we have managed to remain untouched.
It is what we build with what has been placed in our hands.
Perhaps that is why David’s AI-generated TARDIS-sukkah delighted me.
It combined an ancient ritual technology with an absurdly modern one and made something neither could have produced alone.
The ancestors entered the time machine.
The time machine entered the sukkah.
And for a moment, the question was not whether the tool was pure.
The question was where we wanted it to take us.
