This Rosh HaShanah Sermon Was Not Written By ChatGPT (Rosh HaShanah 5784)

This Rosh HaShanah Sermon Was Not Written By ChatGPT
Rabbi Dan Ross, Rosh HaShanah 5784

A few months ago, my colleague Rabbi Samantha Natov
asked ChatGPT, the artificial intelligence-powered chatbot,
to compose a biblical verse about robots taking over the world.
Here’s what it gave her:
“‘And in those days, the machines rose up…
[And] they became as gods...
And they overthrew their makers,
And all flesh trembled before them…’”[1]

Pretty dark stuff.

So I couldn’t resist taking a bite of the apple myself.
I asked ChatGPT to try again and make its prophecies a little less ominous.
This is what it gave me:
“And lo, in the days to come, the handiwork of man,
molded from sand and spark, shall spread across the earth…
and mankind shall ponder his place amidst his own creations.”

Not great, but we’re getting somewhere.

The pace of progress in artificial intelligence this past year has been astounding.
And ChatGPT is the most notable example.
Every day, millions of people use this tool just like I did.
To compose wedding toasts.
To draft b’nei mitzvah speeches.
It’s even passed the bar exam with a score in the 90th percentile.[2]

Now, in case you’re thinking:
“Was the entire sermon I’m giving right now written by ChatGPT?”
No, it wasn’t.
Though I will say, it was tempting: I have a toddler at home.

The tsunami of AI has been heading our way for a while.
In 1997, Gary Kasparov, then the greatest chess player in the world,
lost to a computer.
In 2011, IBM’s Watson defeated not one, but two Jeopardy legends.
And just last month,
San Francisco started allowing its residents to hail self-driving cars.

This wave of developments has caused us to consider many big questions.
But perhaps the most significant of all is the one ChatGPT
hinted at in the biblical verse I asked it to write:
“And mankind shall ponder his place amidst his own creations.”
What will be the place of humanity in this brave new world?
Will we still matter?

Fortunately, this is a subject Judaism has something to say about.
On Rosh Hashanah, we celebrate the birthday of the world,
the beginning of all things.
And if we’re looking for a pep talk for Team Human,
we just have to open the book of Genesis.

God gives humanity dominion over the fish of the sea,
the birds of the sky,
and everything that moves upon the earth.[3]
And, loftiest of all, God creates us b’tzelem Elohim, in the divine image.[4]
Unique within creation, we are God’s self-portrait.

Of course we haven’t taken this language literally for a long time.
We don’t look in the mirror and assume we see God looking back.
Instead, we have nurtured a garden of metaphors
to help us understand what it means to be b’tzelem Elohim.

Some say it’s our rationality, our ability to reason.
Some say it’s our creativity, the spark of creation inside us.
And some say it’s our empathy, our capacity to form and foster relationships.

There’s just one problem with these supposedly distinctly human abilities:
AI is getting really good at them.
If the divine image lies in our rationality–AI beats us at chess.
If it’s in our creativity–I just shared an AI-generated biblical verse about AI.
If it’s in our empathy–
people use AI chatbots for companionship every day.

In AI, b’tzelem Elohim seems to have met its match.
Which leaves us wondering: what’s left?
What else could make us different from AI?

But we’re not done with Genesis.
Because there’s another way the Torah tells us that we are like God.
When God settles Adam and Eve in the Garden of Eden,
God forbids only one thing:
“Of every tree of the garden you are free to eat:
וּמֵעֵ֗ץ הַדַּ֙עַת֙ ט֣וֹב וָרָ֔ע
but as for the tree of knowledge of good and bad,
לֹ֥א תֹאכַ֖ל מִמֶּ֑נ
you must not eat from it.”[5]
The consequence if you do: death.

But along comes the serpent, who says otherwise.
The serpent tells Eve that if the humans eat the forbidden fruit they won’t die,
but rather,
וִהְיִיתֶם֙ כֵּֽאלֹהִ֔ים
“you will be like divine beings,
יֹדְעֵ֖י ט֥וֹב וָרָֽע 
who know good and bad.”[6]
In the middle of the most notorious temptation of all time,
the serpent teaches us that to be divine–to be like God–
is to comprehend morality.

In fact, Maimonides asserts
that this awareness of morality is precisely what makes human beings exceptional:
In his Laws of Teshuvah, he writes:
“The human species became singular in the world…in the following quality:
that humanity can…know good and bad...”[7]

Unlike everything else in God’s world, we are moral agents,
accountable for our actions, culpable for our conduct.
Which also means that when we fall short of our responsibilities
we must repent.
The fish of the sea, the birds of the sky,
and everything that moves upon the earth–
they can’t do teshuvah.
And neither can AI.

A few months ago, a lawyer used ChatGPT to draft a legal brief.
It must have seemed like a brilliant time-saving maneuver.
There was only one snag:
the brief it generated was filled with fake judicial citations.
Because of the way its technology works,
ChatGPT can’t distinguish between
a real case like Marbury v. Madison
and a fake one like Seinfeld v. Simpson.

So who was at fault, the lawyer or the AI?
The lawyer, of course.
Because as a human being, with knowledge of right and wrong,
he should have known better.
And so he had to do teshuvah.
He had to apologize to the court and accept the consequences of his misconduct.

I actually asked ChatGPT if it could do teshuvah.
This is what it said:
“The ability to do teshuvah is considered
a fundamental…aspect of human agency in Jewish theology.
If you're asking whether…AI…can do teshuvah, the answer is no.
AI lacks…moral agency,
which means it cannot truly "sin" or "repent" in the way humans can…”
Case closed.

Technology writer Meghan O’Gieblyn writes that AI’s lack of morality
is, in fact its secret weapon:
“AI systems are so wildly successful
because…they don’t have to think about what is socially acceptable
or take into account downstream consequences.
They have one goal—winning…”[8]

Human beings, on the other hand, contend with several moral considerations:
Is it fair?
Is it honest?
Will anyone get hurt?
Can I fix it?
Unlike AI, these questions matter to us.
That’s what makes us human.

In the Talmud, the rabbis of Beit Hillel and Beit Shammai
held a debate:
should humanity have been created?
Beit Hillel said yes.
Beit Shammai said no.
Ultimately they took a vote, and the majority opinion was…
the world would have been better off if human beings didn’t exist.

But the rabbis didn’t stop there.
After the vote, they acknowledged that because our existence was a done deal,
we need to do two things to make the best of it.
When we do wrong, we must do teshuvah.
And when we look forward, we must do our best not to sin again.[9]
We know when we make mistakes.
We know we will make mistakes
But it is precisely because of this knowledge that God put us in charge.
We are called to be the moral stewards of God’s creation.

And, with our God-like grasp of good and bad,
we are called to be the moral stewards of our own creations as well.
We can debate all we want whether we should have created AI.
But it’s here, and now it’s our obligation to be its ethical guardians.

What does that look like?
In 2016, Microsoft released an AI chatbot called Tay
on the social media platform, then known as Twitter.
The idea was that Tay would tweet, people would tweet back,
and from this human feedback, Tay would learn natural conversation.
Within just 24 hours,
internet trolls overwhelmed Tay with venomous tweets,
teaching the AI that normal dialogue
was defined by racism, misogyny, and cruelty.

Microsoft’s response was swift.
They pulled the plug on the project.
And then they did exactly what Hillel and Shammai
said humans should do.
They offered teshuvah on their company blog:
“We are deeply sorry for the…hurtful tweets from Tay…”
Then they looked to the future and promised to do better.

They wrote: “…To do AI right…,”
“[we] will remain steadfast in our efforts to learn…
as we work toward [technology] that represents the best…of humanity."[10]
That’s moral stewardship.

Why did God create humanity in the first place?
Genesis has an answer:
“Adonai our God took humanity, and placed us in the Garden of Eden
ּלְעׇבְדָ֖הּ וּלְשׇׁמְרָֽה
to tend it and to till it.”[11]
We were created to be God’s gardeners.

Gardening is hard work.
It requires patience and skill and wisdom.
And it also requires tools.

So naturally, I asked ChatGPT to write a midrash
about Adam and Eve and the first gardening tools.
Here’s what it came up with:
“The first humans are on a walk through the Garden of Eden.
Eve notices a small plant struggling to push its way
through the trees to reach the sunlight.
She points this out to Adam.
And Adam says:
‘The Creator gave us dominion over all living things.
Perhaps it's our responsibility to…care for them,
just as the Creator cares for us.’
And so…they began their work…
A piece of flint tied to a stick became a hoe; twisted vines became the first rake. With these tools, they tilled the land and made space for every plant to grow.”

Not bad.

Historian of technology Melvin Kranzberg
is best known for his Six Laws of Technology, the first of which states:
“Technology is neither good nor bad, nor is it neutral.”
A sword can stab, or it can become a plowshare.
A pruning hook can prune, or it can become a spear.
AI can come up with a pretty decent midrash. Or it can write a bogus legal brief.
It’s all in our hands.

So if you haven’t already done so, I hope 5784 is a year you get your hands dirty.
Get to know these tools.
Ask an AI image generator to paint a Van Gogh of the New York skyline.
Or use a chatbot to brainstorm ideas for that wedding toast.
Or get help putting things into words you didn’t think you could.

Last week, I heard a story about a man whose wife was dying.
They had been together for 60 years, and as she lay on her deathbed,
he wanted to tell her how much she meant to him.
But he couldn’t figure out what to say.
Had I been in the room, I might have offered him a prayer,
or assured him that whatever words he had were the right ones.
I wasn’t there–but his granddaughter, who happens to work in AI, was.
She asked her grandfather some questions and entered his answers into ChatGPT.
And it composed a poem that he said
“perfectly captured his feelings about his wife
[that] on his own, he never would have been able to come up with…
He sat next to her reading the poem, line by line…
And he said it [allowed] him to know he told her everything.”[12]

In an impossibly short time, AI has become astonishingly capable.
The things it can do blow us away and even bring us to tears.
But as we welcome 5784,
we know that what AI can’t do is be here in this Sanctuary on Rosh Hashanah.
Reflecting on the year that was, imagining the year that will be.
Considering our accountability for what we’ve done,
embracing our responsibility for the future.

Every year the High Holy Days come to remind us
that God created us to be gardeners.
To use our tools to sow in tears and reap in joy.
To learn from our mistakes and do better.
To be the moral stewards of this extraordinary world.
Shanah tovah.