I’ve been thinking about Golems, the
mythical artificial automata of Jewish mythology, and headed off to the British
Library to dig out Leivick Halpern’s 1921 script Der Golem. It’s a play
based on a story about a sixteenth-century Prague-based Rabbi, Yehudah Ben
Betzalel Loew – known as the Maharal.
The Rabbi, we are told, wanted to create an
automaton capable of solving the problems of his age, hewing water, chopping
wood, defeating antisemites and the like, but in Halpern’s play, Der Golem
doesn’t emerge mighty and ready to serve as “a people's champion.” But rather
as an emotionally needy ..., well he sounds like a teenager. Awakening into
life, in Scene One, Halpern’s Golem shares this;
“Where
is the soul that will be breathed in me? Why is a heart not given me? Where is
the blood that must be poured to flow in me?”
The Rabbi persists with his great plans for
his creation, but, spoiler alert – it doesn’t end well. In Scene Eight the stage
directions describe the moment;
THE
GOLEM sinks to the floor. He starts to tear his hair and clothes. The Sabbath prayers
are heard through the door. The candles on the reading stand flicker. THE
GOLEM, all at once, takes an axe, the axe with which he used to chop wood. He
raises the axe, runs to the window, smashes the glass with the axe and jumps
into the street.
The script
continues with instructions for a cacophony of voices delivering these lines.
The
servant! The Golem! With his axe!
Out
there are two with bloodied heads!
He
is destroying Prague!
You are probably
ahead of me in understanding my interest in Der Golem.
What if the Golem could
serve as an archetype for the new artificial pseudo-life form of today? What if
the Rabbi, could serve as a precursor for Sam Altman or Demis Hasabis? And the
Golem, a sort of Chapt GPT or Claude? What if chopping wood and hewing water are
precursors for folding proteins or creating memes of talking cats? What if the
antisemitism facing the Rabbi in the late sixteenth century Prague was a
precursor to … well, some things don’t change.
I’m not the first
person to reflect on the Golem as an archetype for thinking about AI. I’ve been
beaten to that honour by Gershom Scholem, the greatest scholar of Jewish
mysticism, who, in 1965, insisted the first supercomputer to be built in Israel
– in Rehovot no less - should be named Golem Aleph. In a speech titled The
Golem of Prague and the Golem of Rehovoth, Scholem suggests;
The
creation of a Golem is an affirmation of the productive and creative power of
Man. It repeats, on [a] small scale, the work of creation. [It is] Man [trying
to to] do, in his own small way, what God did in the beginning. [1]
And he concludes,
and again, this is 1965, with this line;
I
resign myself and say to the Golem [of Rehovot]: develop peacefully and don't
destroy the world. Shalom.
We’ve been
worrying about AIs and Golems getting out of control and destroying the world
for a while now. And I’m scared too. I understand, at least partially, the big threat
– the threat of a self-replicating super-intelligence turning against its creator.
I understand too,
at least partially, the more immediate threats; the scraped data, the way AI sucks
up energy and water and jobs and all the rest of it. But I’m not going addressing
those threats sermonically today. Neither am I going to give a sermon on the possibility
of AI curing cancer or solving the problem of global warming. Those things are not
areas of expertise.
Instead, I want to
give a sermon tonight – and actually two sermons tomorrow – on this question. What
should we, normal human beings - well Jews - outside the corridors of power and
control of AI, do in the face of AI? I hope to give us something to do,
something to consider and something to believe in.
The key line in Scholem’s
speech about the Golem of Rehovoth, is this, “The creation of a Golem repeats,
on however small a scale, the work of creation.”
Scholem had in
mind a Midrash from Vayikra Rabba[2]
which recalls that when God created humanity, God gathered dust and kneaded and
formed it into a Golem. And then, having formed the Golem, God breathed into the
Golem a soul – and so the first human was created. The Golem, in the eyes of
the Rabbis, is a step on the way to being fully human, a step a great Rabbi,
like the Maharal of Prague, could match, before the final step – the granting
of a soul – which is beyond even the greatest of us.
There’s another
part of ancient mythology we need – we need an answer to the question, “why are
we here at all, why was humanity created?” From a religious perspective, God
was after something very specific and I think counterintuitive– when God
created the first human being “way, way back, many centuries ago.”
From a religious
perspective, long before even “Jacob, Jacob and sons,” there was God and God’s
angels. And each of God’s angels were responsible for individual specific tasks.
There was the angel responsible for growing the grass, and the angel
responsible for ensuring the sun passed through the sky, and angels whose only
task was swirling round the heavenly throne beating their wings and chanting ‘Kadosh,
Kadosh, Kadosh.’ And each angel knew its shlichut, its task, and
each angel did its task perfectly. You might think that this would suit God
well.
Yet, in the face
of angelic perfection, God chose to create Adam, the precursor to you and me. And
we aren’t as reliable as an angel, or as expert. We make mistakes. We’ve always
made mistakes. God set us up with one law to keep, and we broke it almost
immediately. And even today, 5787 years later, we are still spectacularly fallible
and unreliable.
But let’s suppose
God knew what God was doing, in turning down the company of perfectly competent
angels in search of fallible, unreliable you and me. Let’s suppose that what
God really wanted as a companion in creation, was a being who tried, and
failed, and tried again, failed again and only ever aspired to failing better.[3]
Because we are
never going to get to perfect, you and me, no matter how many times we recite
the Kol Nidrei and pledge ourselves to stop lying and acting presumptuously and
all the rest of it. We are simply not built that way. We are built to explore,
to clamber over obstacles and bang things together to see what happens, even if
we fall off while clambering or bash someone while banging.
This might be part
of the problem with all the Yom Kippur talk of Teshuvah and self-improvement –
we are never going to become perfectly competent humans. I’m never going to be
an angel. The only thing I am ever able to become is a better version of me.
Reb
Zusha was laying on his deathbed surrounded by his disciples. He was crying and
no one could comfort him. One student asked his Rebbe, "Why do you cry?
You were almost as wise as Moses and as kind as Abraham." Reb Zusha
answered, "When I pass from this world and appear before the Heavenly
Tribunal, they won't ask me, 'Zusha, why weren't you as wise as Moses or as
kind as Abraham,' rather, they will ask me, 'Zusha, why weren't you Zusha?' [4]
The problem with
all this talk about Teshuvah is that it can strip us of the thing that makes us
most human. The problem of all this talk of a God judging our every action is
that it could lead us towards an unintended conformity, stifling our uniqueness.
I get that the promise
of automaton can feel attractive – that’s what the Rabbi in Halpern’s play Der
Golem wanted, a perfectly reliable solution to the problems of his time – but
his wish, that wish, doesn’t work. It’s self-defeating because it runs at
cross-purposes to what it means to grow in a changing world – it runs at
cross-purposes to what made us human in the first place.
I’m not suggesting
complacency when we fall short, but rather nurturing delight in being ourselves;
crawling around the playpen of our existence poking and prodding and exploring and,
yes, failing as an essential part, of what it means to be me.
In a world where
the Large Language Models are so efficiently predicting the most likely correct
next token, our task, as humans I think, has to be;
Be less normal.
Don’t assume decisions based on statistically reliable next steps are the right
things we should be doing with our human lives.
Be more odd, be
more creative, be more prepared to fail, be unexpected, be uniquely located in
our own uniqueness.
Be something no
angel and no Golem could ever manage.
In a world where
the Large Language Models are so efficiently predicting the most likely correct
next token, our task is to be human.
In part, this is
about our Jewish identity – we’re supposed, as Jews, not to be same as for
other people. We are supposed to eat odd
food, or not eat any food at all. We are supposed to come to Shul on the night
of Kol Nidrei to pray in an ancient language, beating on our hearts as we
recite a list of sins in the plural. As Jews, we are supposed to .... – it’s a
long list. We are supposed to be very odd indeed.
Kedoshim T’hiyu, teaches the Torah – be holy, Perushim
T’hiyu[5],
the rabbis respond – be different.
But in part it’s
about a way of reconsidering our lives as essentially playful. The world might
be tempting us to do thing I talked about on Rosh Hashanah – incurvatus in
se – fold in on ourselves. But we need to reject that. In the world out
there, we tend to insist on our being right with such conviction that we become
unable to brook disagreement. We are, as humans, increasingly damning the
people who disagree with us for the supposed sin of disagreeing with us. These
might feel like an effective strategy, but in truth, this attitude limits our
ability to adapt – the very secret to our human success.
There’s an irony in
that as artificial learning adapts and learns every more quickly, we are
becoming more fixed, more brittle and more knee-jerking in our reactions to
challenges we haven’t already faced. We are becoming more predictably
automaton, more like, and not in a good way, the angels that God rejected when
God decided to create humanity in the first place.
I know it’s dark
out there, it’s the Yahrzeit of the deaths in Heaton Park last year and the Middle
East is a mess and whole ecosystem is a mess. But we cannot meet the serious
challenges of our time with predictability, with an attempt to become more
productive tokens capable of more accurately working out the next most likely
correct word in a sentence.
We need to be more
like Halpern’s Golem – the creation who wanted to know, “where’s my soul.” We
have to care more about our unique, exquisite and playful soul.
The true call is
to be unique and glorious, more unpredictable than even God could have dreamed,
more human than a machine can replicate, even if we, even as we, fail. That is
I think, the very heart of what it means to be human.
I want to end with
our founding Rabbi, Louis Jacobs.
Louis know
Professor Gershom Scholem, the great scholar of mysticism who considered the
relationship between AI and Golems back in 1965. And on Scholem’s death, Louis
wrote a eulogy.
“On more than one Yom
Kippur,” [wrote Louis] “[Scholem] – attended services at my Synagogue [that
would be us!] It was a frightening experience to have to preach on the great
themes of the day with Gershom Scholem as a member of the congregation” [Amen
to that]. And then Louis attributes an observation to Elaine Thayer, first wife
of the supremely idiosyncratic poet e.e. cummings. Louis uses the quote to
explain the esoteric, hidden world of Kabbalists, and the enigmatic scholarship
of Scholem, their most famous scholar. As a line, I think it captures everything
what we should desire of ourselves and the people around us in this new
artificially assisted, so-called intelligent world.
“I don’t like,”
said Thayer, “people who wish to be understood.”
If I might be so
bold, I suspect God would agree.
Posts
No comments:
Post a Comment