MACHINES of LOVING GRACE
- Jeremy Connell-Waite
- 6 days ago
- 7 min read
“Machines of Loving Grace” is a landmark 14,000-word essay on AI from Anthorpic’s Dario Amodei which everyone should read.
Whether you admire Dario Amodei’s vision for AI or question the ethics of the company he leads (as they cut up thousands of old books), Machines of Loving Grace deserves to be read - not because you have to agree with it – but because this is what serious thinking looks like on the page.
Reading it thoughtfully will help you write better yourself, and to “think in ink.”
The essay uses “The Rhythm of Persuasion”
Abstraction: What’s the big idea?
Mechanism: How could this work?
Consequence: What could change?
Human Life: What does this mean for real people?
It’s a powerful pattern repeated throughout which swings backwards and forwards, from the head to the heart, to make even the most technical parts engaging.
I made this infographic as a cheat sheet you could use yourself.
It’s a 55-minute read (but worth the effort).
“Machines of Loving Grace” https://darioamodei.com/essay/machines-of-loving-grace
--
DEEPER DIVE
Machines of Loving Grace is roughly 14,000 words, highly technical in places, heavily qualified, full of footnotes, and occasionally rather dense. It is not Gladwell. It is not Michael Lewis. There is almost no conventional narrative storytelling. And yet it is an exceptionally effective piece of persuasive writing.
Its great achievement is that Amodei makes an almost preposterously large claim sound sober, reasonable and eventually imaginable. That is a formidable writing skill. It’s a great example of how to write about an extraordinary idea without sounding extraordinary yourself.
1. The whole essay is built around one enormous tension.
The opening is excellent. Amodei begins with the identity readers already associate with him: the man worried about AI risk. He acknowledges the assumption that this makes him a pessimist — a “doomer” — and immediately overturns it.
His central paradox is essentially: I worry so much about AI going wrong precisely because I believe so profoundly in what could happen if it goes right. That is a beautiful piece of argumentative architecture.
It creates what storytelling people might call an expectation gap:
You think I believe X.
Actually, I believe almost the opposite.Let me explain why…
And then comes one of the most important sentences in the essay: most people, he argues, underestimate both the radical upside and the terrible downside of AI. Notice the symmetry: underestimating the upside / underestimating the downside
That antithesis gives the essay its intellectual territory. He refuses both techno-utopianism and doom. He occupies the interesting space between them. That makes us curious.
2. He establishes ethos before making the outrageous claims.
This may be the most useful writing lesson in the entire piece. Before Amodei tells you that AI could help eliminate most cancer, prevent Alzheimer’s, cure genetic disease and perhaps double human lifespan, he spends considerable time telling you why you shouldn’t entirely trust predictions like these.
He says he wants to avoid:
propaganda
grandiosity
science-fiction baggage
excessive certainty
He describes his predictions as guesses. He repeatedly acknowledges that he may be wrong. He identifies areas outside his expertise. He even says the essay should perhaps be considered a starting prompt for people who know more than he does.
Rhetorically, this is concessio — conceding limitations in order to strengthen the argument that remains. And it creates enormous ethos.
The paradox is that his uncertainty makes us more willing to believe his certainty.
Compare: AI will cure cancer with the rhetorical approach Amodei takes:
Here are the reasons this might not happen. Here are the bottlenecks. Here is what intelligence cannot solve. Nevertheless, given what we know, here is why I think something remarkable may happen.
The second voice sounds like a scientist thinking in front of us rather than a salesman selling to us. That distinction matters enormously.
3. He doesn’t just tell us what he thinks. He lets us watch him think.
This is probably my favourite characteristic of the writing. AI-generated business writing often presents conclusions. Great intellectual writing frequently presents reasoning.
Amodei continually uses constructions such as:
“Why do I think this?”
“An obvious question…”
“Another way to say this…”
“Two extreme positions…”
“Let’s take these one by one.”
“What happens if…?”
The prose therefore reproduces the movement of thought.
We don’t feel as though we’re reading a polished corporate position paper. We feel as though we’re sitting beside an unusually intelligent person with a notebook while he works through a problem.
This is something Feynman did beautifully too. The writer isn’t saying: Look how clever I am. He is saying: Come over here. Look at this. I like that. That is a profoundly different relationship with the reader.
4. His masterstroke is the metaphor.
The essay contains several memorable phrases, but the great one is: “a country of geniuses in a datacenter.”
That’s exceptional compression. Before that phrase, Amodei spends hundreds of words defining powerful AI: Nobel-level intelligence across disciplines, millions of copies, autonomous work, computer interfaces, collaboration, enormous speed.
Then he compresses the whole concept into seven words. Suddenly we can see it.
This is what metaphor does at its best. It isn’t decoration. It is cognitive compression.
And notice how much better it is than: “Artificial superintelligence.”
AGI is abstract.
A country has population, labour, expertise, coordination and productive capacity.
A country of geniuses introduces capability.
A country of geniuses in a datacenter creates scale and physicality.
Those seven words carry perhaps a page of argument. That phrase alone is worth the entrance fee.
5. He repeatedly moves from abstraction → mechanism → human consequence.
This is the basic rhythm of the essay.
Take biology for example. He doesn’t simply say: AI → better medicine.
He begins with limiting factors: experimental speed, data quality and biological complexity. Then he explains why merely applying AI to existing datasets isn’t enough. Then he reframes AI as something closer to a virtual biologist capable of directing the entire scientific process. Only after establishing the mechanism does he allow himself the extraordinary consequences: cancer; infectious disease; genetic disease; Alzheimer’s; biological freedom; lifespan.
And then — crucially — he turns those abstractions into people. One of the most effective passages imagines today’s children growing up in a world where disease sounds to them as distant as smallpox, scurvy or bubonic plague sound to us.
That’s storytelling.
Concept → evidence → mechanism → consequence → human life.
The argument moves from head to heart.
6. His language is deliberately anti-hype.
This is fascinating because the subject matter is spectacular. He could write: AI WILL CHANGE EVERYTHING.
Instead you repeatedly encounter:
“my guess…”
“I think…”
“seems…”
“may…”
“probably…”
“could…”
“there’s a good chance…”
“I’m somewhat skeptical…”
“success is not guaranteed…”
Normally we’d warn writers against hedging. Here, the hedging is doing rhetorical work.
Because the claims are loud, the voice needs to be quiet. That’s a wonderful principle for serious writing: The bigger your idea, the less you may need to shout. Amodei allows the implications to supply the drama.
7. He argues against himself.
Another hallmark of excellent intellectual writing. Amodei repeatedly constructs the strongest objection to his own argument before answering it. AI can’t overcome physical experimental times.
AI can’t manufacture missing data. Biology is intrinsically complex. Regulation slows deployment. Human institutions resist change. Corruption exists. Economic systems are enormously complicated.
Dario even cautions that intelligence “isn’t magic fairy dust.” That colloquial phrase is wonderful precisely because it punctures the grandeur of everything around it.
This technique makes the essay feel dialectical rather than declarative:
Thesis.
Objection.
Qualification.
Evidence.
Revised thesis.
The reader keeps thinking: Yes, but what about… and Amodei keeps replying: Quite. Here’s how I’ve thought about that. He's trying to create trust with the reader.
8. Notice the alternation between enormous and ordinary language.
This is one of the subtler stylistic pleasures. The essay contains language like:
“marginal returns to intelligence”
“intrinsic complexity”
“factors of production”
“economic development”
and then suddenly:
“magic fairy dust”
“touch grass”
“the emperor has no clothes”
“pudding does not unstir.”
That tonal variation matters. Without it, 14,000 words of serious speculation could become unbearable. The colloquialisms act almost like little windows being opened in a stuffy lecture theatre. They also humanise the author.
9. The essay gets progressively more human.
This is where I think the storytelling architecture becomes especially interesting. Look at the five major territories:
Biology and health
Neuroscience and mind
Economic development and poverty
Peace and governance
Work and meaning
That’s not an arbitrary list. It expands outward and inward simultaneously: Body > Mind > Society > Civilisation > Meaning.
So underneath all the science and economics lies an increasingly philosophical question: What would all this actually mean for being human? The essay therefore isn’t really about AI. It’s about human flourishing. AI is the inciting technology. Humanity is the protagonist.
10. And then the ending changes register completely.
This is beautiful. After thousands upon thousands of words of analysis, qualifications, economics, biology, neuroscience, statistics and predictions, Amodei ends somewhere almost spiritual. He moves toward fairness, cooperation, curiosity, autonomy, children, disease, freedom and the possibility of humanity becoming better.
Then the long essay suddenly contracts into two remarkably simple sentences:
“Nevertheless, it is a thing of transcendent beauty. We have the opportunity to play some small role in making it real.”
Did you notice what disappeared? The data. The technical language. The economic terminology. The AI jargon. Even much of the argument. After 14,000 words of logos, he permits himself pathos. And because he has earned it, it doesn’t feel sentimental.
That is superb rhetorical timing.
7 Writing Principles you can learn from Dario:
Begin with a contradiction. Challenge what the reader thinks they know about you or the subject.
Earn the right to make a big claim. Establish credibility before reaching for grandeur.
Show your thinking. Let readers experience the reasoning rather than merely receiving the conclusion.
Give abstraction a picture. “A country of geniuses in a datacenter” is worth paragraphs of explanation.
Argue with yourself. Intelligent readers trust writers who acknowledge the strongest case against themselves.
Make the voice quieter as the claim gets louder. Restraint creates credibility.
Save your most emotional language until you’ve earned it. Logos builds the runway; pathos provides the lift.
Great writing is evidence of thought.
That may be the deepest reason this essay feels different from so much AI-generated prose.
Its value isn’t that Amodei found more elegant synonyms, varied his sentence lengths or inserted better metaphors. Those are surface characteristics. You can see a mind making distinctions.
Dario distinguishes intelligence from other factors of production. Optimism from inevitability. Technological possibility from deployment. Discovery from distribution. Physical constraints from intellectual ones. What AI can do from what society should allow it to do.
That is what gives the prose its authority.
The danger of letting AI do our writing isn’t simply that our writing gets worse.
It’s that we stop doing the thinking that good writing requires. Ironically, one of the landmark essays of the AI age is a rather magnificent demonstration of why human writing still matters.
I think Machines of Loving Grace absolutely deserves a place alongside Feynman, Sagan, Carson, Turing and the other pieces I admire—not necessarily because it contains the most beautiful sentences, but because it demonstrates something slightly different and equally valuable: how a serious mind makes an unbelievable future believable.
