Elon Musk

The episode opens on a birth, and on the accident of timing that will follow the child for the rest of his life. Musk's son arrived in early May 2020, in the middle of a global shutdown, and the host frames it in the only way that makes sense: this is a person who will always know that he was born into an anomaly.

You will never forget what is going on in the world when you think about when your child is born.

The birthday itself became a joke, and the joke is the kind Musk enjoys: the fourth of May, and therefore a Star Wars pun that the child will hear for decades.

May the 4th be with him.

The name is not a placeholder, and the division of labour in choosing it is precise. His partner supplied most of it; he supplied the tail end, and the tail end is aeronautical.

Archangel 12, the precursor to the SR-71, coolest plane ever.

Asked whether fatherhood feels different at this age, Musk gives an answer that is more interesting than the standard one: it is not that the experience is new, it is that he is now equipped to notice it.

I think it's better being older and having a kid.

He describes the change in perception rather than the change in circumstance — the way other people's children stop being other people's children once you have your own.

I think of them as these little love packages.

Then the conversation takes the turn that makes this a Musk interview rather than a celebrity one. He has spent years on neural networks, and he cannot watch an infant without seeing the same process in a different substrate: a system that is learning extremely fast, in front of him.

you can sort of see it learning very quickly

The host checks whether he is speaking metaphorically. He is not, quite. The vocabulary of machine learning was borrowed from biology, not the other way round, and he makes the point with a phrase that will not appear in any textbook.

humans are the you know original gangsta

The engineering question underneath the whole episode is whether artificial intelligence is an imitation of a brain or an alternative to one, and the answer Musk gives is a distinction between principles and mechanics.

we don't make a submarine swim like a fish, but we take the principles of hydrodynamics and apply them to a submarine

The architecture, though, is not an analogy at all. Layers of neurons, intermediate steps, back propagation — these are not metaphors for what a brain does, they are descriptions of it, transposed into silicon and trained on data instead of experience.

It's having the multiple layers of neurons and back propagation.

The same distinction governs flight. A bird flaps; an aircraft does not; both generate lift from a wing. The mechanism is shared and the implementation is not, which is the most compact version of his whole position on machine cognition.

just like an aircraft does not fly like a bird

The interview then turns to the decision that had been generating headlines: Musk was disposing of his houses and most of his possessions. The explanation is not asceticism. It is risk management.

Then they're kind of an attack vector.

The logic is that visible property is leverage — something to be pointed at, argued about, used against you. And there is a second, quieter reason: he owned a collection of houses he did not spend time in, which is a poor use of assets that somebody else could be enjoying.

He is not going to be living in a tent. A couple of Teslas stay, on the practical grounds that he has to test the product, and anything with sentimental value stays. Everything else goes, and if he misses it, the money from the sale can buy a replacement.

What he resents is the vocabulary. The word for what he is had become an insult, and he thinks the insult is mostly incoherent — because it conflates two entirely different activities.

billionaire has become a pejorative

The distinction he wants drawn is between spending and allocating. A fortune accumulated by building a company that produces better products is a claim on capital, and the claim exists because the market judged the products. The person who holds it is deciding where capital goes next — not consuming it.

there's a conflation of consumption and capital allocation

His example is Warren Buffett, described with a candour that is not quite hostile: a man he is not a fan of, doing a job he finds boring, reading annual reports to decide whether Coca-Cola or Pepsi deserves more capital. Boring, and also the mechanism by which an economy sorts winners from losers.

I'm not his biggest fan

The sorting rule is simple enough to state in a sentence, which is why he states it in one.

If a company is making compelling products and services, it should get more capital.

The host then draws the line Musk is willing to accept: someone who gets rich designing and engineering a product is not doing the same thing as someone who gets rich moving money around, and lumping them together is a category error that happens to be politically convenient.

From there he makes an argument that is unusual for a billionaire and awkward for his own class: the United States allocates too much of its talent to finance and to law, and not enough to making things.

there's an over-allocation of talent in finance and law

He frames it as both compliment and criticism, which is the only honest way to frame it — the people are smart, and that is precisely the waste. The prescription is arithmetic: fewer lawyers, fewer financiers, more engineers.

And he notes the status inversion that has accompanied it. Manufacturing was once a respected career in America; it is now often treated as a fallback, which he thinks is a mistake both economically and culturally.

manufacturing used to be highly valued in the United States

The pandemic, and the supply chains it exposed, has begun to correct that perception faster than any argument could.

Somebody's got to do the real work.

The houses turn out to have biographies. He owned Gene Wilder's, across the road from his own, and the reason was preservation: it was going to be sold, demolished, and replaced by a three-year construction site.

it was going to get sold and then torn down

He describes the house the way a fan would — handmade cabinets in odd shapes, doors to nowhere, strange corridors, tunnels, odd paintings — and he attached a condition to the sale: whoever buys it does not get to ruin it.

Don't you have Gene Wilder's house?

He accepts that the condition will cost him money on the transaction and does not care. The house is a person's residue, and he thinks the residue is worth a discount.

He'd want his soul.

The other purchases were less sentimental. Once people began climbing over his walls, he bought the neighbouring properties as a buffer, and then began collecting small old houses on the ridge above Sunset with a vague plan to build something on the site.

which is got a great view

Asked what a dream house looks like for him, he answers immediately and without dignity: something out of Tony Stark, with a dome that opens and a stealth helicopter.

some tony stark type shit

The plan died on a comparison. Designing a house would absorb him completely — he says he would get obsessive about the details — and the same hours could go into the rockets. Framed that way, the choice is not close.

Or should I be allocating that time to getting us to Mars?

He is asked whether that is genuinely how his mind works, and he confirms it without embarrassment. The scarcity is not money but attention, and attention has an opportunity cost measured in orbit.

He then offers an unusually concrete accounting of where the attention goes: SpaceX and Tesla take almost all of it, the Boring Company takes less than one percent, and Neuralink takes something like five.

Boring Company does not take less than 1% of brain cycles.

The Neuralink segment begins with a status report. No human trials yet, but the estimate has tightened: a first implant in a person is plausible within a year.