Noland Arbaugh

Rogan opens with a prediction rather than a question: if anyone ever makes a film about how the world changed in 2024, Noland Arbaugh will be in it. Arbaugh, who is the first human recipient of a Neuralink implant, agrees without much ceremony — "I guess so. No, definitely." The exchange sets the register for the next two hours, in which an event of obvious historical weight is discussed by the person it happened to in the tone of a man describing his own hardware.

I have a feeling if there's a movie that they do in the future of how the world changed in 2024, you're going to be in that movie

The joke that follows — that by then the film will be generated rather than acted, and that 'The Great Gatsby' could be run through a video model to produce "the most amazing version" of itself — is the first of many places where the conversation about a brain implant detours into the general strangeness of the moment.

Arbaugh is careful to place Neuralink inside a history. Brain-computer interfaces, he notes, are not new: the Utah array, Synchron's stent-like device threaded up through an artery in the neck and expanded inside a vein, decades of work going back at least to the 1990s. His formulation is conventional and sincere — "we're standing on the shoulders of giants."

we're standing on the shoulders of giants

What he thinks Neuralink adds is scale rather than novelty. The device is "in a league of its own," the name attached to it guarantees attention, and the decision to open-source the work means the whole field gets pulled forward: "I think the whole field is just going to grow exponentially at this point."

It's in a league of its own

Rogan asks about the first such implant and Arbaugh guesses 1998, describing the Utah array as a chip with fixed threads sitting on the brain, reading a section of motor cortex after an open-brain surgery. The conversation then moves to work that tries to reconstruct what a person is seeing or dreaming from scans alone. Jamie pulls up a headline about Japanese researchers recovering dream imagery at around sixty per cent accuracy.

Rogan's reaction is the layman's, and it is the right one: the received wisdom is that nobody knows why we sleep or what dreams are, and here is a lab rendering them. Arbaugh, who has an obvious professional stake in the answer, is more interested than surprised. His own line — "I've heard plenty of people say, Yeah, we still don't know why you even need to sleep" — treats the ignorance as stable background rather than a scandal.

I've always heard that scientists really don't know what dreams are

From dream decoding Rogan moves to the simulation hypothesis, and his argument is not metaphysical but probabilistic and cumulative. The idea seemed ridiculous five years ago, seems less so now, and will in his estimate seem likely in five more — not because of any single proof but because every technology, every quantum result, every increment of understanding makes the whole thing "somehow another real, but not real at the same time."

Neither a simulation nor actual reality, a hybrid of these things

Arbaugh does not argue. He redirects to what he actually knows, which is that the implant is generating an unprecedented stream of data about a human brain: "how much we're going to learn just about the brain from this." The most concrete example is also the most embarrassing for the field, and he tells it without rancour.

how much we're going to learn just about the brain from this

The technical account is plain. Sixty-four threads were implanted, each carrying sixteen electrodes; over the first month many of them retracted, and the signal from those declined. The threads are finer than a human hair, which is the reason nobody could simply look: a scan capable of resolving them "would probably just fry my brain."

there are 64 threads implanted in my brain with 16 electrodes on them each

So the evidence that they were moving was indirect — whether the electrodes were still picking up neuron spikes — and performance declined for a while until it was addressed in software. The explanation, or the leading one, is the part that stunned him: the brain moves more than the designers had assumed.

the brain moves more than they thought it would

His first reaction was disbelief at the ignorance rather than at the fact. "You guys don't know how much the brain moves?" He had not known it moved at all. The brain pulses with every heartbeat because blood is running through it, and the design had been built for roughly a millimetre of excursion.

In his case the measurement came back at three millimetres — three times the tolerance the device and its threads were engineered for. Whether that is unusual is exactly what a single patient cannot answer, and he says so: with more participants they will know whether his brain is an outlier or whether a range exists at all.

they found in my brain that it was moving three millimeters pulsing

Rogan asks whether a faster heart rate means a faster-pulsing brain, and Arbaugh confirms it from the inside: high blood pressure gives him violent headaches. The exchange is a reminder that the data being used to tune an implant is also, for the patient, a body that hurts.

If I have really high blood pressure, my head just gets really, really

The mechanism, as he describes it, is unglamorous: a Bluetooth link, a Neuralink app on the computer, electrodes reading neuron spikes from motor cortex. What makes it work is that the intention survived the injury. He cannot move his hand, but the neurons that would have moved it are still firing; the signal simply has nowhere to go.

There's just a cut off in my spinal cord

A machine-learning layer learns the mapping between that intention and cursor movement, and the crucial consequence is that the mapping is arbitrary. Any motor action can be assigned to any output: a foot kick could move the cursor left, and the on-screen prompt could as easily say "do a little jig."

It could be do a little jig, and that'll move it to the left

Rogan admits he had assumed the control was direct — that Arbaugh simply tells the cursor where to go. The distinction Arbaugh draws in reply is the most interesting technical point in the first hour: attempted movements and imagined movements are different, and he began with the former, deliberately trying to move a hand or lift a finger so the algorithm had something legible to learn.

There are what are called attempted movements and imagined movements

A few weeks in, he tried simply thinking the cursor to a position, with no motor attempt at all, and it moved. That one worked did not surprise him — the signals were intact, only the cord was damaged. That it worked without the attempt did: "I was gitty the entire day."