aleph
IV x MCXIIIV = IVXXXX

aleph is a research lab working on new interfaces for the brain. we believe in speed and scale; we're using ultrasound to interact with the brain. we think our telepathic future is both imminent and wonderful.

WORLD'S FIRST IMAGE OF BRAIN CORTICAL VACULATURE. NONINVASIVE ULTRASOUND
CONTENTS
I.

Human, but more.

We currently have access to only a fraction of what our minds are capable of.

Have you ever tried multiplying roman numerals? It's incredibly, ridiculously difficult. That's why, before the 14th century, everyone thought that multiplication was an incredibly difficult concept, and only for the mathematical elite. Then arabic numerals came along, with their nice place values, and we discovered that even seven-year-olds can handle multiplication just fine. There was nothing difficult about the concept of multiplication—the problem was that numbers, at the time, had a bad user interface.

— Bret Victor

Consider that everything in the world is currently at the stage of roman numerals. Very few things are truly hard, but the act of doing or imagining anything is the act of wading through molasses.

Making a film? Roman numerals. Understanding optics? Roman numerals. So it goes with photoshop, drawing, aerospace engineering, radar, mechanical design, etc. People want to make films, not drag files.

The product

The highest realization of an interface is a direct connection to the mind—telepathy. Telepathy consists in two things: the mapping of our ideas, intentions, and mental activities to the correct external action, where the “action” captures the full fidelity of the idea (mind-to-world), and the mapping of those ideas in full fidelity to another person's mind (mind-to-mind). We'll start with the mind-to-world interface, and then build a mind-to-mind interface. Our mind-to-world interface will enable astonishing experiences:

“I can't believe we used to press buttons.”

— Software Engineer, United States, 2028

In this regime, ordering an Uber is as primitive in experience as moving your index finger; the distance between asking a question and receiving an answer approaches the speed at which you can access your own memories; the connection between you and your tools is so seamless that it is as if you can hold a whole codebase, genome, dataset, company, etc. in your head.

The net effect is the arabic numeralization of everything. the world opens up.

II.

Accelerando.

It is possible now.

Is it possible to achieve telepathy now, and not in decades? Does the answer consist in placing electrodes next to every neuron or genetically modifying brains? How does one conceive of solving the problem of telepathy?

We think this problem is newly tractable.

  1. 1.New modalities of imaging the brain are possible with ultrasound
    1. a.ultrasound can achieve mm-resolution
    2. b.full-brain coverage
    3. c.information can be obtained without surgery
  2. 2.Scaling brain datasets allows astonishing decoding of thoughts

There has not been a single real neurotech product so far. But in recent years the first demos of reading out something interesting from the brain have started to appear, including the neuralink cursor control demos. In our opinion, the most salient results are the MindEye and semantic reconstruction of language papers.

These studies make us believe that decoding actual thoughts with high fidelity based on brain blood flow data is possible.

Ultrasound is a technology that is fundamentally capable of even better resolution than fMRI. We got unbelievably lucky in that ultrasound also happened to be something you can put on a chip inside a device that fits in your pocket.

Left: image shown to participant, right: predicted image from fMRI reconstruction

This was done by training on a dataset of 40 hours of people looking at images. Importantly, the fineness of the reconstructed images improves with increasing amounts of data, and continues to improve.

We see this as a reason to place uncomfortably large bets on data. Recall—the field of ML first started showing interesting behavior on datasets like MNIST and ImageNet. Scaling the models to a dataset the size of the internet led to LLMs.

MindEye is a large dataset from the perspective of neuroscientists, but we're nothing compared to the size of the internet. We think astonishing things will happen when you scale up brain datasets.

To get here, you need high-quality data. It needs to be 1. mm-resolution and 2. cover the whole brain. Ultrasound imaging of the brain allows us to get there today.