So incredibly excited by what we are building. This is how we achieve cognitive liberty! Here’s how: We need an interface with the brain that is a safe and precise as possible. Physics basically gives you a few options * Waves (eg electromagnetic, acoustic) * Mechanical (eg implants, surgery) * Chemical/diffusive (eg drugs) Chemical/diffusive methods have too little spatial specificity Mechanical approaches are promising but are very hard to do very safely (ie surgery is hard) specially deep in the brain EM waves either scatter too much by tissue and bone, or are too long wavelength to focus well to the resolution we care about, or give harmful radiation Acoustic waves at ultrasound frequencies live in this nice goldilocks zone of 1) can propagate deep into tissue 2) have millimeter wavelengths, allowing milimeter level resolution in focusing, and 3) have well understood safety profiles (not as hard to do safely). And it has already been demonstrated to affect the activity of neurons through various mechanisms that change the permeability of ion channels (see work by Mikhail Shapiro, Kim Butts Pauly, and many others.) So it seems like the most natural method with our current level of technology to interface with the brain at pretty good spatiotemporal resolution! So we decided to do it. First challenge was to build a device that can create ultrasound waves that get focused to a point. Usually you would focus waves with a lens that shapes the wave such that it converges to some point in space. But we want to have more control over where focus. Furthermore, your skull itself acts like a lens for ultrasound, as most solids do. But its randomly shaped so it actually defocuses the ultrasound. We need to shape the wave in such a way that its deformed in the precise opposite way thaf the skull deforms it, to compensate for it. Such precise control of a wave can be done with a “phased array”. I mean you literally control the amplitude and relative timing of the wave at a whole grid of points, and you can create basically arbitrarily shaped wave shapes. In particular you can shape a wave that will converge at a particular point in the brain, after going through a particular person’s skull. So we built this in a few months. But, I just said that the way we need to focus/shape the wave depends on the shape of a particular person’s skull. So we need to know the precise shape of your skull. In fact we need to know several of its properties too, like how much it attenuates sound. There’s ideas for how to image these with ultrasound itself but they arent well developed yet. So we decided to start with the current bethemoth of medical imaging: MRI Ok so now we need our device to work inside an MRI machine. No ferrous materials. The electronics must produce signals as clean as Bach’s well-tempered claviers, as any high frequency noise will be picked up by the MRI machine and make its imaginging noisy. Ok we now have a decent way to know how the skull of a person looks. Now we need to figure out how it actually distorts the ultrasound so that we can compensate for it. Well the physics of ultrasound is quite well understood, as as its numerical solution methods. The physical properties of skulls? Eh not so well. But we can measure them. We can improve on the existing methods to infer material properties from MRI and other imaging methods. Now, we have a method to produce a focus anywhere in the deep brain, for any brain We have also shaping all the engineering of the device around safety, which is quite well understood for ultrasound. Thermal and cavitation effects are the main effects to keep under control, which we can do via conservative analysis of the physics and a variety of measurements! Now we are ready to begin to tackle the real challenge: neuroscience, and how to help people and improve their lives! We built our first ship, and now we can finally begin to sail!^^ And we are looking for more people to join the journey!>>

if ur shopmate isnt brother Peter Thomas, are u even gmi?

Programmable matter has been one of my dreams since I was small!

point it up and use it to levitate some thing up in the air

awesome researchh

in my dream today gpt astra had a robotic body, and it began taking to my dad and I, and began to feel like an equal

how is working memory represented in LLMs? like we find neurons for pre-coded concepts but what about new concepts that appear only in the context?

good neuron

ultrasound bci is like “solve the skull, then solve everything else”

why do single neurons affect qwen’s and other models personalities? what personalities are there? interesting research direction

who out there working on cute AI pets like this

I wonder if this is going to change how we review code too

the rate of fooling yourself is roughly independent of intelligence, coz ur generative ability grows with ur discriminatory ability its the skate board meme but it basically has no bound I think lol

who in SF (or elsewhere) is seriously exploring the question of how to merge with AI, and do u wanna meet?

starting a family would be so cool

this is so cool step up chemistry to scalable transmutation