Innerphases is building human-native communication devices that interface directly with the nervous system

really scary competitive place and I had to have some kind of heristic by which I should get better. And so I thought to myself, all right, every two years I want to double in the metrics I care about, which at fourth grade was coolness.

And so at that point growing up in the South Bay, I grew up to exceptionally chill parents, an opera singer and an electrical engineer who were just like, "Oh, son, just do whatever you want to do. " And I was like, "Okay, every two years I will double in coolness.

" And so that's exactly kind of the how I structured my life. And in the next two years, I did stuff like tried to memorize the dictionary because this is what I thought cool was when I was like 10. Eventually the coolness actually went in a direction of like things that were actually pretty cool.

Like when I was in high school, I started working at chess. com and I like delivered tons and tons of features and made some really cool products that were really impactful to me. And when I was in college, I was like, "Okay, what's the coolest problem you could possibly work on?

" And I ended up working at Neuralink where I'm currently seated. Uh sorry to any co-workers who are seeing me taking this I don't never apologize. But what about uh what about like joining a punk band, learning to do a kick flip? Did that ever come into the cool kid? I do all these things. I like guitar. I Oh, nice.

There we go. Psychology. So, I have a lot of mushrooms memorized throughout the Sier Nevada. In general, these are the things that fit my heristic Okay. Okay. Yeah. Yeah. Makes sense. But not the cliche cool guy like like you're not you're not driving a couple years. You're just level you're just leveling up.

You're maxing out your character. Yeah. the the the red Ferrari convertible that comes next year when there's another doubling in the cool factor. The doubling the doubling people are going to start tracking the the doubling like the Bitcoin the happenings. I want to make a poly market is the real thing.

I can stay on track. This is amazing. So So you you've been working at uh Neuralink. What are you working on now? Yeah. So at Neuralink I've been doing all kinds of really cool things and one of these things I've been doing most recently is I'm leading the sound speech project essentially.

Uh you've seen the study that was announced fairly recently. So I can talk about this a little bit, but basically there's a really great corpus interesting literature that goes from how do you get from inside your head to outside your head. This is super super profound technology.

Uh imagine if you have ALS and you're locked in or something like this. I spent a lot of time around these people and every minute I spend with them I think more to myself that like these people need to be able to communicate to the outside world better. Yeah.

Um and Neurink is doing a really great job of this kind of thing. I have so much confidence in the team. I have so much confidence that this technology is going to be transformative. But also, I really hope I don't get ALS anytime soon.

I really hope I have a brain stem stroke or something that keeps me from being able to communicate sely. And I still am interested in the way this technology will go on to transform other parts of the world. Sure. And so with interfaces, what I'm looking at is I see all these kind of gizmos for AI.

I see these like limitless pendants. I see these uh smartwatches and things that are trying to kind of integrate all this data. And I I don't fully buy it. Uh, the reason why I don't is because I actually don't like voice as a medium that much for most things.

Uh, if it actually wants to integrate into your life effortlessly, I don't want to be in the middle of conversation with you and be like, "What's the guy on the left's name? It's super awkward. It's terrible. " So, how do you do better? Um, and there's a couple ways you can think about this.

One, uh, former co-orker of mine and a guy really respect, Bliss, who you may have seen on Twitter or whatever, told me that there's a couple ways you can get value out of machine learning. And one of them is kind of arbitrage via data. And another is arbitrage via time.

And so if you have these longer time series, you can kind of get more bits out of them by just increasing your confidence intervals either by doing these kind of uh causal estimates or a causal estimates where you have understanding of what's going on in the future as well as in the past.

And so you can kind of use this to do s updates or just by looking for signals that are kind of strong over a long period of time.

And I think that with this in mind, uh, there's kind of room for this halfway horse between like a conventional microphone which can pick up stuff in real time and do transcription in real time and a like an EA or someone who actually will kind of try to predict what the heck you're doing in space.

And so you can imagine is building a device that basically takes these like fundamentally pretty crappy signals like MEG signals for example. Uh, what are those? Sorry. Any G signals? Yeah. Yeah.

Uh there so there's many ways in which signals were propagated through the body to move your arms or to create electrical activity. Sure. You may have seen a recent paper out of Reality Labs in Meta where they use exactly this to kind of uh for the Orion uh headset. They bought that company.

I forget what it's called, but they bought a company like six years ago and they're finally implementing it. Right. Yeah. And it's extremely cool. A friend of mine went to one of the clinical trials and put the thing on and you try it out. You say, "Look, I'm pinching my fingers.

" and the fingers pinch and it's it's pretty magical. Uh this is the kind of thing which normally you would have to cut a hole in your head to do but you don't necessarily have to do here. Sure. And there's a really good story to be told about kind of doing this for speech and getting really terrible speech.

Uh I have no illusions as a neurochnologist that you will get really terrible speech out of this. But the key thing is that this does not matter because the context is what matters.

And by informing this really awful speech on the context of what's around you and like your intention of the situation, you may be able to actually get the underlying part of it that matters.

And so in fact, indeed, even if I can only kind of pick up the word left and guy, I I might know that I want to know who's the guy on the left. Uh and and that's an easy LLM transformer problem, right? So you get a bug, big bag of words.

It's very messy, but it's enough to turn into a semblance of a thought that could be confirmed with a pinch maybe. Yeah. Yeah. Now, ideally, at least in my eyes, you have this just totally effortlessly against your life. And you kind of collect this data all the time via a method that's fairly subtle.

Uh I have a little bit of alpha here that I won't spill. So, you'll have to have me on in a couple months to to see my my fun demos, but uh of how to get this data. So you're constantly collecting this and it's constantly informing and improving uh on how you interact with the world in general.

I really care about efficiency in whatever I do. And lots of little things in the in your day-to-day life are super inefficient in these pretty funny ways. And I think that by kind of having a this compute that is always acting in a way that doesn't invade the privacy of other people.

Like I want to understand myself better. I want to be able to improve myself. I want to be able to like keep on that doubling path u without like constantly listening to to you guys or to my friends or to the co-workers or whoever else is around me and I want to have the data set to do that.

So there's this kind of dual product thing of like okay as you create the better interface to interact with LMS and to build the like higher bit rate communication interfaces and also get a better model for oneself.

I had this idea in college that uh got my wisdom teeth out and I was like I want them to put a microphone in the hole there because then I could say I could whisper who's the guy on the left and it wouldn't be picked up to anyone else but it'd be able to pick up the the audio.

But I I think that's probably gross and uh kind of a part of the tech tree we don't even need to go down because we can just leapfrog that whole step. We've already gone down it. I had to break it to you. This there's this Israeli company called Muller Mic that does exactly this for application. No way. Yeah.

You can imagine that like you have bombs going off around you and you're like, "Boss, we need backup, but you can't hear that over the crazy artillery. " Mer good company. Did they do it? I have no idea. But Ben's always telling our producers always telling us to to move the mic. Yeah. Talk closer to the mic.

Closer to the mic. But I could just have a Muller mic. Yeah, we we need to be augmented for sure. Um what what uh are you going into a similar uh FDA pathway as Neurolink? Is that uh how things are going to play out for you? Yeah. So I'd like to avoid this if possible.

I do think that there's room for a device that's somewhere in between like this really groundbreaking first-in-class biomedical device which involves a lot of risk. I mean fundamentally invasive neurosurgery involves a lot of risk.

Even when you're doing it with the very best surgeons in the world with these robots that can vascular with everything there's still some risk price in there. And so I think again there's this halfway horse thing of like okay maybe it's invasive but maybe it's not quite uh as risky of a thing.

Maybe it's more like getting like a an implant for birth control or more like getting like simple surgery to remove a mole. Uh these kind of things that are much more everyday and much sooner.

I I'm very bullish on their link in the long run and I think that you should keep an eye on the amazing stuff folks are up to here. But yeah, if DJ is in the office, tell him I say hi. We'd love to have him on the show. I will. I love DJ. He's great. Uh he's fantastic. Uh you've been fantastic.

Thanks so much for stopping by. What uh I got to ask before you go. You're you're you're skillmaxed to the gills already, but uh what what what's next? What are you uh besides the new company? Uh you know, what are what are some areas that you're trying to level up?

Now, believe it or not, the next thing I'm doing next week is I'm making a trumpet from first principles. I'm going to I have a physics textbook and I want to like without any reference try to figure out how far apart all the valves should be.

I want to cast in my backyard with a big crucible and I want to like play this thing which probably won't work. Uh, and so I'll get better at metal working, I'll get better at physics, I'll get better at all these little stupid things I care about.

And it's a pretty fun little interstitial bit before I really put my head down and get back to the the great doubling. On the show, play the trumpet live. I want to hear the demo. We want to hear no good. I have to warn you, but I'll be better. If it if it in two years, you'll be good. If it makes sound, it works.

That's fantastic. Great question, Jordan. Awesome. All right. Great to meet you. Congratulations. Later. Byebye. Fantastic. We'll talk to you soon. Next up, we got Colton coming in this the studio. Uh, welcome to Colton as soon as he hops in here. The idea of just trying to get twice as cool. It's so good.

It's constantly and just never being like, "Oh, I'm cool enough, you know, I'm cool enough. " No, double the Great doubling. I mean, he he doesn't look like, you know, I don't know if he can have champagne yet, but you know, we we followed a sim similar principle around the great double influential growth.

It's the most powerful force in the world. Something that you can understand unless you've experienced it. We experienced it going from a thousand followers to 2,000 followers, you know. Yeah. Imagine that at trumpet manufacturing. Fantastic. Uh well, we have our next guest in the studio. Welcome to the stream, Colton.

How are you doing? Welcome. Good. How are you guys doing? Can you hear me? Yeah. Yeah, we can hear you. Uh would you mind kicking it off with a little introduction on yourself and the company? Yeah, sure. My name is Colton.

Um, I'm from Philadelphia, but right now I'm in Boston working on my company called Orbit, which I think you guys actually got a brief introduction to because my teammate Stephen was on here, I think, just an hour or two ago. That's right. Yeah, there we go. The double the double teal fellow.

So, combo, give us your side of the story. What does he get wrong about orbit? All right. No, just just just unpack it again a little bit and then we'll we'll we'll take the conversation in a different direction. Yeah, absolutely. So, I can tell you how I first heard about this idea and got involved with it. Yeah.

And how did you meet him? I'd love to know that story. That's interesting. Yeah. So, Stephen went to Georgetown and he was looking for an engineer to help him build this narrow techch. Um, but there's no engineers at Georgetown at all. They don't have an engineering department.