Rob Toews on brain-computer interfaces: from the Utah array to Neuralink and the noninvasive frontier

And our next guest is in the reream rating room. Rob from Radical Ventures coming in to the studio. How are you doing, Rob? Good to see you. Rob. Hey guys. Great to be here. Thanks for having me. Welcome back. Uh what's uh I want to go through BCI. want to go through the latest Forbes piece.

Uh what I mean I feel like you've spent what years at this point writing about AI. Are you just over AI? Is AI done and you're just bored of that and you got to move on? You got to pivot. If you're in AI, pivot to BCI. What's going on? I there there is definitely a lot of a lot of noise in the world of AI right now.

I I think BCI and AI are actually closely related. Um, and BCI will be kind of one of the next really important foundational technologies that impacts how AI gets rolled out. But I do think a big part of our jobs as VCs is to kind of think about opportunities and technologies that are not yet obvious and mainstream.

And I think VCI falls into that category in the sense that it's not ready for prime time yet, but I think it's getting closer. And and I do think it will have a really big impact on how basically human intelligence and artificial intelligence fit together. the merge.

The merge uh it feels like one of those technologies that's always 10 years away like fusion. Uh but take me through the history. Uh what when does this whole story start? Uh there's the Utah array. I want to know about that. Yeah, totally. So BCI is not not a new technology.

Um it's something that's been around in a research context for decades. The very first BCI was actually um in a human was demonstrated in the 70s in the early '7s.

um researchers at UCLA basically using EEG which is non-invasive technology were able to get patients to be able to control a cursor on a computer screen just with their thoughts. That's kind of like what we're seeing from Neurolink, right? It's like not that far off but it's been 55 years now.

So we're super interested like what happened in the intervening years. Yeah. Yeah. Exactly. So that so that was like a very basic unreliable prototype, but it was kind of a concept.

The first invasive BCI and and maybe just as a quick like yeah uh quick definition the difference between invasive and non-invasive these are kind of the two main categories of BCI technology. Invasive technology requires surgery. It involves putting electronics like inside your skull directly on or in your brain.

And so Neuralink is an example of an invasive BCI company where that you know you have to cut open your skull doing using a surgery called a cranottomy um and put the chip inside. non-invasive technology does not require surgery. It's just sensors that you can put outside your head.

You know, they could be embedded in like a consumer device like headphones or a hat or something like that.

So, the trade-off between invasive and non-invasive BCI really does boil down to a trade-off between like accessibility on the one hand and ease of use versus signal quality and like how high fidelity of signal you can get from the brain. So, the the Yeah. And and it's a lot about like value, right?

Like if you if you have a disability and you cannot use a computer and you and somebody says I can give you the surgery and you're going to be able to interact with the computer. That's a great trade. Whereas if you say that to me right now I'm going to say oh wait, you know, I'll use the mouse and keyboard actually.

I'll be fine with the I'll stick to the mouse and keyboard. Yeah. So okay. Yeah. I think my my big question is like at at what point is is the invasive approach like so tried and true everyday people.

So so in the 1970s it was it was non-invasive when did the first invasive stuff start happening because Neurolink like popularized it but didn't invent it. Right. Exactly. Yeah.

So invasive BCI the first invasive BCI was implanted in a human in the 1990s and it was this device called the Utah array that you referenced John. And honestly, I would encourage people listening to like go to Google images and look up Utah, right? Just to see a picture of it.

Like it looks like a medieval torture device. It does. It's so scary. It's literally like a bed of needles. Um, and you literally you like cut open the skull and you just jam this set of needles into the brain.

Um, and so it it it the the Utah rate to be clear was like a really important breakthrough in the history of BCI and it was the state-of-the-art for a long time and it it made it possible to do things like go from thought to text and like you know understand people's uh people's thoughts directly but brain damage basically like if you're if you're jamming these needles into your brain in order to record from some neurons you're also killing a bunch of other neurons and so that was that's the reason why the Utah rate never really like took off from a commercial perspective but it did lay the groundwork for subsequent generations of invasive BCI including Neurolink which obviously the most bas famous famous invasive BCI company.

Okay. So uh now uh was Elon the first person to really take BCI seriously because it feels like there's a proper market map and you have uh Sam's working on stuff Sam Alman founder of OpenAI and then Elon's working on stuff and then uh Nudge also exists.

There's like a there's like a number of like basically like billionaire founders that are all making serious bets.

Are all of them is this one is this like an Anderole type story where like you need a billionaire who's just going to sink a bunch of money into this or are there any companies that have like actually just gone from ground up in the traditional fashion of like they raised a seed round and then a series A like it just feels like a very different path.

So like maybe take me through the market map of what's happening right now. Yeah.

No, it definitely like started as from a startup perspective as a category where like it was largely funded by billionaires like obviously Elon had Neuralink, Bran Johnson had a BCI company called Colonel that he poured a ton of money into. Yeah.

Do do you remember like what that actual was that invasive non-invasive like what the plan was there? He he explored both. It ended up being non-invasive and they developed like a new modality of sensing non-invasively which is actually pretty cool.

you like shoot sensors through your skull, like shoot lasers through your skull to like uh detect the blood flow in your uh in your brain. Uh which is like is still used today.

Um and the company's still going, but um but yeah, it I mean it's worth noting Neuralink is like definitely the most well-known BCI company, but it wasn't the first. Like there were startups before Neuralink. There's one called Synchron out of Australia.

There's one called Paradromics um which did kind of follow that more typical path that you described of like raise a seed raise a series A and certainly in the past few years there there's been an explosion of them. So so now there are a ton of BCI companies both invasive and non-invasive.

What are the uh it feels like Neuralinks had a very like kind of like barbell strategy where they're where they'll tell like a sci-fi story about like you will be able to fully like communicate with AI and merge and whatnot, but then there's also like this ultra practical like if you are a parapillegic, you can use a mouse cursor and click and use a computer.

Uh are there are there other pitches that companies are are offering uh in terms of just like reading information from the brain or writing information to the brain? Like how how do the value props like stack out right now?

Yeah, I think one of the really interesting differentiators of the different invasive BCI companies is like what's the end use case that they're going for and neural what Neuralink has been focused on to date is computer use.

So like all the demos you see of neuralink patients like Nolan Arba the first patient is using a computer like which basically comes down to like mostly controlling a cursor and and potentially keyboard and there's obviously a ton you can do if you can control a computer. So that's like a very valuable use case.

Um but a like a harder more advanced use case which like I kind of think of as the holy grail for BC or one of the holy grails for BCI is decoding language.

like basically being able to instantaneously understand words from a person's brain and like write them or send them via text or email like that basically that basically is telepathy more or less especially if you then like convert it to another person's BCI like so it goes directly from brain to brain and so that is a more ambitious target which there are startups that are going after to this point at least publicly neuralink hasn't published anything related to language decoding of course it's something they're thinking about I'm Um but and but that I think that's one really interesting use case and then another one is being able to control objects like whether it's controlling a robotic arm or like controlling any sort of object in the real world which again it basically is telekinesis and like these terms sound really like um you know fantastical and paranormal but like that is essentially what the the capability that that will enable which is cool to think about.

Yeah. Are the non-invasive BCI companies ever targeting like higher level abstraction?

Like look, we're we're not going to re be able to reliably get an X and Y coordinate for a mouse, but we can tell you that you're stressed out or we can tell you that you're overly tired or we're tell you that you're dehydrated in your brain or something like that.

Something that's more abstract, but still uh something that a consumer could actually adopt, wear a hat, they're not going to go into surgery for it, but more of like a biohacker use case. Yeah, that that's exactly right. And those use cases have already been pretty well proven.

Like there are actually non-invasive BCI products on the market today that like can you know detect if you're stressed or if you're if you need to sleep more or you know if you need to focus better those sorts of things.

I think the the really big question like in my opinion the single most interesting strategic question in the field of BCI right now is like the conventional wisdom has been that invasive BCI is necessary for the most sophisticated use cases like language that we were talking about because you just need to get that high fidelity data but non-invasive approaches have been getting better and better very quickly in part because non-invasive sensors have been getting better but much more so because the underlying AI is just getting a lot more powerful which lets you extract a lot more signal from the same noise.

Sure. And so honestly no one really knows like where the performance ceiling will be for non-invasive BCI. There are there are companies out there today that whose vision is to like build a thought to text language decoding uh technology platform just using non-invasive sensors that you can just put on your head.

Um, it's not yet proven if that will work or not, but like there are definitely people aiming for it and and honestly I think no one no one can say for certain that it won't work.

I think like they're all kind of pursuing this scaling law strategy of like we saw with language models just collect more and more and more data, build bigger and bigger models and like they'll get better and better.

And so far like no one no one yet in the world has collected like a mass a truly massive data set of brain data paired with language data. So like who's to say whether you know what what might be possible if you really try to scale this stuff. Yeah.

How how do you how do you invest in this category right now if you're make money off of this? How do we make money off this? Um no but because the the probably my personal default would just be to try to pick up some Neurolink secondary. Yep.

and just ride with them because it feels it still feels like we're early and if you just fund a bunch of alt Neurolinks I think that that might be kind of a tough tough game but how do you think about it?

Yeah, Neurolink is definitely the like the most well-known the most wellunded obviously it's Elon Musk and it has you know everything that that comes with it in terms of being an Elon company. Um it is also I mean just thinking tactically from an investment perspective it's also extremely highly valued today.

Their most recent round is at $10 billion. Um there are a handful of very credible Neuralink competitors including a couple it could be undervalued. Yeah. Do we have any I'm just messing around. I mean do we have any idea on market size for these things? How do people even underwrite these types of deals?

Are they building up to like the number of parapolgics that need uh you know the the the FDA approved surgery for that indication or is it more of like a consumer uh like TAM? How do people think about that? Yeah, I think you basically have to think about it in two phases. The first phase is the medical phase.

The second phase is the like consumer/humans merge with AI in the singularity phase. Um and I think the TAM math is a lot more concrete for the medical phase. Sure. And just in the medical phase, it is a massive TM.

If you look at like basically if you just take everyone in the world that's paralyzed, that's like tens of millions of people. And you can break it down by different causes, ALS, spinal cord injury, etc. Um, you know, many, many millions of people.

Um, and the industry is kind of circling around this price point of 150k per implant with the expectation that insurance will cover it. Like that's not it. it will be something that um a lot of people will have access to because they won't have to pay out of pocket.

So if you just do the numbers on that, it's you know it can be a many many tens of billions of dollars market. Yeah.

Um, but I do think that like the really big vision, the big prize that ever, you know, that all that is gets people like Elon Musk and Brian Johnson excited is the longer term vision of this being like a consumer product for a more generalized population that lets anyone like seamlessly merge with digital intelligences and artificial intelligences and so forth.

And you know, if you if you think that eventually DCIs will be as widespread as iPhones or something, then obviously, you know, you can you can talk yourself into a huge dam. Imagine not having to open the Sora app, but just having a Sora implant and so you're just sitting there. Yeah. Just at the trough.

It'd be amazing. Are you tracking any of like the intermediate steps? Like uh there was that uh company that launched a demo that it seemed like it would uh put something on like your jawbone and it would be able to hear what you were whispering.

Um we tried the meta-array displays and those were able to pick up on a very low audio signal. You could kind of whisper and it would still be able to interpret it with AI into translation.

um uh like is that is that actually like a critical path in the tech tree or is that kind of just a side side note or do you think that that that whole tech wave will be uh like a big uh precursor? Yeah, there's there are a lot of different opinions on this.

So what you're describing people often use the term silent speech to describe the category which is exactly as you said like you have a sensor somewhere you know outside your your skull. It can be like as part of a headphone or like a a hat or something.

um that can detect the words that you're saying if you just kind of like barely whisper them or or even even if it's subaudible if you're just kind of like barely moving your lips and moving your tongue. Yeah.

And the idea is uh it lets you use all like all the value and possibilities of voice applications but without having to make noise.

And so like when you're in public for instance walking down the street or on a bus or something that like that can be a lot more useful than you know you probably don't want to speak out loud to your phone. Yeah.

Um, I think to your question of like is this just like an intermediate like branch of the tech tree or not, I I think it really depend like bullishness around silent speech I think really depends on how soon you think B true BCI will be ready for prime time because um silent speech isn't recording brain data directly.

It's recording like downstream muscle data. And so if you can actually decode what's coming straight from the brain in an accurate way, you can kind of just leaprog silent speech.

Um, but if you think that there are real limits to what's possible non-invasively with BCI, then silent speech might be like a more practicable path.

Um, and you mentioned this company alter ego out of MIT recently uh had this really buzzy launch video that like they frame it as basically telepathy which is a little misleading because it's not actually reading brain data. It's like the guy is like um mouthing something um and kind of moving his tongue.

But yeah, that is another maybe like uh more achievable and near-term version of this. Yeah. Uh Chris in the chat asked uh what about Chinese in infantry using BCI to reduce PTSD and increase reaction time? Do you know anything about that?

I I haven't heard about that specifically, but there's Yeah, there's there's no doubt in my mind.

I mean the the chi the CCP and Chinese uh startup ecosystem are investing a lot in BCI um in in the US also like one of the big use cases in general one of the big use cases is um like treating neuroscsychiatric diseases so if you can like precisely modulate brain activity you can address a lot of uh health conditions like PTSD but also like OCD depression Parkinson's etc.

It's basically a labbotomy that works, right? Yeah. Which is like kind of crazy to think about. Like are labbotoies was that like somewhat on the right path? Like it's a very hot take, nuclear take potentially. Yeah.

I mean to be clear like the these forms of neuromodulation like you're not actually taking any of the brain out the way you did with blood labbotomy. But yeah, labbotoies are just like were very crude attempts to like modulate the brain. Yeah. Try and turn off certain regions of the brain.

And if you can do that with some sort of electrical frequency or neurom modulation, like you're in the same realm, which is uh people want a lot of FDA trials before they they run that, but I don't know, maybe over in the Chinese infantry, they're they're they're having they'll do it live. Yeah.

Um is is there anything else that's going on in the BCI industry that we didn't cover that uh like where all this goes next? Uh are we like it's certainly not a trend.

And I guess the big the big question for me is like how do you apply the lens of what we're seeing in AI to the BCI world especially with regard to the fact that we might at some point be on a scaling like on a scaling curve.

Uh and we've kind of seen this with like Tesla's now doing like largecale training runs for self-driving technology.

Uh Google just did the cell to sentence project on the GMA 3 models and you can see that they're starting to scale that up but they're not at like we need to build a one gigawatt cluster cluster for AI and bio and I doubt that any of the BCI companies are like we need a 1 gawatt uh data center to do our train because they don't even have the data yet.

So it feels like we're still really early but is that the right lens to be to be tracking this through? or should we just be looking at like how much electricity is going to be going into the training run and then we'll know that it's actually working.

Yeah, I do think like you often hear people comment how they're like we keep discovering new data modalities like there there are more and more modalities than people realize like there's language and image and audio and video but like biology isn't there's a lot of modalities there and robotics and so forth.

I do think neural data, brain data is a new modality that we're just right at the very beginning phases of starting to research into. And so there there really haven't been any large scale efforts to bit to build foundation models for brains as you're describing. I do think that that's inevitable.

Um, but I also think like I think it's I think this stuff is really relevant for the transversing in AI because I do think this is the inevitable next way in which humans interface with AIS and you can imagine like basically having a BCI that is connected to, you know, whatever the latest AI model is where rather than having to pull out your phone and like type a question into chat GBT, you can just, you know, think something and the answer will surface in your mind or if you're like brainstorming in your own head, you'll have like a thought partner there that's brainstorming with you.

Um, you know, you'll be able to instantly communicate with others. Like there once you start kind of playing out the possibilities of what could happen if there are BCIs at scale, it does really change the way humans like interact with information. Basically, I'm sure Udicowski loves this. It's fantastic.

Uh, it is exciting though. Uh, uh, and I appreciate you coming on the show and breaking it all down. Uh, have a great weekend. We'll talk to you soon. Yeah, thanks. Yeah, thanks for having me. Cheers. Bye. Uh, let me tell you about Finn.

AI, the number one AI agent for customer service, number one in performance benchmarks, number one in competitive bake offs, number one ranking on G2. Tyler, are you going to get a BCI uh intern challenge? Get the Utah array installed. Can we pull up a picture of the Utah array? I put it in the title.

It is very scary looking. You've seen it. You've seen this thing. It is just straight up spikes in your head. It's uh uh just a few wires. I feel like like most protein powders, it gives you brain damage. Yeah, this thing. Yeah, this thing.

This feels like just one notch forward from just actually getting a labbotomy, but uh it's it's a rough looking device, but I guess it actually worked in in some ways. I got some data.

But uh how uh you know, I've often said that I would be I wouldn't be the first person to go on the rocket to the moon or the or to Mars, but I'd be like number five or 10. As long as the first couple people make it back, I'm good. Uh what about you? What number of person do you want to be?

Do you want to be the the first person with the millionth? You want to be the millionth uh million Neurolink customer? I I'm probably in the first uh first 50,000 I would say. First 50,000. Yeah. I mean there's an interesting thing.

Um what speaking of like doomers um like I I don't know if it's Elazer himself but a lot of people have said like one possible solution to like the kind of doom scenario is that we it's like the merge like that would actually be prevent like the you know AI taking over because like we are become the AI they can't take us over if we're them.

Yeah. Jord what number are you do you think you're going to wait 100 or 1 million? Big number. We're gonna need a bigger gong for my number. For your number. you going to get up there? I don't know. I mean, I've talked to Nolan from Neuralink. He seems like he's having a great time with it. He gave me a good review.

I'm down. Wire me up. Yeah, the the value right now from what I've heard. Yeah, I'm I'm pretty I'm pretty quick with the mouse and keyboard right now. So, I don't I don't necessarily need the feature that it's offering. Yeah. Yeah.

And I don't necessarily need you to know my I mean I'm sure you can they'd be able to build the right user experiences around this, but my first thought to a question is not necessarily my my best thought, right?

And so if we're here prepping the show and we're just having this instant communication is just like our worst take going this way and then a bad take against a bad take going back. Yeah, it would need a lot of translation. But imagine puppeteering Adio with your with your BCI. That could be the future.

Customer relationship magic. Adios, the AI native CRM that builds, scales, and grows your company to the next level. Can you imagine? So, a lot of these companies, they uh, you know, a common tagline right now is, you know, Adio is the AI native CRM.

Imagine when the BCI era, when AI is old news, you don't want to put that on your landing page. BCI is the new hot keyword. So you say we're the BCI native CRM that builds, scales, and company grows your company to the next level. I'm excited for all the landing pages to update to BCI native. That's what I want.

Uh I also want you to get an eight sleep. And Jordy, uh I heard you had a rough night. I think you got to get a sound effect ready for me cuz I got a 93. I slept for 7 hours and 25 minutes. Let's go. I got a 47. 4 hours 28 minutes. And and dinner last night was in Malibu. I thought you just went right home.

But uh you know the rough nights come for us all from time to time. Left late. Eric Jong says all according to plan. He is highlighting the fact that uh all of his everyone with the same name as him is absolutely either building cooking or at thinking machines or it's a good time to be Eric Jong. Uh congrats to legends.

Uh you legends. We we still have to do the Eric Jong show. Oh and this is Eric Jong is quote tweeting a different Eric Jong. I love it. They're going back and forth on the timeline. Yeah, we got to get all of them on the show. We do. Eric John, you're all invited.

Talk about opening AAI, thinking machines, cooking, and building. Tyler, can you can you work on this? This seems like invite them all. We want we want the six the six up. All of them. Uh um Logan Robinson says, "When Walmart jets take off for store visits, no one but the pilot knows where they're going.

" I wonder if they have them all blindfolded. Rumors rumor is you're only told a few things to help you prep for the trip. domestic or international, how many days you'll be gone, climate of the destination. What does this mean?

What this is like executive I imagine like people at corporate going to visit stores and seeing how they're happening, right?

They don't want to call ahead and be like, "Hey, you got to really this pre prevent executives that are responsible, I imagine, for like certain things within a store to not frontun it and be like, hey, make sure things are clean. Make sure like everybody shows up on time. " Yeah. Yeah, that makes sense.

They want to do a surprise visit. Yeah. So for nearly everyone, including store associates, the visit is a complete surprise. Got to keep them on their toes. Uh when the these store visits happen, the goal is nearly always the same. Inspect what you expect.

Scheduling these trips in secrecy makes it possible to see the stores as they truly are. No fanfare or pretending. I was never lucky enough to travel in one of the jets, but the purpose of these trips was not lost on me. In my cleaning business, we do something similar. Uh anyways, uh cool concept and beautiful jet.

And if you're long Walmart, you're short Walmart, head over to public. com and enjoy multiasset investing, industryleading yields, and they're trusted by millions. Uh Clyde has been at a conference. Looks like he got a Lucy hat. He did. And then a lot of other stuff. This is This is awesome. Wow.

And some breakers there, too. He got samples. He got What conference is this? He got a happy dad hat. Yeah. Uh uh he should start a convenience store, corner store that's only powered by conference. They only sell. Yeah, you could really resell. You could really resell a bunch of this.

Um, all of these brands should head over to adquake. com. Out of home advertising made easy and measurable. Say goodbye to the headaches of out of home advertising. Only adqu combines technology out of home expertise and data to enable efficiencyless ad buying across the globe.

The other thing that you need to do genuinely if you are running a CPG company, you will wind up at a lot of conferences like these. You will be talking to convenience store owners and you'll be talking to distributors.

And as much as we like to joke about luxury watches, it is a very important aspect of the job, it really it really does like when you're shaking hands with people in this particular industry, tech has been, you know, like you got to be an afficionado, you got to like the craftsmanship of a particular watch.

It's a unique choice in tech, but in my experience in consumer package goods, when you go to a CPG conference, uh what's on your wrist actually matters. So, if you're building the CBG business, maybe you should head over to getbbezzle.

com because your bezel concier is available now to source you any watch on the planet. Seriously, any watch. Uh, no joke, I seriously got uh I I became like fast friends with uh one of our major distributors talking about