Cohere's Ray Dolan on 6G and the Future of Physical AI

Full Transcript

Don Kellogg 0m10s

Hello, and welcome to the three hundred and eighteenth episode of the week with Roger, a conversation between analysts about all things telecom, media, and technology by Recon Analytics. I'm Don Kellogg, and with

Roger Entner 0m20s

me as always is Roger Entner. How are doing, Roger?

Guest Speaker 0m22s

I'm great.

Roger Entner 0m24s

So, Roger, this week we

Don Kellogg 0m25s

have Ray Dolan. Ray is the chairman and CEO at Cohere Technologies, and he's here to

Roger Entner 0m30s

talk with us about ISAC and six g. Ray, welcome to the podcast.

Guest Speaker 0m34s

Thanks a lot. Great to be with you both.

Guest Speaker 0m36s

Ray, you know, for me, you're one of these titans of the industry that definitely doesn't get enough of a recognition of what you've done and what you're doing. You were instrumental in developing OFDM with Flareon, which then got bought by Qualcomm and which is the underlying technology for four g and five g. So you're like the father of that technology, and now you had cohere working on similarly transformative technologies and processes of how wireless gets better. So why don't you give us a brief overview of you and what you do at Cohere and maybe how you got there?

Guest Speaker 1m17s

Great, Roger. Well, thank you very much for those kind words. Far from the father, I guess that would be probably Rajiv LaRoya who came out of Bell Labs. I partnered with him. But let me go back because this month celebrates my fortieth anniversary in wireless.

Guest Speaker 1m30s

And so I... What I go into every day with Cohere, I take a lot of my past to inform, like, how do we lead a disruptive technology into an industry that really, really needs it? And so I'll I'll go back as early as the transition from analog to digital. The industry was suffering from a noise problem. All handoffs were breaking down because cells were shrinking, prices were coming down, holding times are coming up, and it was getting in the way of growth.

Guest Speaker 1m55s

And so the industry really paused. We debated the transition to digital, slicing more in frequency, slicing more in time. And it was really Qualcomm, a group of guys outside the industry that said, you're a noise limited system. You need to make noise your friend. You need to create noise as a degree of freedom, and you need to use a spread spectrum technology to ride the noise and encode the noise.

Guest Speaker 2m19s

And that's what created CDMA. Of course, the soft handoff was another iconic contribution. And, of course, that ended up scaling the entire voice industry into your pocket, and now we give it away for free. So I watched that, and I had a ringside seat to that because we were an investor, corporate investor in Qualcomm. And I watched it, and I thought two things.

Guest Speaker 2m39s

Wow. It's amazing if you really understand the problem that you're solving and you have the kind of leverage that they had. And two, I just was turned on to being an entrepreneur. So that's what led me to join Rajiv when he took the company out of Bell Labs. I was not inside Bell Labs.

Guest Speaker 2m53s

I joined as he escaped, so to speak, and created a series a with some great VCs. So at that time, the industry was going from voice to data. And everyone thought it was a fat pipe because they started the journey in 1989, 1990 with a technology that they labeled IMT 2,000. And, of course, this industry sets a vision out ten years, but the problem is the pace of change is much faster than that. And ironically, right in the middle of that ten year period, the Internet became pervasive.

Guest Speaker 3m22s

But they kept designing to a fat pipe, not TCPIP. As a result, they missed unwiring the Internet and they unwired data and they had to put things into spoof data like brew and like all kinds of Symbian and WAP, if you remember those days. I'm sure you do. WAP is crap. Right, but it didn't matter because it was coming out of the biggest voices in the industry, which had the biggest leverage.

Guest Speaker 3m46s

And so what Rajiv realized, and I was really thrilled to join him, and we led this team together for about eight years, right into the belly of the meltdown in 2000 and all the way through the carnage of, you know, all of those companies either going away or becoming in the aggregate smaller companies. We unwired the Internet. And what we did was we created a fat pipe in OFDM with 10,000 tones and we hopped around to look like spread spectrum, but we allowed TCPIP to get an ACK or a NAC, which is so important when you go into wireless because the error rate goes from, like, 10 to the minus 16 to 10 to the minus two. Right? And so you had to make TCP IP happy.

Guest Speaker 4m24s

That's why OFDM won. It wasn't that much fatter than CDMA from the standpoint of a fat pipe, but it was IP from the ground up. Again, I wanna talk about what problem are you solving and then why are you using the tools you're using? So fast forward, we sell that to Qualcomm. It ends up being a great outcome.

Guest Speaker 4m42s

It becomes the seed of LTE, which becomes the seed of five g, and ironically is probably gonna be in 3GPP the seed of six g, which is insane.

Guest Speaker 4m52s

Yes. Because... Sorry. No disrespect, but can't we find a better technology twenty something years later?

Guest Speaker 4m59s

No disrespect taken. I'm advocating for new technology, and I was blessed once again to meet two other iconic founders, Shlomo Rakib and Roni Haddani, who years ago, the company started by those two folks in 2012, so we're about 15 years old, almost about 14. And they realized that the fundamental problem we're trying to solve now is reflectivity. Okay? And reflectivity creates this appearance of a fading channel.

Guest Speaker 5m25s

But the channel doesn't fade. It just bounces off things. And those bounces, those reflections come in out of phase and out of time and out of angle of arrival, and they look really weird and complicated like a snow globe. But actually, if you've ever played the game of pool, the first guy breaks with as much energy as he can, and the second guy uses intelligence to deliberately bounce things off of other balls or the sides of the table. Right?

Guest Speaker 5m47s

And you just think about it that way. So the problem initially was noise. The second problem was making the Internet protocols happy. The third problem is turning reflectivity into a friend. So they used basically what people are now realizing is a radar approach.

Guest Speaker 6m2s

That's what we're doing. It's called the delay Doppler domain instead of the time frequency domain. You get from time and frequency to delay in Doppler through Fourier transforms, inverse Fourier transforms. Everyone's comfortable with that math. It scales incredibly well for compute and power, and it reveals the reflective environment before you transmit.

Guest Speaker 6m21s

It takes this problem called channel estimation, and it tames it. It tames it in computing power, and it makes spectrum more efficient and allows us to start beamforming. But along the way, what becomes extremely important is sensing, And sensing is basically a radar approach. So all of a sudden, a company that was born in the radar domain to make communications better at every layer of the stack, it's completely interwoven, arrives into an industry that says, can anybody do integrated sensing and communication? I have a counter UAS problem that's urgent now and will evolve into robotics and other enterprise and consumer reasons and use cases.

Guest Speaker 7m3s

And we raise our hand and say, not only are we able to do it, we've been doing it for ten years. We've got all the essential IP. We've got products that work in partnership with some of the biggest companies in the world. And so the timing of the focus on ISAC as the driver of six gs couldn't be better for Kohir.

Guest Speaker 7m20s

And I think what you're doing also helps on the cell edge, right? Absolutely. Not only so Because it's always the complicated thing, right? Everybody is good close to the tower and the transmitter, but it's like near the edge. For example, my house, I'm near the sell edge of like all three carriers, and my coverage in my home is challenging to say the least.

Guest Speaker 7m44s

Right? So I live at the sell edge.

Guest Speaker 7m47s

Right. So one of the things that folks in the industry, at least the engineering folks know, is that half of the performance limitations in cellular come from interference between users in the cell, and half of the performance limitations come from users around the cell. Right? And most people with the scheduler, the way it's designed today, only can tame intracellular interference. They can't really tame intercell interference unless you move the scheduler away from the base station, one node into the network so that it can see across the network and controls inter cell interference.

Guest Speaker 8m21s

So not only do we reveal the reflective environment with the radar technology and create a sparse representation of the channel for every user at scale, But that channel estimate is good for fifty to a hundred milliseconds. And over the last ten years, the edge of the cloud has moved closer and closer and closer to the tower. Now you can get... From any tower, can get round trip to the cloud in ten milliseconds or in some cases less. So we've now created an alternative architecture that sees the channel before transmission, and that channel estimate is good for long enough to where, a, you can move the scheduler to the cloud, and, b, you can insert AI into the scheduling process because it gives AI time to think.

Guest Speaker 9m7s

Okay? So we bought time for digital for signal processing, and we bought time for architectural changes. Now people have said, oh, we're not gonna move the architecture. We don't have enough fiber connectivity. Well, that's evolving very quickly.

Guest Speaker 9m22s

Yes. Your markets. And, oh, we don't wanna do that because we've tried it for ten years, so we don't think it'll work. Well, I've run into that problem before. Qualcomm faced the same challenges.

Guest Speaker 9m31s

You just have to go boldly into the night and prove them wrong. Okay? But the most important thing is people are realizing for ISAC, they're all using what they call multi static architecture. So they're all moving the decision making process from the base station into the core so that they can do things like triangulate. And once they do that, they've moved the architecture of six g into what we call comp in the industry coordinated multipoint.

Guest Speaker 9m56s

Now you solve the edge problem that you started this threat on, which is... And by the way, that's a huge issue. This industry has always had a problem being able to create homogenous user experiences from the near field to the far field. And so this is gonna be solved as well. This is exciting because what I say to people when I'm in technology environments is, hey.

Guest Speaker 10m17s

Are there any radar engineers in the room? And I mean, I'd say, like, 10% of the room raises their hand. It's like, congratulations, guys. It's revenge of the nerds. The radar engineers are going to come in and help the comm engineers understand sensing, which you've always understood.

Guest Speaker 10m30s

And Cohere has built, you can think of as a fully integrated radar and comm solution at every layer of the stack. And I gotta tell you, Roger, I was excited to lead Clarion. I'm super excited to lead Cohere.

Guest Speaker 10m44s

So how will this make a difference in a customer's life?

Guest Speaker 10m50s

So there's layers of the stack. So initially when we failed to hit five gs, and that's what happened when I joined the company in 2018 when Shlomo Aroni had tried to become five g in 3GPP. And I wasn't very surprised that it didn't work. I don't say that in a condescending way. It's just like I've seen this movie before.

Guest Speaker 11m8s

Right? So they said, hey. Can you help? And I was already on the board of directors. I jumped in.

Guest Speaker 11m12s

We bifurcated our solution into standard compliant and standard non compliant. Alright. So the physical layer waveform, you can call that six g. That's called OTFS. But we exposed a higher layer solution, which we call our USM.

Guest Speaker 11m25s

And we've already put that in customer networks at small scale, in some cases by ourselves, in some cases with a partner, and we can make spectral efficiency significantly better. They can slice in time and frequency. We gave them the ability to slice in space where we can create high signals to noise ratios between two users. It'd be like a lighting director on stage that can create complete light and darkness relative to two actors, and it's like you have two spotlights on two actors. So that's one thing we do.

Guest Speaker 11m54s

And we do that already, and we're doing that in partnership with one of the largest OEMs in the world. It's not yet announced. And eventually, we'll do it with everybody. But what's exciting is that ISAC is coming on like a banshee, and we just recently were awarded by OUSW in The US a $30,000,000 project to accelerate ISAC and build what they call a multi waveform base station, which is, if you remember DSS or dynamic spectrum sharing, some of the largest players in the world tried to put four g and five g in the same box and it didn't work. And it was because all the way up and down the stack, they were competing with each other.

Guest Speaker 12m26s

They hadn't really isolated each layer of the stack logically. We've done that. So we're gonna build the world's first real multi waveform base station. It'll have five g NR and OTFS in the same box, and it'll prove that the government can ride on a five g standard platform along with whatever other waveform they wanna put in there. Okay?

Guest Speaker 12m47s

Including some of their proprietary defense waveforms. So that's the next thing we're gonna do, and that's gonna scale ISAC. And that's basically a two year program. We're gonna try to get that done early. We likely will because we're ahead of schedule.

Guest Speaker 12m58s

And it's driving as hard as we can into the LA Olympics with some of the biggest companies in the world that I can't yet name, but that are very, very supportive of what we're doing.

Guest Speaker 13m7s

And a lot of that comes with drones and airspace detection.

Guest Speaker 13m11s

It starts with drones, because they're trying to figure out a problem that is like literally mission critical at scale everywhere in the world. There's not a single government that I've spoken to, and I've spoken to a bunch by myself and with our US partners that doesn't see counter UAS and infrastructure protection as massively critical right now. And they see the cellular network as perfectly positioned. And just to be clear, this isn't gonna be cellular only. This will be fused with other solutions both from space and other terrestrial solutions like LIDAR and radar.

Guest Speaker 13m47s

But the cellular infrastructure at a 100 to 150 feet is perfectly positioned to see across the landscape as a network and track things that they sense in Zone one as they transition from zone one to zone 50 across a cellular network. And they'll see with their eyes across that network. And, of course, the people are trying to figure out what the solutions are to... And where the permissions are to shoot them down if they're hostile, track them if they're not, etcetera. That's the use case.

Guest Speaker 14m16s

Well, we're seeing on a daily basis drone invasions into NATO territory from the neighbors in the East. Right? And, you know, we saw the failed attack in Germany on a Ukrainian plane. And if we have this all integrated into a wireless network, you have a perfect sensing environment.

Guest Speaker 14m38s

Correct. And that grows into... And you start exploring how far beyond the warfighters needs can I get? Can I get into public safety? Things like first responders and things like FirstNet in The US, which is starting to propagate as an idea beyond The US.

Guest Speaker 14m52s

Committed spectrum to public safety protection. And then you start thinking, well, can I do this in the enterprise for industrial automation? Can I do this in robotics? And then what you end up with is a very high resolution image of the physical world, which is the foundation for physical AI. So what I'm excited about is that Cohere can help the existing cellular infrastructure transition from primarily being in the comms business to being primarily in the intelligence business.

Guest Speaker 15m25s

Wow. And I'll actually make a prediction that's possible. I'm not saying it's going to happen. But remember, we thought when we started forty years ago that our mission was to mobilize voice, and we did it. And we put it in your pocket.

Guest Speaker 15m38s

We drove it down in scale, and it's actually free now, believe it or not, because it became the enabling technology to mobilize the Internet Yep. Which we've now put in your pocket at scale, and it's getting cheaper and cheaper and cheaper. And I'm gonna predict for you that at some point in time, it may be free. Because a byproduct of giving you free Internet access is creating massive intelligence in physical AI at scale, and that is the most monetizable asset that can only be seen by the reflective environment. So the other two mobilized markets that were created on a wire, this is the first one that's gonna be created out of whole cloth in a wireless environment.

Guest Speaker 16m19s

And it's going to usher in the age of physical AI, and it's going to enable people that are in the business of connectivity to consider whether they want to keep charging for connectivity or they wanna use that as a tool to charge for something at a higher level, which is intelligence.

Guest Speaker 16m35s

Absolutely fascinating, Ray. I wish we could spend more time, but thank you for coming on the show and telling us about what's coming and how Coher is making a difference here. Thank you. Well, Roger and Don, it's good to

Guest Speaker 16m50s

talk to you again. And we'd be happy to talk over time as this continues to evolve. Things are happening very fast. I appreciate the opportunity to share our story.

Roger Entner 16m58s

Fantastic. Thank you. Fascinating stuff, Ray. Thanks, Roger. We'll talk next week.

Guest Speaker 17m3s

Talk to you next week.