Sebastian Hassinger  •  00:00
Hi Aaron, thank you very much for joining me. I'm very pleased to have you on the show and and to be at this KPMG event. I'm looking forward to our panel tomorrow. You just uh assumed a new role. The the the title you have now is Senior Director of Quantum Research and Enterprise Innovation. To my mind, as an observer of the GSIs and consulting firms in quantum, what KPMG is doing right now feels like a really strong signal. in the belief of the importance of the topic and the timing of of the development of of the technology. Is that an accurate read? Do you think that's Right. 

Aaron Kemp  •  00:40
I think that's very accurate. I think we've been at quantum for five years now. We started our quantum program five long years ago. And we really had lumped everything into a singular bucket around quantum. We hadn't broken it out into peaky sensing optimization and research. It was me alone doing quantum with the belief that it would come into its own I I jokingly say Quantum's been five years away for 30 years. Yes, it has. It's not a joke, Aaron. 

Sebastian Hassinger  •  01:09
That's exactly accurate. 

Aaron Kemp  •  01:15
jaded look at quantum of when is the carpet gonna get pulled out on this level of this technological advance. What's our next hurdle? Yeah. Um but this one feels real. Uh and we got some new leadership. We've got new CEO that came on in July that is really pushing the company to become more agile. We're 130 years old. We we do turn a bit like an oil tanker at this point. And um you've you've been around technological advancement. You know, if you're in the midst of the company, the immune system kills you. Um so we're out on the edge and we've been afforded that opportunity to kind of stay forward and My new boss and our leadership with this new agile mentality was very much quantum is coming and we need to have the people in the right place to do that. So we've segregated our program and separated it. We've offered and moved our PQC offering into our Cybertech Risk Group and that's monetized and great things. Yeah. But there was the acknowledgement of we need to keep going on quantum. There's a next step after we get over our tenure migration hurl. Yes. And it was okay, let's hop let's let's commit to some some time with the not acknowledgement that we're not gonna have ROI on this immediately, but that we need to build the talent pool, we need to bring in the right staff, and we need to understand what we can do with client quantum for our clients. Before we go to our clients and say what we can what do you want to do. So we're actively playing in the space, but leadership has been great in saying this is one of those things we do see future hire the team. 

Sebastian Hassinger  •  02:51
Right, right. And you mentioned research as a separate focus, distinct. Um you just put out a paper um with uh Keepu Quantum and IBM as co-authors Uh like how did that collaboration come about and and is that sort of the model that you're going to be pursuing in that in that research uh lane? 

Aaron Kemp  •  03:13
So that came about from ironically a PBS video watching the about the American chestnut tree and you know We used to have four billion chestnut trees in the Northeast and we were IBM quantum partners and we've been looking for a project. And I'd watched this. And was like, you know, I wonder if you could find these, you know, they're trying to find the trees because those are naturally resistant to the blight and they want to bring them back. I was like, I wonder if we could find those. And I'd been working with Kipu on some other projects and they had this quantum machine learning feature selection tool. And we kind of got together and started thinking about it. We were like, let's do some multi-spectral analysis of trees to see what we can do. Can we be better than classical systems? And it just kind of naturally formed with the tooling and then the IBM computers and then with what us driving the research and the time of Came together and yes, you can find trees better than a classical system using quantum feature selection. 

Sebastian Hassinger  •  04:12
This is satellite imagery uh analysis. And the the result you got was um something like three percent um increased accuracy. Um I I have already interviewed Enrique from Kipu and he's gonna be on the panel tomorrow Uh I know his view, but I'm I'm curious about your perspective on that sort of uh incremental advantage or or uh you know increase of of performance. I mean the The quantum community is always on guard for claims of quantum advantage and then they'll you know sort of try they'll look at the the sort of theoretical underpinning and something like three percent, four percent is often sort of seen as being uh you know, cherry picking or just noise and not relevant. Wha how do you see that? 

Aaron Kemp  •  04:59
To me it's a positive game and I agree that it three percent is not the let's go boil the ocean we've succeeded. Um I'm also very cautious around quantum advancing We are, I was, it's analogous to we are in a cold war. We beat classical 3%, they beat us 3%. We're going back and forth. We're going to do more work in this. But to me, it's the we did a short project. We did a you know 12-15-week sprint and we there were goals and to do uh to at least meet classical was our goal when we started. So when we actually did get three percent, it was repeatable. That was to me was okay there's something there. Let's continue down this path. And you've been around quantum long enough of The best question in quantum is what makes a good quantum problem. Yeah. I think a lot of us spend time debating that. And I I think there's a lot of paths we've tread down where it is a dead end. classical's going to handle this subset of problems for the foreseeable future. Yeah. So it was good to see a little bit of a gain. So for me that's a positive of this isn't a wasted avenue, it's not a dead end. So we can pursue that further. And we have taken that to other logical choices. If that's expandable to you know, hey, what if we did multi-spectral analysis? of power lines and what's growing underneath them so they know how fast they have to cut. If it's slow growth oak, you're probably okay to leave it alone for 15 years. If it's high growth pine trees, you probably need to go trim that area. What can we expand this into? Yeah. And can we make it lightweight enough to put it on drones? Right. You know, can we make it more real-time? So it was just three percent, yeah, I don't think it's a simpler not getting a Nobel Prize for three percent, but we're We found something where there is something. Yeah. So it it's worth pursuing. 

Sebastian Hassinger  •  06:51
Yeah. And and spoiler alert, I tend to agree. I mean I I think the the debate over advantage and you know the theoretical underpinnings is important, but so is the competition back and forth between classical methods and quantum methods. And you know, I mean The reality is most of the most efficient and most powerful classical algorithms weren't found through a theoretical process. They were found through heuristics and just, you know, r uh sweat and toil to come up with a an incremental increase over something and then, you know, you you never know what kind of gold you're gonna trip over. So I think it's all really, really valuable. And in fact I think it's It's really healthy to build a research practice, especially one that's rooted as KPMG is in the needs of your enterprise clients in those kinds of practical concerns. 

Aaron Kemp  •  07:44
Yeah, it's a it's a unique position, I think, in the Big Four right now to have kind of a commitment from leadership around that we understand you're gonna go explore. And please do so. Right. You know, we're not being directed. I mean, we're looking at trees. Not exactly a KPMG accounting type area that you think about in it, but it is more the Can we then take that to clients that would have applicability around imagery? We do have those kind of clients. Right. Now we've got some experience. And that's them letting us develop that naturally and not being forced into or pigeonholed into some kind of it needs to be CPA type research is very refreshing um and I think plays well into the leadership Hey, we gotta be agile. We've gotta be we're in this AI-driven agentic world and it's moving so fast if we don't learn to move with it and then break away from some of our old paradigms of how we how we act, how we behave, we're not gonna be able to keep up with the tech. 

Sebastian Hassinger  •  08:44
Yeah, yeah. Yeah, you brought up AI. Do you th and obviously this is quantum as a machine learning application, are there ways in which you're seeing the AI tools that are being developed? uh you know impacting the way that quantum research is carried out. 

Aaron Kemp  •  08:59
I I absolutely I think even from a paper standpoint of just the sheer number of papers coming out, uh being able to feed those into an AI and pull relevance facts even at that basic level AI. And then move into the coding and you know I know the vibe coding thing is always hit or miss with but if I can get an 80% working code for the computer on a new project in an hour and spend three days tailoring it and getting it working correctly by a month long or six week sprint to get them done by me doing I'm spending more time than working on the actual problem and not the laying out the algorithm, laying out the programming behind it. So I think those gains, which and we've talked about that as this conference of that human in the loop of take the minutiae parts of it that I don't want to spend time on, you know, integrate this other paper. Okay, we're gonna put uh keepoo tool into this, you know, pull all that together and give me the 80% function, and then I will go through and make the changes. And I think the time saving the ability to rapidly ingest changes, the papers coming out, the changes we're seeing in the quantum environment of how fast that's coming. being able to adapt to it. I don't think a singular person could keep could keep up with anything anymore. So I think AI immediately helps us there. I think we're also starting to see You know, we're especially with Keepu and some of what they're doing where we're training AI models watching the quantum. And you can pull the quantum out then And the model can function on a data set that stays within certain parameters and you can get almost the same results as the model. And you retrain it every three months to keep it refreshed. But that's that opens up so much of, you know, small footprints. We can run this on a drone, we can run this on small compute. So I I think we're still figuring out all these neat ways to use AI. I always jokingly say if you want a novel solution for something, give it to an 18-year-old with no rules. And we're doing that. We're putting quantum computers at campus. We were giving them these AI models. So what happens, you know, you and I are the generation that we used to build computers and get magazines and typing code line by line by line that was provided in the magazine. What are our kids going to do? Dr. Dobbs. Yes, exactly. What are these kids going to do with this that don't have, you know, in my case, 35 years of computer experience that's probably has some biases built in out of what that won't work. Right, right. Where these kids don't have that. Yeah. I expect big things and then next four years when these kids have these toys. 

Sebastian Hassinger  •  11:34
I refer to sometimes the these LLM tools as an outboard brain basically. Strap it onto your head and do more faster. 

Aaron Kemp  •  11:44
It's a it's a very good analogy of an augmentation. of allowing me to think about the problem. I think Jensen said it. You know, I need you to solve problems, not write code. 

Sebastian Hassinger  •  11:54
Right. Right, right. So the the that collaboration around that paper, uh keepo's a startup uh with a focus on algorithms and software and IBM obviously hardware vendor. Is that sort of Do you think that'll be a repeating pattern you sort of find a startup and a hardware partner to sort of go after these types of explorations? 

Aaron Kemp  •  12:17
I I I do because I think it offers the best of both worlds. We're gonna have to explore modalities. IBM was where we started they seem to be at the forefront right now when you look at roadmaps and they're hitting their roadmaps regularly um which was what drove a lot of our decisions But even IBM's exploring other modalities like neutral atom. You know, they just bought the spin qubits with HRL. So I we're well aware we're gonna have to play in other spaces. And I think the startups for me personally is they're solving problems. 

Sebastian Hassinger  •  12:45
Those are those 18 year olds. 

Aaron Kemp  •  12:46
Yeah, no solving those problems where we were going to have to figure out how to do that anyway. If you've got that piece of that puzzle, I just add that to our mix. Right. And I I that moves me closer to the problem I'm chasing and I'm not spending time reducing gate depth or chasing getting this algorithm to work right on the computer within a specified amount of time so we don't lose you know, decohere out of it. Right. Um so I think the startups are for I think for the probably the next four or five years, that really is the model as these startups come out. But we've all been around the startup model of Then the big players start to absorb them on the IP and we'll hit that phase like as we start to shed modalities and come to whoever's gonna win this. Right. Right. But for the for serial future I think the teaming up with these startups and leveraging what they're doing. doing these novel ideas is going to be really big for us. 

Sebastian Hassinger  •  13:36
Yeah, yeah. That's really cool. And and I mean, you know, you're you're saying that like you're you have to uh a lot of leeway from leadership and and freedom to experiment, but do you still have a sense of sort of at least the context of the KPMG customer, of sort of an a looking for applications that might intersect with the needs of your customers in the future? 

Aaron Kemp  •  13:57
Absolutely. I think that's, you know, we are, I'm very cognizant. that we are for-profit companies. And that, you know, now I've been given this chance to build this research group, that we are going to have to become profitable. And it was very much the same thing I was able to do. do with PQC. We were building frameworks, we were getting ourselves ready, getting knowledgeable, getting the tooling in place, making the right startup con contacts, making sure we had the right partnerships, and then we were able to monetize. And I see us going the same route. And even the silliness of the tree experiment, you know, the there is applicability in If I can go out and look at farms and tell how a farm yield is going to be off of crops, insurance companies can then forecast if they're going to have payouts because the crops failed. So there is applicability in a lot of what we're doing and searching for those searching for those places where it is a good quantum problem, the second step is now where is it at bookable for our clients? Right. That's really interesting. 

Sebastian Hassinger  •  14:59
It's a good analogy. do large scale accurate analysis of satellite imagery uh and then connecting that to to forecasting and insuring and those sort of types of things. That's Um that makes it's a really good example. Um so okay, you you mentioned PQC and developing that line of business and now it's transitioned into uh the broader cybersecurity practice, which I think is that's that's great. It it means you definitely did your job because it's that it is the time when people are shifting into that sort of Okay, now we gotta figure out what we need to do. And you know, the the migration, helping your customers with migration understand what they need to do, that's a big part of that uh the Q prep um for framework that you that you put together. What what's sort of the the headlines of that for from your perspective, the sort of the the main kind of things that were maybe surprising or or you didn't realize how important the um the aspects that sort of surfaced in the creation of that framework. 

Aaron Kemp  •  16:04
Yeah, so that framework started five years ago. Um, we actually had gotten an RFI from a major US company around PQC and we realized There wasn't a lot out there. There was a lot of talk and there were suggestions. But what was missing was governance and um and some way to put this together. How do you go to a board and say I need to go from where we're at today in our cryptographic journey to a continuous monitoring, quantum resistance, agile spot down the road, and we don't know how to do it. yet and as we were going through the RFI I was taking notes and just going, okay, well those tools don't exist yet. And that doesn't exist yet. And I spent about a year and a half really researching. uh what Canada was doing, what the UK was doing, what Australia was doing, um EU, there was all these thought leadership papers out there. And I just started taking what I thought were the great ideas. I have a cybersecurity background. Right. I'd done time at Lockheed with classified cyber cybersecurity and I really had a good feel for how those labs work and how you how data flows in and out and what it takes to build you know from the day of birth of a lab to the day you take a lab apart and shred all the computers. because they have to go to the dump. How do you do that? I looked at the frameworks and started pulling out what I thought clients needed to do. Like if we were walking in to brief a board, how do you convince a board to fund a project on something that's theoretically going to happen. You know, PKC is another one of the ones. Yes, we think it's going to happen, but it's dependent on us solving our quantum computing issues. We think it's all gonna come together. So we've got the math, we're getting the computers engineering. So we built this framework and it really became data-centric When I thought to myself, we don't have a cryptographic problem. We have a data problem. None of these organizations know where their data flows. We tag. We don't No, we're we've thrown everything into data lakes. Yeah. Um we don't have the isolation we used to have. I've always kind of found auditing around cyber interesting. of well we need to I audit the computers that work on the financial records. And I'm like, yeah, but the Kef Tyria computer that I just ordered my lunch on is tied to the same network. And you know, lateral movement's a thing, so why are we not looking at that computer? And as I started thinking that through this, I was like, so if we knew where all the data was and we knew what we had to protect first, that m that lowers the threshold problem into more manageable chunks of We're not protecting parts of a system then we're trying to protect what's really important. So we built the program and obviously the first thing is starting with a baseline. of what do you have? And for me, that got pulled back through experience with clients. Right. Okay, we may even need to go back to your CMDB of what do you have because it was, oh now there's tech debt. Oh now their software is expired. Yeah. And then as we've grown through it, our framework is adapted. We always have a we always have a lessons learned meeting, which I brought from the Navy. with me at the end of every engagement of where did q prep work? Where did q prep fall short? And how do we fix those shortcomings and make this better. So it is seven steps. You can do about the first four right now. Right. Uh because we're still waiting on certificate uh standards. We're still waiting on you know s enough for the software industry to catch up with PQC that it makes it worthwhile. Yeah. Um I think in the next year that all happens. So the framework's a living document right now and it's it's moving with the times. We just had the executive orders that 

Sebastian Hassinger  •  19:35
reduced what we thought was gonna be a ten year run to five. Yeah. Surprise. Yeah. 

Aaron Kemp  •  19:41
Um but we've also designed it to where the roadmaps we provide the client are five year roadmaps to the best of our ability depending on the size, but they have a two-year accelerator built in in case we have that breakthrough. 

Sebastian Hassinger  •  19:53
I see. It is daunting and what you just said is it you know Step one is actually, you know, taking stock of what you have and what your potential exposure is, which kind of reminds me of, you know, other major inflection points like uh like Y2K for that matter or um when virtualization really hit and and suddenly you know, uh the CIO is being asked like how many physical servers do you actually have running Oracle or whatever other license, right? It is kind of It's interesting that the the step one is actually just assessing what you have, which is a hard problem. When you're moving really quickly, that's kind of one of the first things you get a little bit shaky on. Keeping an exact count and uh inventory. 

Aaron Kemp  •  20:38
And it's a ma and if even worse, take organizations that are acquisition heavy. Yeah. Yeah. That are we've acquired you know 15 companies. used in the last 15 years and we've never unified our compute stack. We did a we did some work with a client that had 5,500 odd pieces of software and when they started looking at it, it'd been through it been through a lot of acquisitions. And what they kept doing was just expanding the licensing to cover the company for the tools. And then somebody realized as we were doing kind of a justification process that They had fifteen pieces of program management software that did the same thing, licensed for every person in the company. 

Sebastian Hassinger  •  21:14
So they immediately Well, they must have been happy that you found that efficiency for them. 

Aaron Kemp  •  21:18
So that was one of those things they were able to do to the board and go, look. First step of the process, we're already saving like what we just spent to do this, we just saved in licensing we're gonna reduce. So like that kind of training has been a A better way to instead of sell this as a mandatory like the sky is falling, cryptography's over, it's we've transitioned it to this is the one of those generational opportunities to clean house. Right. You know, get rid of your tech debt, go in and fix your software, get your licensing in order, get your house in order, and come out the back agile. Yeah. And you know, just your zero trust and everything in place, your cybersecurity professionals will be better off for your bottom line will be better off for it. Your risk 

Sebastian Hassinger  •  22:02
Yeah. Hopefully people will uh take on board the the lesson that security is like a verb. It's not something you just do once and you're done with it. 

Aaron Kemp  •  22:12
Well that's cybersecurity is probably the worst career field ever because Perfect cybersecurity has no RO because nothing happens. Right. And it's difficult to continue justifying a spend when we didn't get breached. That's because we're spending appropriately. Yeah. So yeah, cybersecurity can be very unglamorous when we're successful. 

Sebastian Hassinger  •  22:33
No kidding. And and you mentioned your your um your defense background and and uh top secret sort of or classified environment um experience. That does that Does the perspective of those experiences make you worry that the regular enterprises are not seeing that this is like a national security grade sort of uh um threat model that's coming down the pipe. Um you know the whole harvest now and decrypt later like Of course, national secrets obviously are are more valuable in 10 or 20 years than the average enterprise, but If we've got machines, the oratomic thing, for example, being able to break RSA 248 2048 um in a matter of a week or two, I mean that's still a number of years away, but are we going to adjust to that kind of it's radically different way of thinking about the threat models uh for for civilians. 

Aaron Kemp  •  23:35
I I think we'll we'll get there. I think you know, I think we've gone through this mad rush to the cloud if we're gonna move everything to the cloud and You know, I was talking to somebody at this conference and they're now building a whole business model around bringing AI on prem and isolating and taking away the connections. I'm like, we've come full circle. Um and even in cybersecurity, you know, it used to be defense in depth And then it was zero trust. Right. But we're going back to that walled model. Yeah. Because we're gonna have to keep because of AI um and the threat of things getting into AI that you don't want getting into AI, we're bringing that back in house. And it now data flow is becoming important because you don't want AI models being poisoned or your data being in an AI model. Um so we've got almost got this full circle. You know, I started cybersecurity at Hewlett Packard back in the nineties and all with you know through through the 2000s kind of that initial foundational cybersecurity went away and we went to all these new models and now we're back to wait no we're gonna bring it all back on prem we're gonna isolate it Unplug that modem. Correct. And the set the classified cybersecurity stuff is cybersecurity at on steroids. I mean we talk CMDBs. with a normal organization and they're tracking laptops, we were tracking RAM boards. Right. I could tell you how many RAM boards were in the lab. Wow. Um and you know we had RFID tags on everything. You could just walk the lab and you would get your full inventory scan with the RFID scanner. I don't know that we have to go to that level for anything, but I think people in the space that know how to run security at that level is important because they are thinking of the BitLocker defaults to 128-bit by default. 

Sebastian Hassinger  •  25:13
Nope. 

Aaron Kemp  •  25:15
Why? Yeah. You know, data rest is a big deal. Where is it going? Data in transit. How is it transiting? through the building. Yeah. You know, I had a crazy lab that were separated by a big hallway through the building. And to get it approved because of the classifications on the labs We had to cut a hole under the floor and then that was a sealed chamber that was filled with nitrogen so that if there was it if it detected oxygen in there we knew somebody may have penetrated alarming. I mean nobody's gonna go to that level of security more. I mean a skiff, last time we built skiffs, I think they were almost $2,000 a square foot to build a SCIF. Probably not necessary for most businesses. Right. Thinking at the simpler things of inventories. Yeah. Those lessons I think need to be readjusted. I think we were both yesterday that the um we need to get back to basics. You need to know what you have. Yeah. You need to understand your footprint. You need to understand what you own, what's virtual, what's not virtual. I've been through some incidents where it was we have five thousand servers, are they virtual or we don't know. That's not a great answer. Not great. Um you know how many phones do you laptops have you? You know how many monitors uh you know the Samsung I don't think Samsung's gonna put out a patch and you're gonna have to buy a new monitor. Yeah. So as you talk with the boards and you talk with leadership, it's those Understanding the basics of no your footprint, you know, you mentioned Y2K. I did some research, you know, a large company might have had a hundred thousand devices in Y2K. Tens of millions of devices in a large case. 

Sebastian Hassinger  •  26:55
I think I have a hundred thousand devices in my apartment. 

Aaron Kemp  •  27:00
Last time I looked, there were like seventy-seven devices on my router. I'm like that's my my singular home. Yeah. And all of that has to be patched. 

Sebastian Hassinger  •  27:08
Yeah. Yeah. Why well good thing we have AI. But okay, so so shifting back to sort of the future looking, because I mean You know, so now we're in this process of sort of trying to take steps to uh to guard against the threat of these future quantum devices. Do you see any, you know I mean, where do you see the sort of first value in in the typical KPMG customer for for quantum applications? Um if you think, you know, in that five year Right. 

Aaron Kemp  •  27:41
No, it's a great question because I think w as we're talking to clients, you know, and there's interest from clients. The the the first question What are the use cases? What what are the use cases? Where's an ROI on? Yeah. And we've gotten really good about your ROI right now is experience your team's getting to experience quantum and understanding that it's hard to choose a good problem to be quantum and that's some of the things you especially with the groups that are you know have all your optimization algorithms you know this isn't a blanket you just port that over and it runs. You know it there's going to be work to do that. Um and as we've talked with our partners everything I think one of the interesting things out in the space is three four years ago you would you were seeing a lot of research papers coming out of big financial companies and that has gone very quiet lately and to me that's a very good telltale sign that we're moving past research to where they're starting to see actual things they think 

Sebastian Hassinger  •  28:41
A little less open about the research. Yes. So we used to host a um annual quantum luminary session. 

Aaron Kemp  •  28:47
And it was really bringing the startups and kind of the hardware and software players. And it was behind closed doors. And it was Chatham House rules of let's just understand quantum. And it was one of the university research. researchers that said the sharing is drying up, which means we're getting close to profitability. Right. Um and those very good observations. Markers we're seeing now. Um so when we're talking with clients it's You you you're looking for an ROI. My question back to you is can you afford not to if your peers are? 

Sebastian Hassinger  •  29:16
Yeah. 

Aaron Kemp  •  29:17
Because when the computer, it's not these are scale, you've worked for Then they're scalable. Um we're building these where we're waiting on the qubits to catch up, but we've got the algorithm. 

Sebastian Hassinger  •  29:26
Yeah. 

Aaron Kemp  •  29:27
Which means they're not gonna need a lot of work to pop that on when the computers are at scale. So if your peer has built that and you know, I think the HL SBC paper was a great example of if they're 34% better than you tomorrow at scale, how long does your business model hold? 

Sebastian Hassinger  •  29:42
Yeah. 

Aaron Kemp  •  29:43
Um so it's not a matter of is there an ROI today, that ROI is down the road, is when the computer 

Sebastian Hassinger  •  29:49
gets here. Yeah. Do you see your research team and because you're you're do you're publishing papers, is the purpose of that open science, that open collaboration to map out new areas of potential value or is it to you know create uh an awareness among your customer base that that you are y that you have thought leadership, that you have the skills, that you can help them, that you know to almost as a form of marketing in a sense and also just building the muscle so you are capable of delivering that that kind of value. 

Aaron Kemp  •  30:23
Yeah that's a great that's a great question. And then the answer is both. We as researchers I think all of us are new in what we can do with quantum. I mean, every time I'm at every time I'm with you or with you know moth or with anybody, it's novel things I haven't seen or thought of. And then we've trade ideas, everybody's going to wait, why didn't we think of that? It's very much a communal helping each other. What's going on? Why isn't this working? Um but it is also we're an accounting firm is what we're known as. So having our name in the space, you know, I I woke go up to the Chicago Quantum Exchange a lot and it's we'll go on the hiring days. And the physicists and the kids coming out of college going, wait, why is KPMG at this event? Wait, you're hiring researchers is a new thing in the field. This is but that's also probably another marker of It's getting close enough that firms, you know, we get to leverage our relationships that we have with clients. So if we can have an expertise in that space, they've trust and we've got a trusted relationship. That's an easy thing for them to port over. 

Sebastian Hassinger  •  31:28
Yeah. 

Aaron Kemp  •  31:28
Um so we're doing it for both reasons. 

Sebastian Hassinger  •  31:30
Um I think, I mean, as disruptive as AI is and has been and will be, it's still just an incremental improvement over the classical technologies that we have known and loved and used and hated and been frustrated by whatever for the last now eighty years, this is a radically new way of approaching problems and thinking that you can just uh be you know second or third to implement something. It's like, no, there's a lot of groundwork you have to do. 

Aaron Kemp  •  32:09
we should probably highlight of the talent pool. Yeah. I I think we w when we were at P33, they said there's 16,500 quantum researchers. on the planet. I know. Um and you know, here's this technology rushing down on us that's going to be enormously transformative. And we're not making quantum people fast enough and it's it's It's here. And it expands beyond that of what does the quantum workforce of tomorrow look like? Of you said it earlier, you know, you need PhDs. to run just to keep these computers running. Yeah. What does that workforce look like? How do we get our K through 12 interested in this whole new model that's coming down the pipeline? And we're going to have a big gap. Um you and I are probably prime examples of quantum guys are a hodgepodge of elective engineering and computer science, physics, math Yeah. And you know, we don't have those pure quantum computing degrees. 

Sebastian Hassinger  •  33:03
We're just starting to see them. There's some notable exceptions, but it they're notable because they're rare. I mean UCLA's got a fantastic masters of quantum engineering program that should be a a a model for universities around the world because we can't just wait on the PhD pipeline for the I mean as you said, sixteen thousand, that's it's a drop in the bucket. It's unbelievably small 

Aaron Kemp  •  33:27
Fortune 200 will hire 16,000. 

Sebastian Hassinger  •  33:29
That's right. That's right. And as soon as there's a commercial, there is that ROI. that that town pool's gone gone and the and the supply chain is t completely exhausted and et cetera et cetera et cetera and we're just not ready for the scale at all. 

Aaron Kemp  •  33:45
No and we're I mean we were what 3540 quantum computers active on the planet right now? What happens when that's 4,500 quantum computers? What happens when that's 45,000 computers. It's a daunting it is. It's a daunting ramp up to we often talk about AI is. It's gonna get rid of a lot of jobs, but I'm looking at it going, yeah, but there's this slew of jobs coming down the road that we're not ready for. 

Sebastian Hassinger  •  34:13
That's right. And uh, you know, I mean the point was made earlier today on stage about But uh the models are always is by Mark Cuban, the the models are always lagging behind current knowledge. So, you know, I mean it's I'm amazed that I mean it's it's a testimony to how much the archive, the quantum pH category in the archive encapsulates the the work to date that you can actually get workable answers out of the LLMs. They have trained on that corpus, obviously, and others, but that's all backwards looking. And as you know, I mean Tomorrow's innovations are not in the models that were trained last year. So you need the human brain to be engaged. The more novel the space is, the more you're reliant on human intelligence and not machine intelligence. 

Aaron Kemp  •  34:56
Well and Mark said it. What's the only thing in the loop that knows what's going on today? Yeah. It's that human. And that's the researchers and that's the telephone pool we need. then that's right they're gonna need those kids today. You know, kin the kindergartner starting, they're gonna walk into a world where these are gonna be everywhere. Yeah. Yeah. I still think think they'll be in the cloud. I don't think we're gonna have them at home, but you you're gonna have this compute capability at your fingertips. Um I don't know where it's be in the orchestration layer in 10 years. But we we've got a lot of challenges. Yeah. And that's another part of what I'm doing with the research group. of when we get in front of clients of you need to start locking down your talent now and and developing your talent for those in the house that can. Yes. Because you are going to have a few Yeah. Yeah. They have enough adjacency with their backgrounds that they can shift into the thing. 

Sebastian Hassinger  •  35:47
And just raw like curiosity. I mean people who are novelty seekers. I'm one of them. I'm not a physicist by training. I just love this stuff because it's so interesting to me that I I can't stop basically. And every organization's gonna have some people like that. 

Aaron Kemp  •  36:02
And and g you know, I think what KPMG has done with us and with what we've been able to do with Quantum. You're if you're a major Fortune 200 company, it doesn't cost you anything to let three people explore and get ready and that ROI there, you know, duplicate the model of put some people out on the edge and let them kind of function in a space that's not the norm for the company. Yeah. And then you might you're gonna be more ready for what's happening. Right. and we've we've got more compute stuff coming after quantum. Yeah. You know, the neuromorphics and Thermodynamic you know with the the heat off chip stuff and all the things that are exothermic computing and all the other kind of craziness that we're not talking about yet based quantum. After we've rolled out of the AI hangover. But they agree with you with compute. 

Sebastian Hassinger  •  36:54
Right. I agree with you. I mean I think You know, Moore's law sort of lulled us into this idea that that it was just going to be, you know, Boolean logic semiconductor, uh, you know, CMOS for the rest of time. We were running out of that runway and the answer is a whole bunch of heterogeneous compute engines that are specifically good at X but not Y and this other one's good at Y but not X and That just means more uh skilled and uh flexible and you know natural learners basically means they're gonna have to be more educated, more trained, and more curious. 

Aaron Kemp  •  37:30
Well I love the mechanical computing starting to come back. Yeah. You know what it's easier to design. Babbage will have his day. He will. I mean it it's easier to build a device that does the computing mechanically because it only has to do that and we leave it in the corner and that's what it does. Yeah. Because it doesn't have to do anything It's efficient. Yeah. Um you know, what does our compute stack look like in 35 years? 

Sebastian Hassinger  •  37:52
It's what keeps me absolutely obsessed with this stuff. 

Aaron Kemp  •  37:56
Or I mean really with what we're doing, what does it look like in twelve months? Yeah. Yeah. That's right. It's a it's a and clients are worried. Yeah. And they should be. Yeah. You know, PQCs first. Quantum op quantum optimization and research is second. Quantum sensing's now. Right. Um and you know there's There's just it's m I think AI has gone from LLM to agents to you know, all what all the things we're talking about now to hey we're gonna put it back on prem and isolate it to all these different models, um yeah to multi agents, to orchestration agents. Um yeah it's it's It's it's amazing what we're doing. I don't I don't remember a time in my life where compute was moving this fast. 

Sebastian Hassinger  •  38:42
Totally agree. Totally agree. Well, this has been great, Aaron. I really appreciate your time. I'm I'm really excited to see what the team you're building is gonna um produce. 

Sebastian Hassinger  •  38:53
I think the the the collaborative model and the the curiosity that you're engendering uh and the the questions you're exploring for your customers are really interesting. So um looking forward to seeing it. Well thanks for having me. This has been a lot of fun. Great. Excellent. Perfect template