IBM, Vanguard Tap Quantum to Optimize Portfolios

Bloomberg

Watch IBM, Vanguard Tap Quantum to Optimize Portfolios - Bloomberg on Dailymotion

Transcript

00:00I'll start with you, Scott, because IBM and Vanguard, typically people don't think of those

00:05two companies in partnership with each other. So just talk to us a little bit about how this came

00:10to be. Well, we're always looking for leading innovators in industry who want to explore how

00:16to use quantum computing as a tool for their industries, because we put a lot of work into

00:22building better and better hardware and software as we advance it. But we know equally important

00:26is building better and better algorithms and understanding how those algorithms can be

00:30applied to real world problems. So we were very lucky to be able to partner with Vanguard on this

00:36project. And let's talk about what exactly this yielded, because I believe it was a municipal bond

00:41portfolio, right, that contained 109 securities. So when we say it was optimized, what does that

00:47truly mean here? Yeah, so what that really truly means is we were able to take something that we do

00:53very well today. We optimize portfolios at much larger scale and reframe that problem and put it

00:59into quantum space. And we're able to actually get the same result out of a 109 bond portfolio. We came

01:07into this thinking we'd maybe get 30 to be able to get somewhere close using the quantum technology.

01:12But it was a really promising experiment for us to be able to get up to 109 bonds and have it match

01:21current level results. So for those folks who aren't familiar with how this technology is, just give us a

01:25sense here. Obviously, we're talking about speed, but it's a little bit more than that, right? I mean, it is

01:29about the sequencing of potential returns, potential scenarios in a way that you couldn't do before.

01:35What's different? Yeah, so what's different about this particular case is it allows us to put a lot of

01:41different scenarios up and run in parallel. You're typically running a lot of simulations and scenarios

01:47in order. And it really limits you to how much you can put together. And when we put this through a

01:53quantum machine and find the right quantum algorithm, we can start to let correlations pop

01:59up that we maybe did not know were there before. So it's not just about speed, it's about speed to

02:05insight. And at the heart of quantum computing is probabilities. So Scott, I mean, how did we get here?

02:11And when we talk about the technology behind this, I mean, it's been lingering around for a while.

02:15Right. Why is it now sort of becoming a thing? And the second part of that question, is it cost

02:21effective enough for companies to really start to deploy it? Sure. I think why it's taking off now

02:27is because we now have computers, quantum computers that can run programs that you can't simulate

02:33classically. So it allows us to use it as a tool to explore spaces that you can't simulate. This work

02:40basically showed that this approach by breaking the problem into doing part of it on classic computers

02:46and part of it on quantum computers worked. And it showed the kind of results that we were expecting

02:51and we could scale it. So today it's not better than the classical optimization approach.

02:57But if it continues to scale, it'll be able to solve problems that you can't solve using the

03:02classical approaches today. In terms of cost effectiveness, we have more learning to do.

03:07We have more learning as it scales to understand, does it scale well? Are there more algorithmic

03:12innovations that we can do that even get it better than where we are today? And that's

03:17what we hope to be working on with Vanguard going forward.

03:20Is there a timeline that you can put on that? You talk about how there's more learning to do

03:25when we talk about scaling before you see some of those cost efficiencies. Can you put numbers to

03:31that or is it just too early? So we can put numbers to where we are going to be on the hardware and

03:36software roadmap. So we basically have stated that by next year, we will have the first examples of

03:44what we call quantum advantage, where running it on a quantum computer gives you a better outcome

03:48than any other classical approximate method for the first use cases. Not for everything,

03:53but just for a couple. And then by 2029, building much larger computers that can run much more complex

04:00programs. It's called large scale fault tolerant quantum computers by 2029. What we don't know is

04:06we don't know when the exact date is going to be that leveraging these algorithm algorithmic

04:13approaches are going to show significant advantage over classical. And that's why we do the work with

04:17partners like Vanguard. Well, Scott mentioned use cases. And Paul, I want to talk about where this

04:21could go. As I mentioned, you started with a portfolio of municipal bonds. Muni bonds aren't exactly the most

04:27exciting asset class in the world, not to offend anyone. Where else could you see some of these

04:32quantum processes being applied? Yeah, this is really just the beginning. So we started with

04:37municipal bonds because we had such a strong benchmark to go against. It gave us confidence

04:42that we were really putting quantum up against something we felt was world class already.

04:48Once this can scale, you can start to bring in so many other factors. This could be across multi-asset,

04:54taxable. You could do more combined portfolio stocks, bonds. The sky's the limit. And the technology

05:03is the limit. And it's got quite a bit of learning curve associated with it. So we have to start

05:08thinking now. It really is a different way of thinking and framing the problem. And that's why

05:12we're so excited about it that we can rethink the problem and have the technology to try it out.

05:17So it gets to this idea. I mean, are we going to start applying this soon, do you think, in your

05:22view, to the other asset classes? And is it going to be sort of, I guess, broader than just a behemoth

05:27like Vanguard, which, of course, has the resources and the people and the know-how to do it, maybe down

05:32to, you know, say a boutique or a hedge fund to be able to also use this as well? Yeah, I mean, I think

05:37over time, things, you know, things do become available to the masses. But at the same time, there's a steep,

05:43steep learning curve associated with even just formulating the problem space. It took us nearly

05:48two years to formulate the problem space that we are now. So I don't think it's a turnkey

05:54operation to be realizing that value out of the gate.

Category

Transcript

Be the first to comment

Recommended