WEBVTT - Bloomberg Markets: Intel's Clarke on New Quantum Computing Chip

0:00:00.440 --> 0:00:03.560
<v Speaker 1>I love my computer. You make me feel all right

0:00:04.360 --> 0:00:08.719
<v Speaker 1>every waking ore and every lonely night. Intel's working every

0:00:08.760 --> 0:00:11.880
<v Speaker 1>we're waking out and every waking night to make computers different.

0:00:11.960 --> 0:00:14.280
<v Speaker 1>And some of the advances happening right now are going

0:00:14.320 --> 0:00:16.159
<v Speaker 1>to show up computers a few years from now and

0:00:16.200 --> 0:00:18.159
<v Speaker 1>they're just mind blowing. Jim Clark joins U right now

0:00:18.200 --> 0:00:21.360
<v Speaker 1>as the director of Quantum Hardware. Yes, Quantum Hardware and

0:00:21.480 --> 0:00:24.000
<v Speaker 1>Intel Um and Jim, let's let's just start your your

0:00:24.160 --> 0:00:26.560
<v Speaker 1>base up their in Portland, Oregon, or in probably in Hillsboro, Oregon,

0:00:26.880 --> 0:00:30.280
<v Speaker 1>or intels some massive fabs, the most impressive and ever

0:00:30.320 --> 0:00:32.240
<v Speaker 1>built in the history of time. But what you guys

0:00:32.280 --> 0:00:36.040
<v Speaker 1>are working on going forward is so interesting. Describing what

0:00:36.120 --> 0:00:40.800
<v Speaker 1>quantum hardware is Okay, thanks for having me on the show. Um,

0:00:41.040 --> 0:00:43.920
<v Speaker 1>I'll start with a coin. Let's take a quarter for example.

0:00:44.360 --> 0:00:46.319
<v Speaker 1>If you hold the quarter in your hand, that can

0:00:46.400 --> 0:00:49.720
<v Speaker 1>either be heads or tails. It's one or the other.

0:00:50.040 --> 0:00:52.319
<v Speaker 1>If you think of a transistor like we build here

0:00:52.320 --> 0:00:55.520
<v Speaker 1>in Oregon, it's either on or off. It's a switch

0:00:55.600 --> 0:00:58.840
<v Speaker 1>that's one or the other. Zeros and one for a moment, Yeah,

0:00:58.960 --> 0:01:01.000
<v Speaker 1>for a moment. I want you to think of that

0:01:01.080 --> 0:01:03.680
<v Speaker 1>coin is spinning. Imagine you're flipping it up into the air,

0:01:04.000 --> 0:01:06.600
<v Speaker 1>and I'll ask you whether that's heads or tails while

0:01:06.600 --> 0:01:10.240
<v Speaker 1>it's spinning. The answer is that it's both, perhaps both

0:01:10.280 --> 0:01:13.560
<v Speaker 1>at the same time. That's what a quantum bit of

0:01:13.640 --> 0:01:18.080
<v Speaker 1>information is, two states at the same time. Now, if

0:01:18.080 --> 0:01:22.039
<v Speaker 1>you bring two of these cubits together, you can form

0:01:22.600 --> 0:01:25.880
<v Speaker 1>four states at the same time. If you brought three,

0:01:25.959 --> 0:01:28.679
<v Speaker 1>it would be eight states. So it's exponential in its

0:01:28.720 --> 0:01:33.319
<v Speaker 1>compute power. It would only take fifty of these cubits

0:01:33.520 --> 0:01:38.200
<v Speaker 1>entangled together, if you will, to produce a register file

0:01:38.319 --> 0:01:42.959
<v Speaker 1>that's larger than any supercomputer on Earth can generate. So

0:01:43.120 --> 0:01:47.000
<v Speaker 1>is this so is this essentially squaring the how many

0:01:47.080 --> 0:01:49.880
<v Speaker 1>times you can You know, you instead of having two

0:01:49.880 --> 0:01:52.480
<v Speaker 1>processes happen at the same time, you can happen sort

0:01:52.480 --> 0:01:55.400
<v Speaker 1>of four combinations of a process process happen at a

0:01:55.440 --> 0:01:57.560
<v Speaker 1>single time, and then you have sixteen of a certain

0:01:57.640 --> 0:02:01.720
<v Speaker 1>process happening at the same time. Uh, that's that's pretty close.

0:02:01.840 --> 0:02:05.600
<v Speaker 1>Every time you add another cubit or another bit of information,

0:02:05.680 --> 0:02:08.560
<v Speaker 1>you essentially double the compute power of the system. So

0:02:08.600 --> 0:02:11.400
<v Speaker 1>you can see that it gets powerful very quickly. So

0:02:11.480 --> 0:02:14.720
<v Speaker 1>I gotta ask you where does this fit into when

0:02:14.760 --> 0:02:18.120
<v Speaker 1>we talk about things like artificial intelligence and we talk

0:02:18.160 --> 0:02:24.080
<v Speaker 1>about deep learning, UM, tell me how this fits into that. Yeah.

0:02:24.120 --> 0:02:27.040
<v Speaker 1>So the first applications for quantum computing will be as

0:02:27.080 --> 0:02:30.959
<v Speaker 1>accelerators in the data center. Uh, they will be used

0:02:30.960 --> 0:02:36.079
<v Speaker 1>for things like material modeling or chemistry. At the extreme,

0:02:36.120 --> 0:02:40.520
<v Speaker 1>they could be used for things like cryptography and security,

0:02:40.720 --> 0:02:45.120
<v Speaker 1>um things like artificial intelligence and machine learning. There is

0:02:45.480 --> 0:02:48.880
<v Speaker 1>a lot of potential for quantum computing UM to be

0:02:49.000 --> 0:02:51.919
<v Speaker 1>to be pretty influential in this space. But these are

0:02:51.919 --> 0:02:55.359
<v Speaker 1>active research areas. This is essentially at the scientific stage

0:02:55.360 --> 0:02:57.320
<v Speaker 1>at the moment. So where are you guys in this

0:02:57.360 --> 0:03:00.160
<v Speaker 1>research process because you're working on it and you I

0:03:00.200 --> 0:03:01.839
<v Speaker 1>think you put it. You put out a new chip

0:03:02.000 --> 0:03:05.880
<v Speaker 1>or another test chip last month. That's right, that's our

0:03:05.919 --> 0:03:08.880
<v Speaker 1>seventeen cubit chip. Intel has only been in this space

0:03:08.919 --> 0:03:11.919
<v Speaker 1>for a couple of years, but in that time we've

0:03:12.000 --> 0:03:14.560
<v Speaker 1>essentially caught up to the lead pack and what amounts

0:03:14.600 --> 0:03:17.440
<v Speaker 1>to maya one of a marathon. So this is just

0:03:17.520 --> 0:03:20.240
<v Speaker 1>the first step, and our goal is to get these

0:03:20.440 --> 0:03:25.160
<v Speaker 1>um small chips working together, all of the individual cubits

0:03:25.200 --> 0:03:28.760
<v Speaker 1>working together to build this powerful system that I just described.

0:03:29.240 --> 0:03:31.480
<v Speaker 1>So is how is this different? Here's what I can't

0:03:31.480 --> 0:03:34.639
<v Speaker 1>get my head around, and it's it's my own problem, right,

0:03:34.680 --> 0:03:37.200
<v Speaker 1>But but is I don't understand how this is different

0:03:37.280 --> 0:03:41.160
<v Speaker 1>than having lots of transistors working quickly. You know, if

0:03:41.160 --> 0:03:43.560
<v Speaker 1>you get a seventeen cubit system, then that would be

0:03:43.760 --> 0:03:48.040
<v Speaker 1>like a hundred and thirty thousand processes happened instantaneously. Is

0:03:48.080 --> 0:03:53.040
<v Speaker 1>that right? Yeah, so two to the seventeen it's not

0:03:53.200 --> 0:03:55.440
<v Speaker 1>that impressive of a number. But then I would ask

0:03:55.480 --> 0:03:58.280
<v Speaker 1>you to plug in two to the fifty. That's a

0:03:58.360 --> 0:04:02.680
<v Speaker 1>number that we can't simulate today. Um, I would ask

0:04:02.680 --> 0:04:04.920
<v Speaker 1>you to plug in two to the three hundred. Two

0:04:04.960 --> 0:04:08.760
<v Speaker 1>to the three hundred is the same number of states

0:04:09.240 --> 0:04:12.280
<v Speaker 1>as there are atoms in the universe. But but here's

0:04:12.280 --> 0:04:15.240
<v Speaker 1>an understand how is it? Is it just about being

0:04:15.320 --> 0:04:18.560
<v Speaker 1>quicker or there's a kind of process. It's possible when

0:04:18.560 --> 0:04:21.520
<v Speaker 1>you're sort of multidimensional instead of binary ones and zeros

0:04:21.560 --> 0:04:24.960
<v Speaker 1>on and off two heads of the quarter. Great question.

0:04:25.360 --> 0:04:29.159
<v Speaker 1>We talked about power that naturally we think about speed.

0:04:29.560 --> 0:04:33.440
<v Speaker 1>It's less about speed and more the parallelization. Imagine if

0:04:33.480 --> 0:04:36.200
<v Speaker 1>I have a huge string of numbers two to the

0:04:36.240 --> 0:04:39.200
<v Speaker 1>fifty and I have to sequentially go through each one

0:04:39.760 --> 0:04:43.720
<v Speaker 1>to find perhaps the number I'm looking for. But imagine

0:04:43.800 --> 0:04:48.080
<v Speaker 1>if instead, all of these states are coupled together, so

0:04:48.200 --> 0:04:50.960
<v Speaker 1>essentially I can search all two to the fifty of

0:04:51.000 --> 0:04:54.039
<v Speaker 1>them at the same time. That's where the power of

0:04:54.120 --> 0:04:57.040
<v Speaker 1>quantum computing comes from. All right, So I'm thinking some

0:04:57.080 --> 0:04:59.600
<v Speaker 1>of our listeners are pulling over saying, what the heck

0:04:59.720 --> 0:05:03.040
<v Speaker 1>is he talking about? Um? I mean, just tell me

0:05:03.480 --> 0:05:07.440
<v Speaker 1>the practical use of this um and where it needs

0:05:07.440 --> 0:05:09.440
<v Speaker 1>to get to what level, Because from what I understand

0:05:09.440 --> 0:05:12.000
<v Speaker 1>is you've got to have processors need to be much

0:05:12.080 --> 0:05:16.440
<v Speaker 1>larger than the fifty cubits to be capable of useful work.

0:05:16.520 --> 0:05:19.120
<v Speaker 1>So it has a long way to go. Correct, it

0:05:19.160 --> 0:05:22.800
<v Speaker 1>has a long way to go. Um quantum mechanics play

0:05:22.880 --> 0:05:25.560
<v Speaker 1>some tricks on us. These quantum states. They don't stick

0:05:25.600 --> 0:05:29.440
<v Speaker 1>around for very long. It's what we call decoherence. Uh,

0:05:29.560 --> 0:05:32.480
<v Speaker 1>these are very fragile, so we're going to need to

0:05:32.800 --> 0:05:35.800
<v Speaker 1>probably string millions of them, if not more, together to

0:05:35.880 --> 0:05:39.160
<v Speaker 1>do something realistic. That's why we're just such a at

0:05:39.200 --> 0:05:42.640
<v Speaker 1>such an early stage of the race. But once we

0:05:42.680 --> 0:05:47.520
<v Speaker 1>have these chips, they should be able to do calculations

0:05:47.520 --> 0:05:50.240
<v Speaker 1>that a conventional computer would take billions of years to

0:05:50.320 --> 0:05:53.599
<v Speaker 1>do because of the exponential scaleing. Jim just got about

0:05:53.600 --> 0:05:55.800
<v Speaker 1>thirty seconds. So is this something that could be potentially

0:05:55.880 --> 0:05:58.440
<v Speaker 1>much more advanced? We're talking awful lot about IBM Watson.

0:05:58.640 --> 0:06:00.800
<v Speaker 1>Is this something that would be much more advanced? And

0:06:00.839 --> 0:06:03.920
<v Speaker 1>just got about twenty five seconds. Yeah, it's a different

0:06:03.920 --> 0:06:08.359
<v Speaker 1>type of computer. For certain calculations, UM calculations that benefit

0:06:08.520 --> 0:06:14.839
<v Speaker 1>from parallelization, matrix multiplication, etcetera quantum computer would outperform something

0:06:15.440 --> 0:06:19.960
<v Speaker 1>like Watson, But for certain UH, for certain types of

0:06:20.440 --> 0:06:26.160
<v Speaker 1>i'll say more neuromorphic um or artificial intelligence applications, I

0:06:26.160 --> 0:06:31.800
<v Speaker 1>would still expect specialized chip to be quite useful. Wild stuff.

0:06:31.920 --> 0:06:33.960
<v Speaker 1>Jim Clark. Jim Clark a good name of technology always,

0:06:34.040 --> 0:06:37.880
<v Speaker 1>especially now. UH is the director of Quantum Hardware and Intel.

0:06:38.400 --> 0:06:40.240
<v Speaker 1>The good Look of the Future, doesn't a Bloomber Markets

0:06:40.240 --> 0:06:42.799
<v Speaker 1>on BLOOMBERGNA have Corey Johnson. She's Carol Masster at Corey

0:06:42.800 --> 0:06:45.400
<v Speaker 1>TV and Twitter. She's at Carol Masser on Twitter, And

0:06:45.440 --> 0:06:46.279
<v Speaker 1>this is Bloomberg