WEBVTT

1
00:00:16.739 --> 00:00:25.289
Silence.

2
00:00:26.460 --> 00:00:33.509
Okay, good afternoon. People hope you can hear me.

3
00:00:33.509 --> 00:00:42.719
So this is, I guess, class 21 for quantum computing, Monday, November, 16th, 2020.

4
00:00:42.719 --> 00:00:49.890
Okay, so you had a good weekend. Let's see if I can get sharing working here.

5
00:00:52.920 --> 00:00:56.850
Good. Wow.

6
00:00:56.850 --> 00:01:03.090
Was pleasantly surprised when computers actually work.

7
00:01:06.150 --> 00:01:15.719
Class 22. sorry. Okay. So what is happening now? Various random announcements.

8
00:01:16.224 --> 00:01:29.754
And that's better. Okay. 1st, I've got 3 guest speakers scheduled other professors at this is all tentative.

9
00:01:29.754 --> 00:01:44.275
Of course, it may change, but this Thursday at full tongue is talking next Monday arena Wang, and Monday after that. So there are more physics related people. So, I'm a software person, even though I actually did take.

10
00:01:44.579 --> 00:01:49.409
More physics than anything else is an undergrad.

11
00:01:49.409 --> 00:01:54.090
Well, I passed out in my 1st company as a computer science course is actually, so.

12
00:01:54.090 --> 00:01:57.329
This will give you the electrical engineers.

13
00:01:57.329 --> 00:02:01.439
Some more physics based understanding of what's happening.

14
00:02:01.439 --> 00:02:07.290
2nd, for people who want to have just.

15
00:02:07.290 --> 00:02:10.560
Colors here.

16
00:02:10.560 --> 00:02:14.460
Come on not what I wanted.

17
00:02:16.289 --> 00:02:21.240
That's what I wanted for people who want to learn more about D wave.

18
00:02:21.240 --> 00:02:24.840
D, wave is running a.

19
00:02:24.840 --> 00:02:31.949
Seminar tomorrow morning, 11 o'clock, Eastern time talking about.

20
00:02:31.949 --> 00:02:46.080
D wave computers and why you should buy 1 of them. I expect, but in any case, you have to register D wave to see this Patriots. Why? I didn't put the link online, but I think it's free and.

21
00:02:46.080 --> 00:02:53.490
So, if you want to learn more D, wave register, goes to the D wave Web site register for this and.

22
00:02:53.490 --> 00:02:57.090
And then tell me what they say, basically.

23
00:02:57.090 --> 00:03:08.310
So, you can have fun with that, right? So, obviously it's a sales book, but it still might have something technically.

24
00:03:08.634 --> 00:03:23.485
Interesting. Okay now, 1 thing I find useful for sales presentations is, I don't know suppose I was trying to write a proposal or trying to talk about quantum computing to someone.

25
00:03:23.759 --> 00:03:27.900
I could actually go to 1 of these sales documents and.

26
00:03:27.900 --> 00:03:37.349
You know, get buzzwords and stuff like this. So I can mine sales documents for useful selling information.

27
00:03:37.349 --> 00:03:48.840
1 thing I want to do here. Okay. I got the chat window open. Now. In case anyone wants to tell me stuff, stop working and so on. Okay.

28
00:03:48.840 --> 00:03:54.449
So, that's what I had under the quantum news thing. Okay so what's happening now.

29
00:03:55.224 --> 00:04:07.525
What I want to do some status stuff I tried to show the Google document on Thursday. We had trouble got through about 10 minutes, but she's such a good speaker. So such a good talk.

30
00:04:07.525 --> 00:04:11.724
I want to continue on with this and.

31
00:04:12.389 --> 00:04:17.639
And I showed you, I think I showed you 1 thing on quantum trapped ions on.

32
00:04:17.639 --> 00:04:26.249
Thursday, I may show some more. Oh, I'll show you this Cleo conference owners and electro optics.

33
00:04:26.249 --> 00:04:37.408
This is again, this is guys, Christopher Monroe. He's a really good speaker. So this is a really good talk. He's better than me. They all are. That's why I show them to you. So I'll show you this.

34
00:04:38.519 --> 00:04:47.728
But before I got to lighten things up a little bit before I get into serious stuff, then obviously we've got Honeywell and so on.

35
00:04:47.728 --> 00:05:02.488
But I want to show you this Minecraft on a quantum computer 1st actually to life and people up. But before I do that, I want to show you. I've also put another homework online. Where did it go here?

36
00:05:03.509 --> 00:05:14.153
And here we call, so this is to continue this week's homework where you had, like, use a simulator on Amazon bracket.

37
00:05:14.153 --> 00:05:19.644
Now, I want you to use actual hardware both D, wave the quantum.

38
00:05:21.564 --> 00:05:35.723
The Optimizer thing, and then 2nd, to use actual ai on cure regarding hardware and then to compare the 2 of them. So I'm giving you 2 weeks while week and a half. If you subtract Thanksgiving.

39
00:05:36.024 --> 00:05:47.334
So, this well, I'm a chance to use actual computers and via Amazon, the easy way to do it. So, by the end of this semester, we used IBM used.

40
00:05:47.968 --> 00:05:59.309
The D, wave is the I, on Q3, different types of quantum computers so you can honestly put on your resume that you've used 3 different.

41
00:05:59.309 --> 00:06:05.728
Quantum computer architecture and the actual quantum computers and.

42
00:06:05.728 --> 00:06:09.329
Who knows, it might have a positive correlation to getting a job.

43
00:06:09.329 --> 00:06:18.988
So, but if you do that, okay, this time, we do not need another performing arts center at. How about an engineering building this time? Okay.

44
00:06:20.038 --> 00:06:23.069
So.

45
00:06:23.069 --> 00:06:28.319
I want to show you something light. 1st, before I show you the.

46
00:06:28.319 --> 00:06:32.579
He stopped, so this is an IBM video.

47
00:06:32.579 --> 00:06:35.819
Minecraft on a quantum computer.

48
00:06:39.204 --> 00:06:51.233
She hello guys ever loaded 1 of those procedually generated games and are stuck sitting for a while. Log generates a new world for. You afford to sometimes generate patterns that you feel like you've seen in every other game. That does this.

49
00:06:51.353 --> 00:06:59.483
Well, it's always been a dream of mine to generate truly wonderful kind, but fully explored will complex worlds and that lightning speeds.

50
00:06:59.579 --> 00:07:09.713
Well, you see this, as I envision, it is something that only a quantum computer can do, because even some of the fastest computers that we have struggle, which I, and generate these complex worlds.

51
00:07:09.834 --> 00:07:23.963
But, thankfully, today's video is sponsored by, which means we get to try and make my long lasting dreams. Come. True. Because not only does kids can have quantum computers according to some sources kissed kit, which is a quantum team within.

52
00:07:23.963 --> 00:07:37.314
Ibm is leaving grace on quantum computing. And it makes sense to me, they open source their quantum computers. They've made such expensive technology, free for people like you, and I, to play around with. So, as far as I'm concerned.

53
00:07:38.608 --> 00:07:52.108
Lead on lead on, but to get started with today's project, I'm going to 1st, need to know the basics of quantum mechanics because without that, how can I expect to be able to understand how a program for a quantum computer.

54
00:07:52.553 --> 00:08:06.863
Oh, right. I should probably explain what it is a tool that allows even us bedroom developers to write quantum computing algorithms and run those algorithms on in real life, quantum computers and get all the new age benefits.

55
00:08:06.863 --> 00:08:13.524
That computers offers, like, massive speed ups, entanglement, parallelization and more.

56
00:08:13.584 --> 00:08:27.113
But you see, there's a problem quantum computing is still really new and writing a quantum algorithm is a bit tricky but thankfully kids can has an amazing that helps make writing quantum algorithms. Really simple.

57
00:08:27.178 --> 00:08:38.094
Look pretty graphics look at it, believe it or not. But, what you're looking at is a quantum awkward and what I loved about learning how to write quantum algorithms to kiss kids because they have a lot of really good resources.

58
00:08:38.124 --> 00:08:47.844
That really helps give intuition on what's going on and kiss, get his written in Python. And because we're friends, I'm going to give you a bit of a cheat sheet.

59
00:08:47.903 --> 00:08:59.484
The challenge writing quantum algorithms is basically just doing a bunch of mathematical rotations on a cubic. And if you can figure out how to rotate a collection of Cubics with a certain touch, let's just say, you can make some really magical things happen.

60
00:08:59.693 --> 00:09:03.923
But this is not about to be a walk in part.

61
00:09:04.104 --> 00:09:05.094
But then again,

62
00:09:05.124 --> 00:09:05.724
I mean,

63
00:09:05.754 --> 00:09:07.254
I used to edit for physics girl,

64
00:09:07.283 --> 00:09:20.183
I even edited both of the episodes she did on quantum mechanics and what I remember from those experiences is that 2 States particle wave.

65
00:09:21.443 --> 00:09:24.653
Okay. All right. Fine. I'll do my research.

66
00:09:27.599 --> 00:09:42.024
Okay, okay okay. All right we're now about 3 weeks and I still really haven't done anything with the quantum computer yet, other than just watch numerous videos and articles on quantum mechanics and do about a 1M. Hello?

67
00:09:42.024 --> 00:09:55.913
Constant world projects with kiss gets resources. I know. Have a much better understanding on both quantum mechanics and content repeating and I'm officially ready to begin pulling my hair from all the quantum bugs we will eventually create. Ha. Ha.

68
00:09:56.004 --> 00:10:04.823
So look, we need to start here. This is a single house that is generated using the Minecraft method and if we were to generate more in this area, we would have a small town.

69
00:10:05.099 --> 00:10:19.734
Generated world, everyone. Cool. Great. I mean, it gets the point across what you see, Minecraft generates houses by selecting 1 of a few different houses that were all Pre made by humans and they only replace a house on the terrain. If the terrain is flat enough.

70
00:10:19.943 --> 00:10:25.793
Now, this method is quite common in video games, because it's very simple, but also very effective.

71
00:10:25.913 --> 00:10:39.744
But why don't this method is because once you've seen a couple of generated houses, Minecraft, you've basically seen not to mention this method, makes the terrain, the most important thing over everything else because that's simply what's generated.

72
00:10:39.744 --> 00:10:46.823
1st, which means if the terrain is no good. Then the town will be no good as well. We all know D***.

73
00:10:46.823 --> 00:10:47.063
Well,

74
00:10:47.063 --> 00:10:48.744
there's some rough terrain is not going to stop,

75
00:10:48.744 --> 00:10:53.004
man from conquer or can I say humans are okay but most importantly,

76
00:10:53.094 --> 00:10:53.933
this method,

77
00:10:53.933 --> 00:10:58.224
as far as I can imagine is not compatible with a quantum computer,

78
00:10:58.313 --> 00:11:03.144
trying to figure out how to write a quantum program to generate this may be extremely hard,

79
00:11:03.144 --> 00:11:04.344
if not impossible.

80
00:11:04.673 --> 00:11:13.764
Okay. Listen, in the middle of editing, this video is actually proven to work. Dr Jim, who was actually my point of contact for all the technical quantum computing stuff.

81
00:11:13.764 --> 00:11:27.864
Just published an article detailing how he helped a game dev team procedually generate terrain using a quantum computing simulator. Very cool. I'm no expert. I still fully understand how quantum algorithms work, so you should take everything.

82
00:11:27.864 --> 00:11:42.144
I'd tell you with grain of salt, but back to the video, or we need instead is an algorithm that treats the entire map as 1, singular piece. It treats the map as a true quantum object in which the terrain plants houses, build roads. You name it.

83
00:11:42.144 --> 00:11:45.173
I'll have equal value in the eyes of the generation algorithm.

84
00:11:45.269 --> 00:11:48.509
All working together to make it seamless, but believable map.

85
00:11:48.509 --> 00:11:52.619
Thankfully, we don't have to search far for this algorithm.

86
00:11:53.729 --> 00:12:07.889
There is a new, very magical generation algorithm called the wave function collapse algorithm, created by computer scientists named maximum groom in 2016 and I personally believe it to be the ultimate generation algorithm of our time.

87
00:12:07.889 --> 00:12:17.879
We can feed it a data set to learn from and it will generate whatever we want inspired by that data, which makes it a machine learning algorithm. But from my experience is.

88
00:12:17.879 --> 00:12:32.009
With a lot more control, this is how it works. This map is in what's called superposition. Now superposition is something that happens in quantum mechanics, which is basically means that this map currently exists as every possible map imaginable.

89
00:12:32.009 --> 00:12:36.509
From the randomly assembled maps all the way to the seamless cohesive maps.

90
00:12:36.509 --> 00:12:42.839
Here's a bit of idea what a mapping superposition would look like. No, we can never truly visualize a quantum object.

91
00:12:42.839 --> 00:12:57.264
Maine Einstein called it spooky for a reason anyhow, with the wave function collapse algorithm we simply create X amount of states in this example. It's only 3 the dirt state, grass state and nothing state. Then we give each state very simple rules that they must follow.

92
00:12:57.293 --> 00:13:05.903
For example, the grass state must have a nothing stay above it must have a dirt state under it. It must have another grass state on every side of it.

93
00:13:05.933 --> 00:13:12.953
And when we give rules to all of our States and tell it to generate, it will remove all impossible States from the neighbors at this position.

94
00:13:13.048 --> 00:13:23.969
Which then will make their neighbors remove impossible States from their possible States and this keeps on going until all positions on the map have exactly. 1 possible state. And.

95
00:13:23.969 --> 00:13:33.629
Walla you have a map generated with the aid of quantum mechanics. However, this algorithm doesn't become actual magic until you have a giant list of possible States.

96
00:13:33.629 --> 00:13:38.908
Create a loan data set and feed it to the algorithm to generate beautiful houses. Like you see here.

97
00:13:38.908 --> 00:13:44.879
It's so beautiful. In fact, I think we need to take a quick montage while I go cry in the corner from its beauty.

98
00:13:50.124 --> 00:14:03.714
But now, what really drops jaws with this machine learning algorithm is that we get to do things like this. We can simply use the same possible States as before, but change the dataset from this house that requires flat land to this house.

99
00:14:03.714 --> 00:14:12.683
That is built into a hill feed it to the algorithm, and it will learn from our data set and generate a whole new type of house. I have no words.

100
00:14:12.833 --> 00:14:22.614
And, of course, there's absolutely nothing stopping us from increasing the size of our map to generate a beautiful, completely seamless town like this.

101
00:14:29.573 --> 00:14:30.053
You'll see,

102
00:14:30.053 --> 00:14:30.833
in this now,

103
00:14:30.833 --> 00:14:33.443
initially I plan to add a bunch of different states,

104
00:14:33.653 --> 00:14:48.533
so that the algorithm can make things like parks and backyards and rooftops and restaurants and stores parking lots things like this but under estimated just how long it takes to get this whole system set up this single data set map took me some good

105
00:14:48.533 --> 00:14:49.943
hours to put together.

106
00:14:50.033 --> 00:15:00.354
Mostly detailed verbal that map editor I made was but, I mean, there's nothing stopping me from adding more and more state to this currently. It's just this is as far as I was able to get for this video, but.

107
00:15:02.663 --> 00:15:12.384
I have to come clean. I have not been honest with you guys and gals generating the individual houses. Isn't that big of a deal? It took maybe 5 to 10 minutes each.

108
00:15:12.413 --> 00:15:26.004
But every town that I've shown you a 2nd, a minimum of 2 hours to generate imagine buying a new game Minecraft, and then having to wait that long every time you want to play a new game, most people would want their money back. So quick.

109
00:15:26.214 --> 00:15:29.903
It's stuck on this screen for, like, 1015 minutes now.

110
00:15:31.139 --> 00:15:43.469
How long after way, every single time a quantum computer will be able to reduce this generation that takes 11 hours down to a few seconds for minutes. Like something you'd expect with mind credit world generation.

111
00:15:43.469 --> 00:15:55.374
The only problem is, I tried for a solid week straight, and I'm talking sunup to sundown, trying to turn this regular algorithm into a quantum algorithm. And it was just too difficult for me to do in that time span.

112
00:15:55.464 --> 00:16:05.634
So, I instead how to wait forever and generate those complex worlds on a regular computer. But I still believe that this is possible to do with the quantum computers the IBM has running today.

113
00:16:05.724 --> 00:16:17.094
In fact, I'm still trying to figure it out in my free time, because it's quite a fun challenge but I have no idea just how long is going to take me figure that out. If it's even possible at all. And the show how to go on.

114
00:16:17.369 --> 00:16:29.933
So, hey, gotta be some slack guy, but knowing what I know about computers, I am very excited about how they might shake things up in the years to come. Sure. There's potential threats to our cyber security that quantum computers might be able to destroy.

115
00:16:29.994 --> 00:16:43.673
And we probably should focus on that, and there's also quantum drug development that might save a lot of lives, but just imagine using a computer to generate a new and unique, fortnight size map with all its detail at the click of a button.

116
00:16:44.038 --> 00:16:56.548
All right, I want to give her huge thanks to kids learning about quantum computing has been a whole lot of fun and I cannot wait to see where they take this technology. Quantum computing is still very.

117
00:16:56.548 --> 00:17:11.003
Very new, and just like, no networks in the early days it's working on trying to figure out. Exactly where, and when it will be useful so shelf solid, quantum computing researchers out there. If you want to support me in this channel. Please go check out kiss, gets you to channel there.

118
00:17:11.003 --> 00:17:18.534
You can learn all types of stuff on how to program for their concept computers. It was a grail for me when I got started and more resources, or in the description.

119
00:17:20.338 --> 00:17:24.868
Oh, okay. I hope you had fun with that. So.

120
00:17:27.778 --> 00:17:32.999
To where did I put it here?

121
00:17:36.328 --> 00:17:40.288
Started up again, just a 2nd to.

122
00:17:43.344 --> 00:17:55.943
Okay, well, 1 thing I forgot to mention before we so I forgot to started to talk about the final project. So this is what we're looking at tentatively for the rest of the class.

123
00:17:56.189 --> 00:18:06.778
3 more guest speakers probably and to the extent, they don't take up the full class. I'll fill it in with other stuff. Then the last 3 classes you guys get to talk.

124
00:18:06.778 --> 00:18:09.929
So trying to talk about the final project.

125
00:18:09.929 --> 00:18:13.078
Syllabus mentioned to.

126
00:18:13.078 --> 00:18:21.298
2 rounds of quick student presentations, but was 19 students. There's no time for that. So I'm just had the 1 round, which I called homework. 3.

127
00:18:21.298 --> 00:18:33.989
But we'll put it in the grade separately and then so work a little harder on the final project. So, the final project do it in teams of up to 3 students. By the way I've noticed you guys actually don't like working in teams, but that's okay.

128
00:18:35.874 --> 00:18:49.314
Deliverables will be a video presentation to class 1 of the last 3 class days. So this is, you know, proximate lined up. So we partition it. So, 6 students a day or something.

129
00:18:49.588 --> 00:18:54.568
7 students on 1 day, so something like that.

130
00:18:54.568 --> 00:19:04.499
And the deliverables, so you give a presentation, you could might be easier to prepare a video actually, and to show the video.

131
00:19:04.499 --> 00:19:11.038
And then an 8 page paper, describing the project, and written in the style of an Tripoli.

132
00:19:11.038 --> 00:19:24.328
Conference journal paper inside as a PDF file the last day and for the talk, you can email me just your team members and preferred dates.

133
00:19:24.328 --> 00:19:28.739
And it's helpful if you put quantum in the subject line or something.

134
00:19:28.739 --> 00:19:32.939
And then.

135
00:19:32.939 --> 00:19:36.148
Oh, the topic that's the least important part of it.

136
00:19:36.148 --> 00:19:40.409
Something in relating to quantum computing, so.

137
00:19:41.578 --> 00:19:48.929
And I'll entertain questions and so on and think about that. So, this would be so you'll notice that you'll be giving your talk.

138
00:19:48.929 --> 00:19:57.058
Before you hand in your paper, your paper is like the last legal day, which is the last class day and.

139
00:19:57.058 --> 00:20:03.659
I think it's a Friday or something, and but you'll be giving talks over the previous week and a half.

140
00:20:03.659 --> 00:20:08.368
Any questions on that at the moment? No. Okay.

141
00:20:08.368 --> 00:20:12.058
So, I want to go back to.

142
00:20:13.769 --> 00:20:18.088
A Marius casino, because she's such a good lecturer and I'll start around 10.

143
00:20:18.088 --> 00:20:25.259
Or something, that's where we ended up so.

144
00:20:30.568 --> 00:20:36.538
A back a minute or 2 here. Yeah.

145
00:20:41.878 --> 00:20:49.648
Okay, logical keep it manual, but that's still pretty far away.

146
00:20:49.648 --> 00:20:54.298
There's another interesting spot though. We know. Is that.

147
00:20:54.298 --> 00:21:01.318
Your laptop can simulate and emulate in essence, a small group of but.

148
00:21:01.318 --> 00:21:09.953
Once you get to, around 15 Cubans, there is no classical computer that can actually simulate that system. So it's such a system if you can control it.

149
00:21:09.953 --> 00:21:21.713
And if those are high performance Cubans, they reach into a space that no other tool can reach, which makes them very interesting. And we could imagine that there may be some interesting applications between.

150
00:21:21.959 --> 00:21:36.358
That 50 Cuban area, and the known useful, Universal fault, tolerant, quantum computer added a 1M cubes that means or that suggests that this 50 inch cubic range would be a good place to stick our prototype. And so indeed.

151
00:21:36.358 --> 00:21:47.909
For our prototype, we decided we would aim for something around 50 units, and we would make it a 2 dimensional array so that it would be forward compatible with the air correction schemes would like to pursue in the future.

152
00:21:47.909 --> 00:21:57.058
Then, of course, we need to have a good interface with the people who would like to use this prototype and and its successors. So.

153
00:21:57.058 --> 00:22:05.878
As, as I mentioned, there's this intermediate MSC error, we call that noisy, intermediate scale, quantum regime and.

154
00:22:05.878 --> 00:22:10.979
That's a space before we've reached a fully fully are corrected system.

155
00:22:10.979 --> 00:22:17.338
But where we could hope that there is some interesting algorithms to be developed and discovered and.

156
00:22:17.338 --> 00:22:20.788
By necessity, these algorithms will be hardware specific.

157
00:22:21.838 --> 00:22:30.179
The the universal fault, tolerant, quantum computer, you can imagine some level of abstraction for your interactions with it. But during the nisc era.

158
00:22:30.179 --> 00:22:35.848
And the algorithm developer really needs to be quite aware of the specific strengths and limitations.

159
00:22:35.848 --> 00:22:45.203
Of the process that they're developing the algorithm for you, you can't be trying to make 2 Cubans on opposite sides of the ship interact with each other. For example, that that wouldn't work.

160
00:22:45.413 --> 00:22:52.284
So we need a way for hardware and algorithm developers to handshake with each other.

161
00:22:52.558 --> 00:23:05.159
And to do that, we started developing this platform called Cirque. It was released a couple years ago at this point and as an open source Python framework for these nisc era algorithms that.

162
00:23:05.159 --> 00:23:18.838
Force and enable algorithm developers to be quite aware of the specific hardware they're developing for complete with a little ASCII art to help to see what the algorithm is that you're developing. So.

163
00:23:18.838 --> 00:23:25.679
Um, that was projects that we developed to facilitate this interface with all the developers.

164
00:23:25.679 --> 00:23:34.769
Now, let's forge ahead into the prospect of actually building this prototype building. There's 50 cubic 2 D Ray and.

165
00:23:34.769 --> 00:23:40.108
Demonstrating system wide control so I.

166
00:23:40.108 --> 00:23:45.689
Nathan meant to ask you if people have questions, do they just blurt them out? Or how, how do you do this?

167
00:23:47.759 --> 00:24:00.509
Yeah, typically people would just unmute themselves and say, hi, Marissa and ask a question, but feel free to pause as well if appropriate apologies I didn't mention that.

168
00:24:00.509 --> 00:24:03.628
That's okay. I.

169
00:24:05.939 --> 00:24:11.818
I will, I will certainly take questions at the end and if people have questions in the middle, please feel free to, uh, to ask them.

170
00:24:13.378 --> 00:24:24.929
With that before, I don't know how much familiarity that's audience has with superconducting microwave, keep it. So obviously some of you will be quite familiar with them, but I don't know.

171
00:24:24.929 --> 00:24:32.969
I, I think that an introduction is fine there. You know, the audience ranges from.

172
00:24:32.969 --> 00:24:37.648
Our computer scientists, 3, 2 hardware specialists so they're going to be, um.

173
00:24:37.648 --> 00:24:45.509
Um, a wide range. Great then we'll just do this very, very brief introduction to superconducting microwave Cuba.

174
00:24:45.509 --> 00:24:51.689
It's easy to think about this by starting from a more familiar. A friend.

175
00:24:51.689 --> 00:24:56.608
The LLC harmonic also later and this guy has.

176
00:24:56.608 --> 00:25:01.288
What we call a parabolic potential that means in essence that the energy.

177
00:25:01.288 --> 00:25:05.308
Levels of resonance will be equally spaced.

178
00:25:05.308 --> 00:25:15.328
With a fixed resident frequency, and this is not a cubic you could imagine trying to assign levels 0, 1, 2, 3 and so on.

179
00:25:15.328 --> 00:25:20.189
To those different energy, if you could imagine trying to assign.

180
00:25:20.189 --> 00:25:32.638
Computational identities to those different levels, but because they are all evenly spaced, you have no control of where you are in the ladder where you're going up or down. So what we would like, instead is a.

181
00:25:32.638 --> 00:25:43.949
A different kind of oscillator with a different potential. This isn't an harmonic oscillator. And what we've done is basically just replaced that indicator in our harmonic oscillator with a nonlinear doctor.

182
00:25:43.949 --> 00:25:49.019
That gives this whole thing, a CO sign potential, which means it pulls the levels or pushes then.

183
00:25:49.019 --> 00:26:00.538
Part into an even spacing and now you can see that the, the difference between the 0T and the 1 level has a different energy gap and between the 1 and the 2 level and this allows us to control which state we're in.

184
00:26:00.538 --> 00:26:04.528
And effectively restrict ourselves to something like a 2 level system.

185
00:26:04.528 --> 00:26:09.568
Which is then our Cuban so in this particular setup.

186
00:26:09.568 --> 00:26:20.608
The quantum information is represented by the amount of energy stored in the Cuba. So, the difference between being in a 0T state, or the 1 state is whether or not, you have.

187
00:26:20.608 --> 00:26:26.009
1, photon at around 5 gigahertz sloshing around inside of your also later.

188
00:26:26.009 --> 00:26:29.159
And to come back to that in a bit.

189
00:26:29.159 --> 00:26:38.368
How do we physically represent this? And so we've got some physicist model, but how do we really make it in the lab? We make.

190
00:26:38.368 --> 00:26:42.028
Our Cubans out of aluminum on Silicon substrate.

191
00:26:42.028 --> 00:26:46.318
And pattern it in a cleaner and facility this.

192
00:26:46.318 --> 00:26:58.108
Capacitor is of this part that I'm highlighting in orange here is the capacitance between that center island and the ground playing around the outside. So the light color is aluminum. And the dark outline is the Silicon that you can see.

193
00:26:58.108 --> 00:27:09.239
Which has been left after the aluminum pattern was established, and then this nonlinear element, this squid, which is 2, Joseph junctions to.

194
00:27:09.239 --> 00:27:12.959
Nonlinear elements are in that blue box.

195
00:27:12.959 --> 00:27:25.528
So, that's how we physically build our documents and we also need to build some of these standard and we build those often actually, by just making a very long.

196
00:27:25.528 --> 00:27:29.009
String if you will of transmission line, so.

197
00:27:29.009 --> 00:27:34.798
In the same way that you could imagine a string being bound on 1 side and.

198
00:27:35.124 --> 00:27:49.733
Having certain modes that it supports, you can have a transmission line resonator up operating much the same way. And that supports a certain set of harmonics and we will pick 1 of those to be our residents for these resonators.

199
00:27:49.733 --> 00:27:53.844
So, what that ends up looking, like, on a chip is something like this.

200
00:27:54.058 --> 00:28:05.788
You have your Cuban actually, each Cuban needs to have an associated for the process of measuring it state and then we also have some control lines that drive expectations in the.

201
00:28:06.624 --> 00:28:20.663
Centralize that drive read out and as you can see, this is all filling up quite a bit of space on the chip. And I mentioned before that our goal was to make a 2 D array. Right now we've got a 1. we've got a linear array and we were already filling up quite a bit of space.

202
00:28:20.663 --> 00:28:33.203
So we needed to do in order to make a 2 dimensional array is find some way to cram all of this into a unit sell that we can then tile across. And you can tile as many times as you want, making more.

203
00:28:33.203 --> 00:28:47.814
Cubans is as simple as copy, pasting that unit cell assuming you've made a good enough unit selling the 1st place. And so, what we've actually done was pull the control wiring and these. So, the yellow and red.

204
00:28:48.058 --> 00:28:52.318
Onto 1 shift and the Cubans onto a separate chip.

205
00:28:52.318 --> 00:28:55.348
Now, we have 2 chips and we can switch them together.

206
00:28:55.348 --> 00:29:02.729
Connect them using, like, an Indian bump on process, for example, and that allows us to.

207
00:29:02.729 --> 00:29:13.259
Cram more we have sort of a quasi 3 dimensional system at that point and this that allows us to cram our readout and our control wiring into more space.

208
00:29:13.259 --> 00:29:21.719
Sorry into less space, close to the queue as and allow us to, to build this unit cell that we can tile. So we do all of that fabrication.

209
00:29:21.719 --> 00:29:32.038
We make this chip sandwich that you can see here, the small square is the Cuban chip on top of our carrier chip with the control wiring. Then we put it into a package.

210
00:29:32.038 --> 00:29:43.558
And that's because, of course, you have cables that come down your Christ, and they have connectors on the ends and you want to connect it to a Silicon ship, which does not have connectors on the end. So that the packages that fan out interface between those 2.

211
00:29:43.558 --> 00:29:52.229
You mount the whole thing on the base of dilution refrigerator and you pull it down to 2000 only Kelvin and then, as I sometimes like to say the real work begins and that that's a little bit of.

212
00:29:52.229 --> 00:30:03.328
It's not quite fair to say that of course, there's an insane amount of work that goes into just producing the thing on this slide but just because you've built your system with.

213
00:30:03.328 --> 00:30:07.588
50, some does not by any by any means.

214
00:30:07.588 --> 00:30:16.679
Mean that you have a full working system of 50, some, Cuba's to do that. We need to spend quite a bit of time with a process called calibration.

215
00:30:16.679 --> 00:30:21.028
And a very simple level actually, this, this was an analogy.

216
00:30:21.028 --> 00:30:33.179
That 1 of my colleagues likes to use he's a very good musician. And so I've borrowed this from him because I just, I like this analogy a lot. You can think of it as though we are learning to play music for our.

217
00:30:33.179 --> 00:30:37.919
Just, as a musician is an expert.

218
00:30:37.919 --> 00:30:48.568
With their instrument, they're very capable of producing exactly the sound waves that they want so that the listener will hear a particular note.

219
00:30:48.568 --> 00:30:54.659
Similarly, the process of calibration is figuring out how to play our control electronics.

220
00:30:54.659 --> 00:31:03.388
So that they will produce exactly the correct electrical waves and cause the Cubans to hear, or feel exactly the gate or the manipulation that we want.

221
00:31:03.388 --> 00:31:07.558
So, we send some electronic pulse out of our control electronics.

222
00:31:07.558 --> 00:31:16.348
Which then causes the Cuban to experience a rotation around the blocks here, for example, and that is a quantum 8.

223
00:31:16.348 --> 00:31:19.348
We put those gates together into an algorithm.

224
00:31:19.348 --> 00:31:26.009
And so, these little blocks that you see, the White blocks are single cubic Gates and the blue ones are 2 cubic Gates.

225
00:31:26.009 --> 00:31:29.489
All coming from these well, calibrated electrical pulses.

226
00:31:29.489 --> 00:31:39.929
So, the process of actually figuring out what each electrical pulse looks like is not trivial. And in particular, when you have lots and lots of them, it becomes.

227
00:31:39.929 --> 00:31:48.659
A lot of lot of work, this is a graph that we've drawn to represent the process of bringing up of calibrating.

228
00:31:48.659 --> 00:31:57.989
The electrical signals that we need to control 1 single cubic, and we, during this as a graph, because the only way to manage.

229
00:31:57.989 --> 00:32:04.078
Calibration of 50 some is to do it automated. You can't. There's, there's no.

230
00:32:04.078 --> 00:32:07.979
Time and not enough patients in the world.

231
00:32:07.979 --> 00:32:19.288
To bring up all of those Cubics manually with a person in front of an telescope and spectrum analyzer you just you can't do it. So we needed to to write software to teach.

232
00:32:19.288 --> 00:32:32.429
The software how to calibrate, and each of these dots, each of these colored dots is, in essence, like an experiment and the results of 1 experiment, enable us to proceed to the next 1.

233
00:32:32.429 --> 00:32:36.298
So, we can eventually traverse this entire graph and.

234
00:32:36.298 --> 00:32:40.078
Come out at the end with knowledge of what.

235
00:32:40.078 --> 00:32:43.199
Each gate, we might need on that single cube that looks like.

236
00:32:43.199 --> 00:32:55.138
And then we want to check. So, is the gate doing that? The Gateway turned up? Is it doing what we thought? We thought it should do 1 way of checking. This is what we call cross entropy, benchmarking.

237
00:32:55.138 --> 00:33:08.009
And for a moment, just to think of this f, as an accuracy score, the cross cross entropy, benchmarking fidelity. But it's an accuracy score where 0T means really bad. And 1 is ideal.

238
00:33:08.009 --> 00:33:13.588
And the way we would exercise it really we would use it is, for example.

239
00:33:13.588 --> 00:33:22.138
We would run a date that we wish to run on a cubic and we measure the outcome that will be either 0T or 1. that's how quantum mechanics works now.

240
00:33:22.138 --> 00:33:27.328
We should know because we know what that gate should be. We know what.

241
00:33:27.328 --> 00:33:37.378
Our expected outcome is we know how many what fraction of the time we expect to measure 0T and what fraction of the time we expect to measure 1. so we can compare what the quantum.

242
00:33:37.378 --> 00:33:40.499
It actually does with what it should do.

243
00:33:40.499 --> 00:33:52.134
And that will allow us to derive this accuracy score. We can do that with all sorts of different gates that we might like to run on that Cuba and figure out for each 1 what is the accuracy score?

244
00:33:52.463 --> 00:33:55.733
We can start running multiple Gates 1 after the other.

245
00:33:56.009 --> 00:33:59.278
And so we, we run and lots of Gates.

246
00:33:59.278 --> 00:34:11.699
Again, we have to know in each case what the expected outcome will be, then we can compare it with what we actually get and we can see how well as the quantum computer actually doing. And of course, we can do this with 2 cubic Gates as well.

247
00:34:12.838 --> 00:34:24.898
So, I should also mention that the process of tuning a 2 cubic gate is, of course, even more complicated than the 1 cubic gate. So, here's the graph for a 2 cubic gain.

248
00:34:24.898 --> 00:34:34.018
And now keep in mind that we have to do single cubic gates for all 54 of our cubes. And then each pair of neighboring.

249
00:34:34.018 --> 00:34:41.548
Um, again, as I said, before, we wanted to automate this, because it's really just too much work to try to do by hand.

250
00:34:41.548 --> 00:34:49.139
And my 1 colleague, who was who, who built this graph in the 1st place.

251
00:34:49.139 --> 00:34:58.318
Sometimes, this, this 1 particular pair of vertices was actually his pH. D. and so he likes to I joke. So that.

252
00:34:58.318 --> 00:35:11.278
Now, this algorithm is doing these pH. D. multiple times automatically every time we bring up Cubans, we need the automation. There's just no way around it. And so, 1 thing that's.

253
00:35:11.278 --> 00:35:14.398
It's not enough to have.

254
00:35:14.398 --> 00:35:24.898
Your best effort of the day I need the Cubans to be a little bit better calibrated. Well, I'll just work a little harder today that doesn't work when everything's automated. So you have to make sure that you write good software.

255
00:35:24.898 --> 00:35:28.378
And, in fact, we have.

256
00:35:28.378 --> 00:35:38.188
Managed to get this whole system working so that the colored axes are the color denotes the error of the single cubic Gates.

257
00:35:38.188 --> 00:35:49.168
For each of those documents, and the little boxes between them, those little rectangles denote the error of the 2 pubic Gates between those pairs of neighboring Cubans.

258
00:35:49.168 --> 00:35:59.338
And as you can see, we've got some pretty impressive error rates and I should mention that not only are these error rates. Very good.

259
00:35:59.338 --> 00:36:07.289
For just single gains, considered an isolation. These are actually measured with the entire chip running at the same time.

260
00:36:07.289 --> 00:36:13.978
And that's important as well because you have the whole chip, you need to be able to run it at the same time. So.

261
00:36:15.028 --> 00:36:23.099
This this value is referring to the way this cubic performs while all of its neighbors also are expected to be performing.

262
00:36:23.099 --> 00:36:27.719
And that's a critical piece of our error measurement.

263
00:36:27.719 --> 00:36:31.619
So, with that we've.

264
00:36:31.619 --> 00:36:35.398
Built this quantum system, we have verified it in some.

265
00:36:35.398 --> 00:36:47.458
I deputized way I guess so. We've looked at each of these Cubans individually, but we would like to still see this thing behave as a big quantum system, a big system.

266
00:36:47.458 --> 00:36:51.898
With the system that is quantum with a lot of constituent parts.

267
00:36:51.898 --> 00:36:55.829
And so we put it all together.

268
00:36:55.829 --> 00:37:02.099
And now we would like to identify a small milestone, computational task that we can have.

269
00:37:02.099 --> 00:37:11.579
This thing do, and this highlights that we're really trying to push into this, this regime we, and we want to really be a classical computer at something.

270
00:37:11.579 --> 00:37:16.829
I'm not going to say it's going to be something interesting, but it is there's going to be something useful.

271
00:37:16.829 --> 00:37:20.668
But it is going to be something interesting in so far as.

272
00:37:20.668 --> 00:37:29.938
It explores this hypothesis, or this thesis in computer science that, in essence advocates.

273
00:37:29.938 --> 00:37:33.568
And a classical computer are.

274
00:37:33.568 --> 00:37:39.539
Efficiently they can efficiently simulate each other. So, in, in a computational perspective.

275
00:37:39.539 --> 00:37:44.219
They are in some ways the same, but a quantum computer.

276
00:37:44.219 --> 00:37:58.679
Sits outside of that, a quantum computer is a fundamentally different animal from the Abacus or the traditional classical computer. And so we'd like to kind of push in that direction. And to that end, we describe.

277
00:37:58.679 --> 00:38:04.528
Or we define this milestone computational task called quantum supremacy.

278
00:38:04.528 --> 00:38:07.829
We say it's a well defined computer science milestone.

279
00:38:07.829 --> 00:38:19.409
And it can be achieved when a quantum computer performs, but ask, that may be a contrived to task. But a task that would take too long, or would require too much resources.

280
00:38:19.409 --> 00:38:31.050
On any classical computing machine and okay, so we've got this little prototype. We would like to use it to demonstrate supremacy. How should we do that?

281
00:38:31.050 --> 00:38:35.159
Uh, well, an ideal task would be something that we could solve with.

282
00:38:35.159 --> 00:38:38.280
Using that quantum hardware using that that prototype.

283
00:38:38.280 --> 00:38:50.730
Should be something that's very difficult to solve on a classical machine. So the classical machine should struggle and the quantum machine should succeed and ideally, we would also be able to check whether the quantum machine has.

284
00:38:50.730 --> 00:38:57.659
Done the task correctly, using classical hardware. So, for example, it would be nice if we could use just a full task.

285
00:38:57.659 --> 00:39:05.519
Like, throwing something that we know quantum Peter, something good at and of course, it's very easy to check.

286
00:39:05.519 --> 00:39:12.239
The computers were still defender correctly. Well, just mobile 2 factors that gave you together and see because you get the.

287
00:39:12.239 --> 00:39:22.110
Quotient that I told you that you should get the product that it told you, you should get another 1. it would be easy to check is function and version.

288
00:39:22.110 --> 00:39:34.829
Unfortunately, the quantum harder that we have today isn't yet sophisticated enough to run these kinds of algorithms. And so we need to revise our hopes a little bit for what this task for demonstrating quantum supremacy might look like.

289
00:39:34.829 --> 00:39:46.889
Of course, the 1st, 2 are non negotiables it needs to be solved Harper. It needs to be very difficult to solve the hospital, but we can wait a little bit on this easy check using classical hardware.

290
00:39:46.889 --> 00:39:56.130
Yes, your audio is breaking up a tiny bit at the moment. Maybe. Would you like to turn off your video and see if that to.

291
00:39:56.130 --> 00:40:00.420
It says that I can do that.

292
00:40:00.420 --> 00:40:04.710
Silence.

293
00:40:04.710 --> 00:40:09.900
I'm not sure what else to try. So, let's that guys.

294
00:40:11.400 --> 00:40:16.199
Thanks.

295
00:40:18.059 --> 00:40:21.960
So, what this computational task.

296
00:40:21.960 --> 00:40:29.250
Look like then making that concession well, we can consider the following computational task.

297
00:40:29.250 --> 00:40:35.219
And it almost feels a little bit obvious, because it's exactly the thing that quantum computers are good at doing.

298
00:40:35.219 --> 00:40:41.550
We'll say that the task is to sample the output distribution from a carefully chosen quantum circuit.

299
00:40:41.550 --> 00:40:47.070
That is, we will pick some gates that that should be run on a quantum computer.

300
00:40:47.070 --> 00:40:55.949
And please predict the output distribution, the chance of measuring each of these strings of 0T and 1.

301
00:40:55.949 --> 00:41:07.110
In the output now, why would we pick such a task? Well, okay. 1st of all it fulfills that 1st requirement. You can solve it with quantum hardware simply by running the algorithm.

302
00:41:07.110 --> 00:41:15.690
Gave a choice on the quantum hardware, but it turns out it's also interesting because it has been carefully analyzed in.

303
00:41:15.690 --> 00:41:19.440
Classical complexity theories, it's actually a well study a problem.

304
00:41:19.440 --> 00:41:26.369
That if we choose these gains, we choose the identity of each of these single human Gates from certain distribution.

305
00:41:26.369 --> 00:41:35.489
That will guarantee that this problem is very difficult for a classical computer to solve. And in fact, if we choose these gates randomly.

306
00:41:35.489 --> 00:41:39.510
We, we have a task with high computational complexity.

307
00:41:39.510 --> 00:41:42.960
That our quantum computer should be able to.

308
00:41:42.960 --> 00:41:52.289
Performing, so, what does what is the process for demonstrating supremacy look like with this task? Well, the 1st thing you do is you choose a circuit.

309
00:41:52.289 --> 00:42:00.780
So, we need a random number generator somewhere. It generates the identity of each of these different dates. We write that down. That's our input to the program.

310
00:42:00.780 --> 00:42:09.480
And now we want to sample the output distribution. We have a quantum computer on the 1 hand, and we have a classical computer on the other hand.

311
00:42:09.480 --> 00:42:19.110
Um, so the best strategy for the quantum computers, of course, just to run that circuit for the classical computer, the best strategy known is actually to simulate quantum mechanics.

312
00:42:19.110 --> 00:42:24.539
And, of course, now you are trying to simulate keeping track of zillions of amplitude.

313
00:42:24.539 --> 00:42:27.539
That blows up pretty quickly and.

314
00:42:27.539 --> 00:42:40.110
Give some indication as to why this is hard for a classical computer and then finally, after we, after we do this, we can determine the classical cost of the work that that quantum machine did.

315
00:42:40.110 --> 00:42:48.570
Because we just saw how hard it was for the classical computer to do the same thing. If that cost is too high we have achieved 1 of supremacy.

316
00:42:48.570 --> 00:42:58.380
Well, if the cost is not too high, then we haven't achieved quantum supremacy, but it's also possible for us to check that. The quantum computer is doing what it's supposed to do.

317
00:42:58.380 --> 00:43:04.289
Now, on the cost does get to be too high, then we can send you achieved quantum supremacy. But at that point.

318
00:43:04.289 --> 00:43:09.119
We can't check that. The quantum computer is performing incorrectly anymore. So.

319
00:43:09.119 --> 00:43:19.440
To actually convince ourselves that the quantum computer is still performing correctly inside of the supremacy regime. We need to do some verification and gymnastics and.

320
00:43:19.440 --> 00:43:25.440
This is to convince ourselves basically that the quantum computer is running in the way we expect it to.

321
00:43:25.440 --> 00:43:31.260
Basically, what we do for these verification gymnastics is on the on the 1 hand.

322
00:43:31.260 --> 00:43:35.340
We model the errors, so we have all of these, these.

323
00:43:35.340 --> 00:43:41.369
Error models that we we saw on that previous slide. We know the single Cuban into cubic errors.

324
00:43:41.369 --> 00:43:52.289
That were measured when we brought up those gains, and we can compare that using an error model with the data we see. And verify that. Yes the errors we measured.

325
00:43:52.289 --> 00:43:56.340
Are representative of the way the system is performing at scale.

326
00:43:56.340 --> 00:44:06.000
And then, in addition we can run some simplified circuits simplified in such a way that it makes it much easier for the classical computer to track.

327
00:44:06.000 --> 00:44:16.860
And to keep up with the quantum computer, but ideally simplified in such a way that it doesn't really make the quantum computers job any easier. And then we can compare and and see.

328
00:44:16.860 --> 00:44:26.760
How well is the quantum computer doing on the simplified circuits and have some projection as to how well, it's probably doing on those big full circuits that we can't actually compare to.

329
00:44:26.760 --> 00:44:30.090
So, let's look at that modeling for a moment.

330
00:44:30.090 --> 00:44:33.360
We wrote down a pretty simple.

331
00:44:33.360 --> 00:44:37.349
Fidelity error model for.

332
00:44:37.349 --> 00:44:41.699
How good should the fidelity of a multi Cuban be.

333
00:44:41.699 --> 00:44:49.710
Based on the fidelity of the individual, single and 2 cubic Gates and of the readout, our measurement fidelity and.

334
00:44:49.710 --> 00:45:00.659
Then we were able to build up this model so, on the Y, axis, we have the fidelity that is the chance that the quantum computer is performed without an error.

335
00:45:00.659 --> 00:45:04.559
And then we just expand to.

336
00:45:04.559 --> 00:45:09.269
Up to the entire chip, so we take a little subsection and we.

337
00:45:09.269 --> 00:45:14.099
Calculate all of those errors together, and we get a point on our model and we grow that section and we.

338
00:45:14.099 --> 00:45:23.940
Draw out this whole model. Okay. And then, of course, the next thing to do is to check whether the model is in fact, correct because of course, this model only considers individual.

339
00:45:23.940 --> 00:45:29.849
Cuban single Cuban into cubic Gates so if there's any additional error mechanisms that show up from.

340
00:45:29.849 --> 00:45:34.980
The fact that all of these keywords are sitting together on the same s***. We haven't included that in our model yet.

341
00:45:34.980 --> 00:45:40.619
And how do we check this? Well, we can actually use this same.

342
00:45:40.619 --> 00:45:46.769
Or score, accuracy score that we use for the single human Gates. Now, we just use it for.

343
00:45:46.769 --> 00:45:51.300
The larger circuits altogether we, we try to run a circuit.

344
00:45:51.300 --> 00:45:54.510
We measure the bit strengths that the quantum computer gives us.

345
00:45:54.510 --> 00:46:00.420
And then we feed all of that information, both the identities of the gates that we used, and the strings that we measured.

346
00:46:00.420 --> 00:46:04.260
Into a classical computer, and it tells us how well is the quantum computer doing.

347
00:46:04.260 --> 00:46:07.739
And, in fact, the data that we get, um.

348
00:46:07.739 --> 00:46:14.880
Agrees remarkably well with the model so you can see we have these red circles, which are full circuits.

349
00:46:14.880 --> 00:46:19.530
And you can see they're really aligning very well with the model.

350
00:46:19.530 --> 00:46:30.929
Even though, that model is not including anything besides single interview errors and you can see, we also have these green axis, which are simplified circuits. And I mentioned this before.

351
00:46:30.929 --> 00:46:40.530
We run both the full circuit and some slightly simplified circuit. So we, we shift some of the 2 cubic gains around or we.

352
00:46:40.530 --> 00:46:47.400
We'll choose a different gate here or there to make the classical computers job easier. And the reason we have to do that is because.

353
00:46:47.400 --> 00:46:50.489
At some point, we otherwise won't be able to.

354
00:46:50.489 --> 00:46:58.710
Check to really run this part of the algorithm anymore. We can't check that. The quantum computer is performing correctly.

355
00:46:58.710 --> 00:47:10.559
If it's task is too hard for the classical computer you can see that's already starting to happen here where you've got red data, red, red, red, red, red, and then there's this block. That's because this 1 last dot at the end.

356
00:47:10.559 --> 00:47:15.210
Is there's 1 last got in the end? Actually.

357
00:47:15.210 --> 00:47:27.960
5 hours with 1M course to verify and once you start doing too much of this, you get calls from the Google data center. People asking why are you using so much of their compute resources? So.

358
00:47:27.960 --> 00:47:39.300
you know you have to be careful with how many of these you do and you can't sustain doing all of this data and that's what we have just this one point here but as you can see the full data the data for .

359
00:47:39.300 --> 00:47:47.789
The performance of a full circuit, and the data for the performance of a simplified circuit overlap very well and they both overlap very well with the model. So.

360
00:47:47.789 --> 00:47:53.699
This suggests to us that the corner processor is actually working and it's working. Well, in a way that we understand.

361
00:47:53.699 --> 00:48:02.250
That emboldened us to go venture into the supremacy regime and of course, here, we can't run the full circuits anymore.

362
00:48:02.250 --> 00:48:11.280
We have stored archived the bit strings that are quantum computer produced and written them down with the hopes that.

363
00:48:11.280 --> 00:48:24.659
When it, hopefully in a not too distant future classical computing will improve to the point that those circuits can be simulated, then it will be possible for anyone to check the performance of our quantum computer. That is.

364
00:48:24.659 --> 00:48:33.300
In today's supremacy regime, but the thing that we can do is still continue to check these simplified circuits and we see that.

365
00:48:33.300 --> 00:48:36.690
Even in this higher.

366
00:48:36.690 --> 00:48:39.809
This is supremacy regime, higher numbers of cycles.

367
00:48:39.809 --> 00:48:49.289
So deeper circuits, you can still see that the, the model is well, predicting the performance that we observe with the quantum computer on those simplified circuits.

368
00:48:49.289 --> 00:48:54.090
So, with that, we say that we have achieved quantum supremacy.

369
00:48:55.644 --> 00:49:03.744
Now, I've mentioned quite a bit about the classical machines that we use to verify the performance of the quantum computer.

370
00:49:03.744 --> 00:49:11.065
There's actually several different classical machines and also a couple different algorithms that we use, depending on.

371
00:49:11.159 --> 00:49:21.480
What kind of where we are in this premise here, whether it's a many many Cubans or many many cycles and so these are a few of the.

372
00:49:21.480 --> 00:49:25.829
These are the supercomputers that we've collaborated with.

373
00:49:26.574 --> 00:49:33.655
And here are a couple of the simulation strategies are the 2 simulation strategies that those classical computers can use.

374
00:49:34.045 --> 00:49:42.684
1 was called the shrink or simulation and this, in essence stores all the amplitude is it just tries to simulate the full circuit.

375
00:49:43.019 --> 00:49:47.789
And so it produces all of the probability attitudes that you would need.

376
00:49:47.789 --> 00:49:53.039
And for the biggest circuits that we've run, if you had a machine.

377
00:49:53.039 --> 00:49:57.300
With dozens of petabytes of memory, you could run this in a few days.

378
00:49:57.300 --> 00:50:08.400
If you had a machine with dozens of petabytes of memory, so you don't. And so what you might want to do instead is break this circuit up into pieces.

379
00:50:08.400 --> 00:50:11.730
Simulate those pieces and then stitched them back together.

380
00:50:11.730 --> 00:50:18.329
And that stitching process works just 1 amplitude at a time. So, in essence, your trading.

381
00:50:18.329 --> 00:50:21.570
Memory for time and.

382
00:50:21.570 --> 00:50:28.920
That trade is not particularly favorable if you are working on a machine where you have just terabytes of memory, then.

383
00:50:28.920 --> 00:50:32.849
Unfortunately, the runtime explodes and now you're, you're.

384
00:50:32.849 --> 00:50:45.030
Working in the thousands of years, receive regime so either way you split it. Literally the resources that you need in order to simulate these circuits using a classical machine are are enormous.

385
00:50:45.030 --> 00:50:48.119
So, if the take home message, then.

386
00:50:48.119 --> 00:50:55.590
Here is we have our processor that we've called sycamore. It can run 53 cube circuits.

387
00:50:55.590 --> 00:51:00.869
With non 0T, fidelity and okay, that sounds a little bit funny. I'm saying that the quantum computer.

388
00:51:00.869 --> 00:51:06.000
Perform it gives me the right answer at all ever. In fact.

389
00:51:06.000 --> 00:51:13.500
The this algorithm is very sensitive to errors in the quantum computer and you do see that this.

390
00:51:13.500 --> 00:51:17.039
This the numbers on this Y, axis are.

391
00:51:17.039 --> 00:51:30.659
Pretty small that said these are the numbers that were predicted from those gate errors that we measured at the beginning. So we can run these these 53 human circuits and we get a non 0T fidelity.

392
00:51:30.659 --> 00:51:34.230
And the fidelity is predicted well, by our simple model.

393
00:51:34.230 --> 00:51:40.380
Furthermore, if we were to try to reproduce that work, and we were to try to get the classical computer.

394
00:51:40.380 --> 00:51:50.579
To run that same algorithm and just to reach the same teeny, tiny fidelity that would require a ridiculous amount of resources. And so if that, we say we have achieved.

395
00:51:50.579 --> 00:51:57.599
On supremacy, we've achieved our small milestone computational task to demonstrate this prototype.

396
00:51:57.599 --> 00:52:02.820
Can do something interesting taking a step back what.

397
00:52:02.820 --> 00:52:12.750
What's really going on here from from a big picture perspective. We kind of like to look at this on 2 levels. 1 is a computer science level. This is.

398
00:52:12.750 --> 00:52:16.590
A piece of physical demonstration.

399
00:52:16.590 --> 00:52:25.380
That a quantum computer seems to be fundamentally different from a classical computer and in particulars is interesting because.

400
00:52:25.380 --> 00:52:29.099
The difference between 52 and 53.

401
00:52:29.099 --> 00:52:33.030
Is just 1 human honor chip, but it's.

402
00:52:33.030 --> 00:52:40.679
A, very significant change for the work of simulating that system on the classical computer.

403
00:52:40.679 --> 00:52:47.460
And from a physics perspective, it's it's quite interesting that we can even control.

404
00:52:47.460 --> 00:52:53.639
I heard space of 2 to the 53 and that that as a quantum system with 53.

405
00:52:53.639 --> 00:53:01.050
Constituent quantum parts still works quantum mechanics still works for such a highly complex system and.

406
00:53:01.050 --> 00:53:08.130
There was some doubt in the community as to whether this would be possible and therefore, whether quantum computing would be possible.

407
00:53:08.130 --> 00:53:14.400
So, from a physics perspective, this is a very interesting proof of principle that indeed.

408
00:53:14.400 --> 00:53:18.480
At least to the size of 53 elements.

409
00:53:18.480 --> 00:53:23.070
We can indeed a quantum mechanics still continues to work.

410
00:53:23.070 --> 00:53:26.400
And with that, I would like to think.

411
00:53:26.400 --> 00:53:33.510
The great team that I put this whole experiment together, and I would like to take whatever questions.

412
00:53:33.510 --> 00:53:42.659
All right, thank you very much Marisa. Fantastic book.

413
00:53:42.659 --> 00:53:53.909
And I'm going to take this opportunity as in all of the online talks at the moment to upload on everyone's perhaps I, thank you very much.

414
00:53:53.909 --> 00:54:04.530
Um, but, yeah, like to open up to questions, uh, now if people have questions, please unmute yourself and ask I see that Alexis has already unmuted himself. So, do you have a question.

415
00:54:04.530 --> 00:54:08.849
Yes, yes, I've seen that diagram of.

416
00:54:08.849 --> 00:54:12.329
The QB several times and I've just been wondering.

417
00:54:12.329 --> 00:54:23.820
What is the missing? What happened? So what happened there that the Cuban itself as far as we know is only fine.

418
00:54:23.820 --> 00:54:28.769
The problem if you will, was that every so often.

419
00:54:28.769 --> 00:54:32.099
Things do work on the 1st shot.

420
00:54:32.099 --> 00:54:37.110
And so the, the broken piece is actually the line of the package.

421
00:54:37.110 --> 00:54:40.769
And we had our very 1st device.

422
00:54:40.769 --> 00:54:53.070
And our, we thought, well, we'll mounted and we'll, we'll cool it down and something is bound to not work. We've got this 1 package that has 1 broken line. We could.

423
00:54:53.070 --> 00:55:03.000
We could use our other package, but we'd really rather have that package for the 2nd cool down when, when we're going to fix all the bugs from the 1st cool down. So we'll use this broken package and.

424
00:55:03.000 --> 00:55:08.550
Then we pull it down and actually everything that worked and we never got to go back and fix it. So.

425
00:55:08.550 --> 00:55:17.010
That was going to be my question. 1st question actually. All right next question.

426
00:55:23.070 --> 00:55:33.329
I have a question, so I said, I was wondering, what's the time scale difference between the measurement and the, to cubic gates are.

427
00:55:33.329 --> 00:55:37.440
Like, are they roughly the same order magnitude? Or are they.

428
00:55:37.440 --> 00:55:44.820
Different now the measurement as much slower. So the, the gates are on the order of.

429
00:55:44.820 --> 00:55:53.699
29 seconds, I think the single cubic gates are about 12 or 15 seconds. The gates are closer to 20 and then measurement.

430
00:55:53.699 --> 00:55:57.329
Takes on the order of microsecond here.

431
00:55:59.460 --> 00:56:03.000
Is that a like a fundamental thing or a like a.

432
00:56:03.000 --> 00:56:09.659
Like, how like, optimistically, like, how fast do you expect to be able to produce to magical time?

433
00:56:11.070 --> 00:56:15.000
That is the, that is a big question. Um.

434
00:56:15.000 --> 00:56:18.539
Of course, we would like the measurement time to be much lower.

435
00:56:18.539 --> 00:56:22.139
And it's a product of how much more.

436
00:56:22.139 --> 00:56:31.079
How many more characters are involved in the process of measurement as compared to Gates if you want to do a gate on a on.

437
00:56:31.079 --> 00:56:38.579
You send some I carry down the line and zap the Cupid and you have to zap it in just the right way and be.

438
00:56:38.579 --> 00:56:44.550
Kind and hurtful and sensitive to that cubic, but it's, it's really a well configured SAP.

439
00:56:44.550 --> 00:56:57.090
Readout on the other hand, requires the coordination of that readout resonator and some kind of filtering. And of course, what you're actually trying to do is detect the presence or absence of 1 foot on at 5 gigahertz, which is.

440
00:56:57.090 --> 00:57:05.610
Ridiculous a small amount of energy in vacuum. So there's a lot it's more complicated in terms of the number of.

441
00:57:05.610 --> 00:57:13.949
Of objects involved, and we are optimistic that we'll be able to reduce the measurement time. We're playing around with some.

442
00:57:13.949 --> 00:57:18.659
Different approaches to filtering and to amplifying.

443
00:57:18.659 --> 00:57:24.090
But we don't have any any great data on that published yet.

444
00:57:24.090 --> 00:57:32.849
Okay, yeah, thank you. Sorry I wish I had more just following that day.

445
00:57:32.849 --> 00:57:36.510
So, like, uh, obviously a microsecond.

446
00:57:36.510 --> 00:57:42.210
Is a fairly typical time for measuring typical cube it, but there are many.

447
00:57:42.210 --> 00:57:45.719
There were there police being a number of.

448
00:57:45.719 --> 00:57:55.230
Experiments where people have pushed those times down towards the 100 to 5 seconds, or even just below. Sorry? Hundreds of.

449
00:57:55.230 --> 00:58:00.750
90 seconds of like, what's the what's the difference between.

450
00:58:00.750 --> 00:58:04.949
Measurement in these devices and say the bar experiments.

451
00:58:04.949 --> 00:58:14.579
Full Circle, half it I guess it's partly to do with the fact that you're measuring 53 Cubics and you're trying to measure 9 Cubans at a time on a line. So that.

452
00:58:14.579 --> 00:58:19.769
Where are the limitations of the moment? In some ways it's the space.

453
00:58:19.769 --> 00:58:27.389
So, in 1 of the recent experiments, they put a filter on every single cube, for example.

454
00:58:27.389 --> 00:58:32.010
So each Cuba had 2 resonators associated with it.

455
00:58:32.010 --> 00:58:40.889
For example, and that requires a lot more space. There's resonators are big and we try we're trying to cram.

456
00:58:40.889 --> 00:58:47.849
All of the Cubans in and together, and 1 of the most tricky things about the layout is just figuring out.

457
00:58:47.849 --> 00:58:55.920
How to fit everything into the smallest amount of space and then keep it shielded enough that crosstalk doesn't dominate.

458
00:58:55.920 --> 00:58:59.429
The whole system, so.

459
00:58:59.429 --> 00:59:03.389
It's a very delicate balancing act of.

460
00:59:04.409 --> 00:59:12.000
Adding and so it's a scale of favoring together. It's it's a problem with scale. Yes.

461
00:59:12.000 --> 00:59:18.030
So so, do you still have a cell filter at all like the like you did with the 9? Keep it many a device.

462
00:59:18.030 --> 00:59:23.369
Yes, okay. So there are still the cell filters, but just not this many.

463
00:59:24.389 --> 00:59:28.380
Notice sorry.

464
00:59:28.380 --> 00:59:37.469
Not just personalized for each Cuban right? We are, we are so multiplexing we're still doing multiplex. Readout.

465
00:59:37.469 --> 00:59:43.320
Also, as I indicated before we've got 6, keep it on our outline.

466
00:59:43.320 --> 00:59:49.469
And so there is a frequency spacing issue to be considering there as well.

467
00:59:49.469 --> 00:59:54.780
Yeah, all right next question.

468
01:00:03.269 --> 01:00:10.019
Okay, so I might ask a question about your calibration a process, I guess.

469
01:00:10.019 --> 01:00:14.250
Um, I mean, this is calibration.

470
01:00:14.250 --> 01:00:17.670
Uh, graph that you developed.

471
01:00:17.670 --> 01:00:23.880
Is is, I guess.

472
01:00:23.880 --> 01:00:28.739
Very, it's obviously a very sophisticated process. It has to happen completely automated.

473
01:00:28.739 --> 01:00:31.920
What, I mean.

474
01:00:31.920 --> 01:00:36.750
Without giving away too many trade secrets.

475
01:00:36.750 --> 01:00:44.550
What are some of the key principles that you need to follow to try and create.

476
01:00:44.550 --> 01:00:52.199
Calibration routine like that. I would say probably 1 of the biggest ones is.

477
01:00:52.199 --> 01:00:57.239
Getting a group of people who can work together in a collaborative code environment.

478
01:00:57.239 --> 01:01:00.360
That may sound a bit trivial, but.

479
01:01:00.360 --> 01:01:03.989
This only works because we can build on.

480
01:01:03.989 --> 01:01:07.349
All of our colleagues and all of our former selves.

481
01:01:07.349 --> 01:01:13.050
We have 1 giant code repository where all of those calibration experiments live.

482
01:01:13.050 --> 01:01:20.280
And probably the biggest delta between.

483
01:01:20.280 --> 01:01:24.329
The calibration work that we were doing in 2016.

484
01:01:24.329 --> 01:01:28.289
And the work that we're doing today is just that the code got cleaner.

485
01:01:30.150 --> 01:01:33.150
So, it's been a lot of.

486
01:01:33.150 --> 01:01:36.150
Learning how to work.

487
01:01:36.150 --> 01:01:39.750
Together in, in some ways, learning how to.

488
01:01:39.750 --> 01:01:44.969
Right and experiment in such a way that anyone can pick it up and read it and continue it.

489
01:01:44.969 --> 01:01:51.570
And it sounds a bit trivial, but that's probably the biggest.

490
01:01:51.570 --> 01:01:54.599
Delta there was a really a clear.

491
01:01:54.599 --> 01:02:06.989
Take off in our productivity when we cleaned up, we had 1 marathon code, cleanup session and after that things really kind of took off in terms of how efficient and effective the calibration is. And.

492
01:02:06.989 --> 01:02:13.289
I mean, I shouldn't say that everything is clean and perfect and beautiful. We, we certainly still have times where somebody.

493
01:02:13.289 --> 01:02:22.559
Pushes some code to master and it breaks something and then everybody else is upset because why what happened there who reviewed this? But it's.

494
01:02:22.559 --> 01:02:27.329
It's so much cleaner than it has been in the past that.

495
01:02:27.329 --> 01:02:30.630
That really made a big difference. That's really interesting.

496
01:02:30.630 --> 01:02:38.460
We have a question in the chat. I'm not sure whether you're able to mute yourself and ask the.

497
01:02:38.460 --> 01:02:47.670
Question otherwise I'll ask for you. Okay, so our question is, what are the next steps and are there plans for building? Large chips?

498
01:02:49.050 --> 01:02:57.809
Yes, so the next steps are to or a, a next step is to start trying to do.

499
01:02:57.809 --> 01:03:03.690
To demonstrate the principles of error correction in practice. So.

500
01:03:03.690 --> 01:03:11.849
We, as I mentioned before 1000 Cubans is currently what we expect to need a 1000 physical.

501
01:03:11.849 --> 01:03:18.329
In order to reach a single logical cubic however, we should be able to demonstrate that.

502
01:03:18.329 --> 01:03:22.650
A slightly bigger array can sustain.

503
01:03:22.650 --> 01:03:29.340
It can produce a longer lived logical cubic than a smaller grid. And so.

504
01:03:29.340 --> 01:03:36.329
Our goal is to make a error correction prototype, if you will, or where we can show that.

505
01:03:36.329 --> 01:03:40.530
Yes, we can, in fact, add more and.

506
01:03:40.530 --> 01:03:45.929
Finagle a way that that will reduce the error for the whole logical cube. It.

507
01:03:45.929 --> 01:03:50.489
In order to do that, actually the, the.

508
01:03:50.489 --> 01:03:53.909
Chips that we are building are and about a skill that we could.

509
01:03:53.909 --> 01:04:00.360
Already tried to start doing that. We are planning to build some larger chips in the process of doing that as well.

510
01:04:00.360 --> 01:04:08.400
And so, yes, we're interested in building larger chips, but moreover, we're interested in improving our.

511
01:04:08.400 --> 01:04:15.869
Error rates and our readouts and our read out speed even further. And so those things have to be taken.

512
01:04:15.869 --> 01:04:18.929
You know, balanced fashion, it wouldn't make sense to.

513
01:04:18.929 --> 01:04:22.469
Focus only on building the biggest possible chip.

514
01:04:22.469 --> 01:04:25.650
And at the expense of.

515
01:04:25.650 --> 01:04:30.630
Building good quality Cubics and properly optimizing the technology.

516
01:04:30.630 --> 01:04:38.699
At the same time, there's something to be learned from pushing the cubic number. There's a lot of technology that has to be developed.

517
01:04:38.699 --> 01:04:42.239
Just to support a larger cubic number and that was.

518
01:04:42.239 --> 01:04:49.409
A big piece of the work that took us, it brought us to this prototype in the 1st place. We had to rethink the way we did Fab.

519
01:04:49.409 --> 01:04:56.909
To get from the linear chain of Cubans to that to D array. We had to build new packages. We had to automate our calibration.

520
01:04:56.909 --> 01:05:05.280
And there there are all of these things, all of these technological developments that needed to happen, regardless of whether or not, we would be successful with the quantum supremacy experiment.

521
01:05:05.280 --> 01:05:08.369
So that will happen moving forward as well.

522
01:05:08.369 --> 01:05:13.619
And then the, the need is just to balance.

523
01:05:13.619 --> 01:05:22.920
How much time do we spend building up the next technology versus making sure that the Cuban performance is really there to support a larger chip?

524
01:05:24.780 --> 01:05:28.530
I guess that opens up, um.

525
01:05:28.530 --> 01:05:35.219
So there clearly 2 different directions that you're working on 1 is to expand the size of the chip and 1 is to improve the performance.

526
01:05:35.219 --> 01:05:39.630
With regard to improving the performance.

527
01:05:39.630 --> 01:05:44.039
What do you think the.

528
01:05:44.039 --> 01:05:49.230
This sort of it's going to have the biggest impact on that. At the moment. Are we talking.

529
01:05:49.230 --> 01:05:54.030
At lots more fat development, or are we talking a lot more control?

530
01:05:54.030 --> 01:05:58.409
Strategies being improved, or what.

531
01:05:58.409 --> 01:06:01.500
What's your sense as to where you're going to get the best high off?

532
01:06:01.500 --> 01:06:14.099
Probably both and actually our team is broken right now. We've kind of exercises a few different ways of organizing ourselves, but but right now we have.

533
01:06:14.099 --> 01:06:22.019
Kind of the people who work full speed ahead on calibration. We have the people who who build the devices and we have our team.

534
01:06:22.019 --> 01:06:25.949
And we also have in in within our.

535
01:06:25.949 --> 01:06:34.769
Theory group a physics team and so there's been a sort of a really nice handshake where the physics team will see.

536
01:06:34.769 --> 01:06:38.789
Some effects that they think would make the calibration of it better.

537
01:06:38.789 --> 01:06:45.119
And then the, the, the calibration team does a big step of improving the calibration.

538
01:06:45.119 --> 01:06:51.719
Say they get it to 98% and then the team gets a hold of it and they, they twiddle around and they get it to.

539
01:06:51.719 --> 01:07:05.159
99%, you know, that last 2 extra percent is really hard to eke out and then the physics team finds oh, here's 1, little little physics aspect that you really need to incorporate very conscientiously into.

540
01:07:05.159 --> 01:07:13.920
Your calibration and then they get it to 99.8% or something like that. So, there, there's this interplay between what levels of.

541
01:07:13.920 --> 01:07:17.969
Problems you're trying to solve, I guess, and.

542
01:07:17.969 --> 01:07:22.260
At the moment, at least, it seems like we've managed a group ourselves into such a way that.

543
01:07:22.260 --> 01:07:25.559
We can work together on.

544
01:07:25.559 --> 01:07:28.619
Tackling each of these different kinds of problems.

545
01:07:28.619 --> 01:07:33.210
Somewhat effectively, hopefully that will persist into the future and we.

546
01:07:33.210 --> 01:07:37.889
We certainly will be working to improve the to have.

547
01:07:37.889 --> 01:07:43.320
But also the design, I mean, just making a better chip in includes.

548
01:07:43.320 --> 01:07:49.739
Designing a better chip, but also having the materials at your disposal, being being able to.

549
01:07:49.739 --> 01:07:54.900
Actually, that Fab process to really create that chip.

550
01:07:54.900 --> 01:07:59.039
So there.

551
01:07:59.039 --> 01:08:04.440
There's work to be done across the board. All right.

552
01:08:04.440 --> 01:08:10.650
Uh, it's not picking a host yet to this. It seems to answer.

553
01:08:10.650 --> 01:08:21.359
So, I'm all right, it looks like we're reaching a natural, um, right point in the questions coming in. I can't see anyone else.

554
01:08:21.359 --> 01:08:27.239
On meeting, so unless there's a a desperate call, I'd like to.

555
01:08:27.239 --> 01:08:32.640
Get everyone to thank Mercer again. It was a really fantastic tool.

556
01:08:32.640 --> 01:08:42.060
Uh, it's been a pleasure having you join our online seminar series and look forward to hearing more from your, um.

557
01:08:42.060 --> 01:08:45.630
Uh, from the team and all the exciting work very soon.

558
01:08:45.630 --> 01:08:49.199
Thank you very much. Thank you very much for the invitation.

559
01:08:49.199 --> 01:08:52.470
I wish I could be with you in person, but it's a pleasure to be.

560
01:08:52.470 --> 01:08:56.220
Here all right.

561
01:08:56.220 --> 01:09:01.500
Hello.

562
01:09:01.500 --> 01:09:06.060
Oh, okay. I hope you enjoyed that learned a lot about Google.

563
01:09:06.060 --> 01:09:11.909
And so no time, I don't want to start the other long 1.

564
01:09:12.930 --> 01:09:17.760
Right now anything short I can show you.

565
01:09:19.470 --> 01:09:22.890
I'm sure if I showed you this thing about it.

566
01:09:22.890 --> 01:09:26.550
I on cue or not. Let's see.

567
01:09:27.869 --> 01:09:39.359
Startup I on Q and Q, so I'm guessing they want me to pronounce that as I think I showed you this next gen, quantum computing system.

568
01:09:39.359 --> 01:09:47.069
Which it is calling the world's most powerful claiming to have set a new record by bursting through the 4M quantum volume barrier.

569
01:09:47.069 --> 01:09:54.390
Cloning forms on a machine featuring 32 Cubans, the quantum equivalent of classical computing bits.

570
01:09:54.390 --> 01:10:07.829
Ironic says it has achieved an expected quantum volume greater than 4M figure vaults ahead of the previous record. A quantum volume of 128 announced 1 day prior by Honeywell, the industrial conglomerate.

571
01:10:07.829 --> 01:10:20.784
Peter Chapman, chief executive said that, as the company releases newer of its machines in the years ahead updated measures will be required. The number will come civil large. We'll have to leave content volume behind.

572
01:10:20.784 --> 01:10:28.944
He said quantum volume attempts to grade content computers on a combination of metrics, including a machines number of Cubics their activity and error rates.

573
01:10:29.274 --> 01:10:37.194
Ibm a rival quantum computing pioneer, introduce the artistic 3 years ago in an effort to create a more holistic ranking system for quantum computing engineers.

574
01:10:37.470 --> 01:10:51.505
Cubans tend to be unstable, but in an ideal world, each additional 1 adds exponential power to a quantum machine, because of unique hardware design. It says it is able to tap into those exponential increases, helping push its machine far ahead of the pack.

575
01:10:51.805 --> 01:11:02.814
At least according to quantum volume is tilting against tech giants many times. The company size such as IBM. Google Honeywell, Intel and Microsoft the 5 year old startup based in college park.

576
01:11:02.814 --> 01:11:11.305
Maryland is racing, like the others to give businesses a computing edge and domain such as chemistry, financial modeling, medicine and artificial intelligence and.

577
01:11:12.060 --> 01:11:16.710
Oh, okay.

578
01:11:18.930 --> 01:11:26.970
So that's enough for today on just to remind you again homework do after Thanksgiving.

579
01:11:26.970 --> 01:11:33.630
This Thursday at photon, we'll be talking if there's.

580
01:11:33.630 --> 01:11:37.439
Time after he's finished, I will start talking about.

581
01:11:37.439 --> 01:11:43.289
Presenting some of this stuff on optical quantum computing and so on.

582
01:11:43.289 --> 01:11:49.619
Oh, okay. Have a good week. See you Thursday and as usual.

583
01:11:49.619 --> 01:11:53.159
I'll stay here for a while and answer questions.

584
01:11:53.159 --> 01:12:01.470
You're welcome.

585
01:12:05.039 --> 01:12:20.039
Silence.

586
01:12:20.039 --> 01:12:25.890
Oh, John, I haven't read it yet. So, let me check it after the.

587
01:12:25.890 --> 01:12:31.470
After class, or could you tell me what, what's the problem? Do you want to on mute and tell me about it or.

588
01:12:33.210 --> 01:12:47.725
Can you hear me? Yes, I do a little low, but I can hear you. All right yeah. Yeah. No, I submitted my my task to run this morning and I've just been waiting in the queue all day so I was just going to ask.

589
01:12:47.755 --> 01:12:50.784
Can I just submit whenever the queue finishes? Absolutely.

590
01:12:51.060 --> 01:12:57.539
Okay, so thank you if I haven't permitted in grade scope, tell me and I'll permit it. So.

591
01:12:58.314 --> 01:13:09.805
Yeah, I mean, I think I can still submit anytime after the submission deadline. So I can't actually so, tell me what was the, um, it's getting overloaded or something.

592
01:13:10.404 --> 01:13:13.465
I'm not really sure it just shows, um.

593
01:13:13.770 --> 01:13:23.970
There's sort of like a, um, you could see all the available devices and it just shows, uh, devices online and then you can see.

594
01:13:23.970 --> 01:13:32.880
Sort of like your task queue and it just shows tasks queued. So I'm not really sure when the, when it, it'll execute.

595
01:13:32.880 --> 01:13:44.489
But, hopefully, it shouldn't take too long. Are you advancing in the queue? The machine is actually running. It doesn't even show. It doesn't even give you like a like, what number you are in the queue. Oh.

596
01:13:44.489 --> 01:13:47.880
Yeah, you just have like, a task ID and just a status.

597
01:13:47.880 --> 01:13:55.590
Oh, okay. Well hope the machine's not down. Well, it says it says it's online.

598
01:13:55.590 --> 01:14:07.079
But I checked it before class and it said, like like, next it's an online, but then below it, it said next available in 15 hours or something like that.

599
01:14:07.079 --> 01:14:20.880
So, I guess we'll see quantum computing is popular. Yeah. Well, actually, they don't raise their prices. No I mean, at least for now, it's pretty cheap. It's like 30 cents per run or something like that. Yeah. Okay.

600
01:14:20.880 --> 01:14:28.409
Thank you anyone else have any comments questions feedback.

601
01:14:30.060 --> 01:14:33.930
No, okay. See you Thursday then.

602
01:14:44.189 --> 01:14:47.670
Silence.