WEBVTT

1
00:00:01.480 --> 00:00:05.599
<v Speaker 1>Welcome to the quarre Side. Quantum Physics podcast, an exploration

2
00:00:05.679 --> 00:00:10.839
<v Speaker 1>of the fundamental structure of reality where quantum laws govern matter, energy,

3
00:00:11.199 --> 00:00:16.239
<v Speaker 1>and information. Here, uncertainty is a feature, not a flaw,

4
00:00:16.719 --> 00:00:19.239
<v Speaker 1>and understanding begins of the smallest scales.

5
00:00:28.320 --> 00:00:32.920
<v Speaker 2>Imagine you're holding like the single most powerful piece of technology.

6
00:00:32.359 --> 00:00:35.320
<v Speaker 3>In human history, always a fun thought experiment, right, Like.

7
00:00:35.679 --> 00:00:39.880
<v Speaker 2>A machine that's capable of simulating the molecular structure of

8
00:00:40.119 --> 00:00:46.719
<v Speaker 2>complex life saving drugs, or cracking cryptographic codes that would

9
00:00:46.759 --> 00:00:49.719
<v Speaker 2>take classical computer's millennia to break.

10
00:00:49.799 --> 00:00:52.799
<v Speaker 3>Basically modeling the very fabric of the universe exactly.

11
00:00:52.880 --> 00:00:54.880
<v Speaker 2>So you power it up, you input the variables, you

12
00:00:54.880 --> 00:00:57.079
<v Speaker 2>get ready to change the world, and then you discover

13
00:00:57.159 --> 00:01:00.679
<v Speaker 2>your multimillion dollar machine has the memory of a old fish.

14
00:01:00.759 --> 00:01:03.520
<v Speaker 3>Yeah, that is a That's the great paradox of modern

15
00:01:03.520 --> 00:01:08.079
<v Speaker 3>physics right there. I mean, we spend billions developing machines

16
00:01:08.120 --> 00:01:13.000
<v Speaker 3>with almost unimaginable computational promise, but their current hardware instability

17
00:01:13.079 --> 00:01:15.280
<v Speaker 3>is just it's the absolute bottleneck.

18
00:01:14.799 --> 00:01:16.560
<v Speaker 2>Because they literally forget what they're doing.

19
00:01:16.719 --> 00:01:19.879
<v Speaker 3>Right If the processor literally forgets the math problem halfway

20
00:01:19.879 --> 00:01:22.959
<v Speaker 3>through solving it, the theoretical power just doesn't matter.

21
00:01:23.519 --> 00:01:26.000
<v Speaker 2>So today we're going to explore this Achilles heel of

22
00:01:26.079 --> 00:01:30.959
<v Speaker 2>quantum computers. We're getting into the unpredictable fluctuation of quantum memory, and.

23
00:01:30.959 --> 00:01:37.879
<v Speaker 3>More importantly, unpacking a massive new breakthrough that has accelerated

24
00:01:38.239 --> 00:01:41.879
<v Speaker 3>our ability to measure this information loss by more than

25
00:01:41.920 --> 00:01:43.040
<v Speaker 3>one hundred times.

26
00:01:43.040 --> 00:01:45.680
<v Speaker 2>Which is just a staggering leap. I mean, the conversation

27
00:01:45.719 --> 00:01:49.000
<v Speaker 2>around quantum computing usually gets bogged down in theory.

28
00:01:48.760 --> 00:01:52.079
<v Speaker 3>Right, oh, constantly right, But the reality of engineering these

29
00:01:52.120 --> 00:01:57.959
<v Speaker 3>devices requires confronting some incredibly harsh microscopic physics. To build

30
00:01:58.000 --> 00:02:01.879
<v Speaker 3>a reliable machine, you have to measure its unreliability first.

31
00:02:02.200 --> 00:02:04.599
<v Speaker 3>Makes sense, and for a long time our measurement tools

32
00:02:04.599 --> 00:02:06.640
<v Speaker 3>were just well, they were simply too slow to see

33
00:02:06.640 --> 00:02:07.519
<v Speaker 3>the actual problems.

34
00:02:07.560 --> 00:02:10.280
<v Speaker 2>So let's get right into the hardware of that unreliability.

35
00:02:10.639 --> 00:02:12.840
<v Speaker 2>For the audience listening, we all know what equib it is,

36
00:02:12.879 --> 00:02:15.520
<v Speaker 2>and we know they operate in delicate stakes of superposition.

37
00:02:15.599 --> 00:02:17.319
<v Speaker 2>But let's look at the physical architecture.

38
00:02:17.400 --> 00:02:20.240
<v Speaker 3>Yeah, specifically, we're talking about superconducting quibits.

39
00:02:19.879 --> 00:02:23.080
<v Speaker 2>Here, right. These are microscopic electrical circuits, usually made of

40
00:02:23.080 --> 00:02:27.120
<v Speaker 2>metals like aluminium or niobium, printed on a silicon or

41
00:02:27.159 --> 00:02:30.439
<v Speaker 2>sapphire chip, and they are cooled down to fractions of

42
00:02:30.479 --> 00:02:33.080
<v Speaker 2>a degree above absolute zero, Yes.

43
00:02:33.000 --> 00:02:36.159
<v Speaker 3>Around fifteen millik Elvin's to be exact, which is colder

44
00:02:36.199 --> 00:02:39.280
<v Speaker 3>than the vacuum of deep space. That's crazy, it is, Yeah,

45
00:02:39.280 --> 00:02:41.439
<v Speaker 3>And we do that so the electrons can pair up

46
00:02:41.439 --> 00:02:44.520
<v Speaker 3>into what are called Cooper pairs. They flow without any

47
00:02:44.560 --> 00:02:46.120
<v Speaker 3>electrical resistance.

48
00:02:45.759 --> 00:02:49.199
<v Speaker 2>And that frictionless flow is what allows the quantum state

49
00:02:49.280 --> 00:02:52.639
<v Speaker 2>to survive exactly. But clearly it doesn't survive for very long.

50
00:02:52.680 --> 00:02:58.479
<v Speaker 2>There's this constant thread of relaxation or decoherence.

51
00:02:57.840 --> 00:02:59.280
<v Speaker 3>Right, decoherence to the enemy.

52
00:02:59.439 --> 00:03:02.960
<v Speaker 2>Let me flow an analogy here. If maintaining a classical

53
00:03:03.000 --> 00:03:06.840
<v Speaker 2>bit is like laying a heavy coin flat on the table, Okay,

54
00:03:06.919 --> 00:03:09.159
<v Speaker 2>like it takes a lot of force to flip it over,

55
00:03:09.639 --> 00:03:12.800
<v Speaker 2>But a quibit in superposition is more like a spinning coin.

56
00:03:13.280 --> 00:03:13.759
<v Speaker 3>I like that.

57
00:03:14.039 --> 00:03:18.120
<v Speaker 2>Yeah, it holds both heads and tails dynamically, but eventually

58
00:03:18.360 --> 00:03:21.319
<v Speaker 2>microscopic friction from the air or the table, you know,

59
00:03:21.479 --> 00:03:25.039
<v Speaker 2>it SAPs its momentum and it clatters flat. It completely

60
00:03:25.080 --> 00:03:26.560
<v Speaker 2>loses its quantum state.

61
00:03:26.719 --> 00:03:31.840
<v Speaker 3>That captures the physics beautifully actually, and that friction causing

62
00:03:31.840 --> 00:03:34.439
<v Speaker 3>the coin to clatter flat is what we call environmental noise.

63
00:03:34.719 --> 00:03:36.919
<v Speaker 2>Even in a freezer that's colder than space.

64
00:03:37.039 --> 00:03:40.759
<v Speaker 3>Even there inside a dilution refrigerator at fifteen milik Ellen's

65
00:03:41.080 --> 00:03:44.360
<v Speaker 3>the environment is not perfectly silent. You've got stray photons

66
00:03:44.400 --> 00:03:47.360
<v Speaker 3>wandering through the chamber, you have cosmic rays striking the

67
00:03:47.439 --> 00:03:52.039
<v Speaker 3>outer casing. But mostly you have microscopic defects within the

68
00:03:52.080 --> 00:03:56.680
<v Speaker 3>silicon or sapphire substrate itself. Wait really, yeah, tiny atomic

69
00:03:56.719 --> 00:03:59.280
<v Speaker 3>imperfections called two level system defects.

70
00:04:00.199 --> 00:04:02.319
<v Speaker 2>So the noise isn't just coming from the outside. The

71
00:04:02.360 --> 00:04:05.280
<v Speaker 2>ship itself is inherently flawed at the atomic level.

72
00:04:05.520 --> 00:04:07.680
<v Speaker 3>That is the insidious part of the problem. As the

73
00:04:07.719 --> 00:04:14.000
<v Speaker 3>temperature shift slightly or electric fields fluctuate, these atomic defects

74
00:04:14.080 --> 00:04:17.839
<v Speaker 3>within the chip actually absorb energy from the quibit. Oh wow,

75
00:04:17.879 --> 00:04:20.519
<v Speaker 3>they act like microscopic sponges. They just suck the quantum

76
00:04:20.560 --> 00:04:23.600
<v Speaker 3>information right out of the circuit. That's brutal, it really is.

77
00:04:23.959 --> 00:04:26.240
<v Speaker 3>When that happens, the quibit drops out of its excited

78
00:04:26.279 --> 00:04:28.639
<v Speaker 3>state back to its ground state. The information is gone,

79
00:04:29.480 --> 00:04:31.920
<v Speaker 3>and in physics we measure the life stan of that

80
00:04:32.000 --> 00:04:35.279
<v Speaker 3>memory using a metric called T one. The relaxation time,

81
00:04:35.519 --> 00:04:36.680
<v Speaker 3>right the relaxation time.

82
00:04:36.879 --> 00:04:40.199
<v Speaker 2>And according to Jiro and Danen, who's a professor at

83
00:04:40.240 --> 00:04:44.120
<v Speaker 2>NTNU involved in this research, the problem isn't just that

84
00:04:44.160 --> 00:04:49.759
<v Speaker 2>T one is short. It's that T one is wildly unpredictable.

85
00:04:49.120 --> 00:04:52.199
<v Speaker 3>And that is the core engineering nightmare right there. I mean,

86
00:04:52.399 --> 00:04:56.800
<v Speaker 3>if the relaxation time was uniformly short, say every quibit

87
00:04:57.120 --> 00:05:00.879
<v Speaker 3>perfectly held its memory for exactly fifty micros, you could

88
00:05:00.920 --> 00:05:04.279
<v Speaker 3>just work around it exactly. You could meticulously schedule your

89
00:05:04.279 --> 00:05:08.279
<v Speaker 3>calculations to execute and finish in forty nine microseconds.

90
00:05:08.319 --> 00:05:10.839
<v Speaker 2>But it's not a static timer, not at all. Like

91
00:05:10.879 --> 00:05:13.720
<v Speaker 2>the life span of the memory fluctuates randomly over time.

92
00:05:13.839 --> 00:05:17.319
<v Speaker 2>One minute, the quibbit might hold information for one hundred microseconds,

93
00:05:17.319 --> 00:05:20.839
<v Speaker 2>and the next minute it forgets everything after like ten microseconds.

94
00:05:20.920 --> 00:05:24.759
<v Speaker 3>Because those microscopic atomic defects in the chip, they're constantly shifting,

95
00:05:24.800 --> 00:05:27.439
<v Speaker 3>they're coupling and decoupling with equibit. It's just a highly

96
00:05:27.519 --> 00:05:28.439
<v Speaker 3>dynamic environment.

97
00:05:28.560 --> 00:05:30.040
<v Speaker 2>It sounds chaotic, it is.

98
00:05:30.040 --> 00:05:32.839
<v Speaker 3>You're basically trying to build a perplectly synchronized orchestra out

99
00:05:32.839 --> 00:05:35.959
<v Speaker 3>of instruments that randomly change their tuning every few seconds.

100
00:05:36.120 --> 00:05:37.959
<v Speaker 2>Okay, So if I'm an engineer trying to fix this,

101
00:05:38.079 --> 00:05:40.680
<v Speaker 2>my very first instinct is, well, I need to observe

102
00:05:40.720 --> 00:05:43.000
<v Speaker 2>the machine while it's running. I need to find out

103
00:05:43.040 --> 00:05:46.399
<v Speaker 2>exactly when and why the tuning changes and fix that

104
00:05:46.519 --> 00:05:51.040
<v Speaker 2>specific defect naturally. But this brings us to the massive

105
00:05:51.120 --> 00:05:54.439
<v Speaker 2>roadblock that this new research published in Physical Review X

106
00:05:54.720 --> 00:05:59.360
<v Speaker 2>finally solved, because historically scientists were essentially flying blind.

107
00:05:59.759 --> 00:06:02.959
<v Speaker 3>They were because you know, measuring a quantum system is

108
00:06:03.000 --> 00:06:06.759
<v Speaker 3>fundamentally different from measuring a classical system. It used to

109
00:06:06.800 --> 00:06:10.160
<v Speaker 3>take researchers roughly a full second to measure that T

110
00:06:10.319 --> 00:06:11.560
<v Speaker 3>one relaxation time.

111
00:06:11.680 --> 00:06:13.519
<v Speaker 2>Now, let me push back on that a bit, because

112
00:06:13.519 --> 00:06:15.279
<v Speaker 2>I think a lot of people might hear one second

113
00:06:15.319 --> 00:06:18.560
<v Speaker 2>and think, well, a second is incredibly fast. Like if

114
00:06:18.600 --> 00:06:21.800
<v Speaker 2>my laptop diagnosed a hardware error in one second, I'd

115
00:06:21.839 --> 00:06:25.000
<v Speaker 2>be thrilled. Why is a one second measurement lag such

116
00:06:25.000 --> 00:06:28.720
<v Speaker 2>a catastrophic bottleneck When we're measuring quantum computers, you really have.

117
00:06:28.720 --> 00:06:31.439
<v Speaker 3>To scale your perception of time down to the microscopic

118
00:06:31.519 --> 00:06:35.279
<v Speaker 3>round here. A single second in a quantum processor isn't

119
00:06:35.279 --> 00:06:38.959
<v Speaker 3>a quick diagnostic check. It is an absolute eternity, and

120
00:06:39.040 --> 00:06:43.120
<v Speaker 3>eternity Yeah, the environment around that superconducting circuit is shifting,

121
00:06:43.399 --> 00:06:47.639
<v Speaker 3>it's vibrating, and it's changing state millions of times per second.

122
00:06:47.879 --> 00:06:51.079
<v Speaker 2>So a one second measurement means you are just entirely

123
00:06:51.120 --> 00:06:53.279
<v Speaker 2>missing millions of microscopic events.

124
00:06:53.399 --> 00:06:56.480
<v Speaker 3>Yes, you're just completely missing them. But to understand why

125
00:06:56.519 --> 00:06:58.720
<v Speaker 3>it took a full second, we have to look at

126
00:06:58.759 --> 00:07:02.480
<v Speaker 3>the mechanics of quantum op Okay, you can't just point

127
00:07:02.839 --> 00:07:06.120
<v Speaker 3>a continuous sensor at equibit and watch its energy fade

128
00:07:06.199 --> 00:07:10.160
<v Speaker 3>over time. It's not like watching a battery icon drain

129
00:07:10.240 --> 00:07:11.279
<v Speaker 3>on your phone.

130
00:07:11.000 --> 00:07:13.639
<v Speaker 2>Right because of the observer effect. Exactly, if you look

131
00:07:13.680 --> 00:07:16.759
<v Speaker 2>at the spinning coin, you force it to clatter flat immediately.

132
00:07:17.199 --> 00:07:19.319
<v Speaker 2>You force it to choose a definitive state of zero

133
00:07:19.519 --> 00:07:23.199
<v Speaker 2>or one, which destroys that delicate superposition you were trying

134
00:07:23.240 --> 00:07:24.319
<v Speaker 2>to measure in the first place.

135
00:07:24.439 --> 00:07:28.680
<v Speaker 3>Spot On the act of measuring the quibbot destroys the memory.

136
00:07:28.800 --> 00:07:31.120
<v Speaker 3>So how do you find out how long the memory

137
00:07:31.240 --> 00:07:33.920
<v Speaker 3>lasts if looking at it actually kills it.

138
00:07:34.000 --> 00:07:35.040
<v Speaker 2>That sounds impossible.

139
00:07:35.079 --> 00:07:37.279
<v Speaker 3>It's tricky. You have to use a statistical approach called

140
00:07:37.279 --> 00:07:38.279
<v Speaker 3>inversion recovery.

141
00:07:38.480 --> 00:07:42.160
<v Speaker 2>Okay, walk me through that. How does inversion recovery actually work.

142
00:07:42.480 --> 00:07:45.879
<v Speaker 3>So you send a microwave pulse to excite the quibbit.

143
00:07:46.519 --> 00:07:47.560
<v Speaker 3>Basically you spin the.

144
00:07:47.480 --> 00:07:48.839
<v Speaker 2>Coin okay, coin of spinning.

145
00:07:49.079 --> 00:07:52.800
<v Speaker 3>Then you purposefully wait for a specific duration, let's say

146
00:07:53.120 --> 00:07:57.959
<v Speaker 3>ten microseconds. Right after exactly ten microseconds, you hit it

147
00:07:57.959 --> 00:08:00.480
<v Speaker 3>with a measurement pulse to see if it is still

148
00:08:00.519 --> 00:08:02.360
<v Speaker 3>spinning or if it fell flat.

149
00:08:02.680 --> 00:08:05.279
<v Speaker 2>But that only tells you the result for that single trial.

150
00:08:05.319 --> 00:08:07.079
<v Speaker 2>It doesn't give you the average lifespan.

151
00:08:07.240 --> 00:08:09.079
<v Speaker 3>Correct. So you have to reset the quibit and do

152
00:08:09.160 --> 00:08:12.839
<v Speaker 3>that exact same ten microsecond waight time thousands of times

153
00:08:13.279 --> 00:08:16.639
<v Speaker 3>just to build a reliable probability thousands of times thousands

154
00:08:16.759 --> 00:08:19.399
<v Speaker 3>And then you have to increase the weight time to

155
00:08:19.399 --> 00:08:23.560
<v Speaker 3>say fifteen microseconds, and run another thousand trials. Oh my god,

156
00:08:23.759 --> 00:08:28.560
<v Speaker 3>then twenty microseconds, another thousand trials. You painstakingly sweep through

157
00:08:28.639 --> 00:08:31.639
<v Speaker 3>all these different weight times, plotting the probabilities until you

158
00:08:31.680 --> 00:08:35.480
<v Speaker 3>build this beautiful, smooth, exponential decay curve, and.

159
00:08:35.440 --> 00:08:37.440
<v Speaker 2>The slope of that curve tells you the t one

160
00:08:37.559 --> 00:08:40.320
<v Speaker 2>relaxation time. Wow, Okay, I see the problem.

161
00:08:40.360 --> 00:08:40.600
<v Speaker 3>Now.

162
00:08:40.960 --> 00:08:43.840
<v Speaker 2>It's not that a single measurement takes a second. It's

163
00:08:43.840 --> 00:08:47.159
<v Speaker 2>that you have to run tens of thousands of individual

164
00:08:47.200 --> 00:08:50.799
<v Speaker 2>excite weight measure reset cycles just to plot a single curve.

165
00:08:51.399 --> 00:08:54.519
<v Speaker 2>And that entire ensemble process is what takes a full second.

166
00:08:54.559 --> 00:08:57.559
<v Speaker 3>And here is where the physics really bite you. During

167
00:08:57.600 --> 00:09:00.679
<v Speaker 3>that one second you spent gathering all those thousands of

168
00:09:00.759 --> 00:09:04.639
<v Speaker 3>data points, the atomic defects in the chip were actively shifting.

169
00:09:04.879 --> 00:09:07.840
<v Speaker 2>The environment was fluctuating, right, which means the curve you

170
00:09:07.919 --> 00:09:09.440
<v Speaker 2>just claweded is basically a.

171
00:09:09.360 --> 00:09:11.759
<v Speaker 3>Lie, or at least a massive blur.

172
00:09:12.080 --> 00:09:12.399
<v Speaker 1>Yeah.

173
00:09:12.480 --> 00:09:14.279
<v Speaker 2>It's like trying to take a photograph of a chaotic

174
00:09:14.360 --> 00:09:16.679
<v Speaker 2>mosh pit, but you leave the camera shutter open for

175
00:09:16.720 --> 00:09:17.440
<v Speaker 2>a full minute.

176
00:09:17.519 --> 00:09:18.639
<v Speaker 3>Oh, that's a great way to put it.

177
00:09:18.720 --> 00:09:21.240
<v Speaker 2>You wouldn't see any of the individual collisions or the

178
00:09:21.320 --> 00:09:25.320
<v Speaker 2>dynamic movement. You just get the smeared, useless blob of

179
00:09:25.399 --> 00:09:26.200
<v Speaker 2>average motion.

180
00:09:26.480 --> 00:09:31.399
<v Speaker 3>That is a phenomenal conceptualization. Over that one second integration time,

181
00:09:31.759 --> 00:09:35.360
<v Speaker 3>the quibot's actual relaxation rate might have spiked and plummeted

182
00:09:35.440 --> 00:09:39.080
<v Speaker 3>one hundred times. Wow, But your one second ensemble measurement

183
00:09:39.440 --> 00:09:43.360
<v Speaker 3>just spits out a single smeared average. You know the

184
00:09:43.399 --> 00:09:47.159
<v Speaker 3>memory fluctuates, but you have no idea what specific microscopic

185
00:09:47.200 --> 00:09:51.159
<v Speaker 3>event triggered the loss in that exact fraction of a millisecond, And.

186
00:09:51.120 --> 00:09:53.320
<v Speaker 2>If you can't see the details of the noise, you

187
00:09:53.679 --> 00:09:56.600
<v Speaker 2>obviously can't engineer a defense against it. You just can't,

188
00:09:56.840 --> 00:10:00.559
<v Speaker 2>which brings us to the breakthrough the t at the

189
00:10:00.600 --> 00:10:05.159
<v Speaker 2>niels Bore Institute in Copenhagen, led by Fabrizio Barita, alongside

190
00:10:05.240 --> 00:10:09.279
<v Speaker 2>Jero and Dan and the NTNU researchers, they basically decided

191
00:10:09.320 --> 00:10:12.159
<v Speaker 2>that staring at a smeared mosh pit wasn't going to

192
00:10:12.200 --> 00:10:13.440
<v Speaker 2>scale quantum computing.

193
00:10:13.559 --> 00:10:14.600
<v Speaker 3>No, it certainly wasn't.

194
00:10:14.720 --> 00:10:17.320
<v Speaker 2>They needed a high speed camera and they achieved something

195
00:10:17.360 --> 00:10:20.960
<v Speaker 2>called real time adaptive tracking. They took that agonizing one

196
00:10:21.000 --> 00:10:25.960
<v Speaker 2>second measurement window and slashed it down to roughly ten milliseconds.

197
00:10:25.279 --> 00:10:27.600
<v Speaker 3>Which is a speed increase of more than one hundred times.

198
00:10:27.759 --> 00:10:30.440
<v Speaker 2>That's insane. So how do you actually do that If

199
00:10:30.480 --> 00:10:32.960
<v Speaker 2>the rules of quantum mechanics dictate that looking at the

200
00:10:33.000 --> 00:10:37.399
<v Speaker 2>quivot destroys the state and you're forced to run thousands

201
00:10:37.399 --> 00:10:40.799
<v Speaker 2>of trials to build that decay curve, how do you

202
00:10:40.879 --> 00:10:43.960
<v Speaker 2>suddenly do it one hundred times faster? Do they just

203
00:10:44.000 --> 00:10:47.480
<v Speaker 2>like buy a faster microwave pulse generator to cycle, the

204
00:10:47.480 --> 00:10:48.240
<v Speaker 2>resets quicker.

205
00:10:48.360 --> 00:10:51.919
<v Speaker 3>No. No, the hardware pulsing speed is already pretty much

206
00:10:51.960 --> 00:10:55.720
<v Speaker 3>near its physical limit. The breakthrough wasn't about pulsing faster,

207
00:10:56.279 --> 00:10:59.879
<v Speaker 3>It was about measuring smarter. Okay. They completely threw out

208
00:11:00.000 --> 00:11:03.600
<v Speaker 3>the traditional brute force method of sweeping through all the

209
00:11:03.639 --> 00:11:08.240
<v Speaker 3>wait times from scratch every single time. Instead, they integrated

210
00:11:08.279 --> 00:11:11.360
<v Speaker 3>an active machine learning inspired feedback loop.

211
00:11:11.399 --> 00:11:13.440
<v Speaker 2>Okay, break that down for me. What does the algorithm

212
00:11:13.480 --> 00:11:14.360
<v Speaker 2>actually do differently?

213
00:11:14.399 --> 00:11:17.360
<v Speaker 3>Think about the traditional method. You're blindly testing weight times

214
00:11:17.360 --> 00:11:20.759
<v Speaker 3>of one microsecond, two microseconds, three microseconds, all the way

215
00:11:20.840 --> 00:11:23.200
<v Speaker 3>up right. But if you already know from the last

216
00:11:23.240 --> 00:11:27.639
<v Speaker 3>measurement that the Quivitt generally survives for fifty microseconds, testing

217
00:11:27.639 --> 00:11:30.080
<v Speaker 3>a one microsecond wait time is a complete waste of

218
00:11:30.080 --> 00:11:30.879
<v Speaker 3>physical time.

219
00:11:30.720 --> 00:11:32.399
<v Speaker 2>Because you already know it's going to survive that long.

220
00:11:32.399 --> 00:11:33.919
<v Speaker 3>You already know the probability is going to be near

221
00:11:33.919 --> 00:11:34.679
<v Speaker 3>one hundred percent.

222
00:11:35.080 --> 00:11:39.360
<v Speaker 2>Ah, So the algorithm stopped searching the entire haystack. It

223
00:11:39.399 --> 00:11:42.159
<v Speaker 2>remembers where the needle was a millisecond ago, and only

224
00:11:42.200 --> 00:11:43.840
<v Speaker 2>searches in that immediate vicinity.

225
00:11:43.960 --> 00:11:47.639
<v Speaker 3>Precisely, they optimize the tracking by using what we can

226
00:11:47.679 --> 00:11:52.080
<v Speaker 3>call a hot and cold approach. The software dynamically calculates

227
00:11:52.120 --> 00:11:55.960
<v Speaker 3>this deepest part of the exponential decay curve, the absolute

228
00:11:56.039 --> 00:11:59.559
<v Speaker 3>most sensitive weight time, where the probability is roughly fifty

229
00:11:59.559 --> 00:12:03.519
<v Speaker 3>to fifty okay, and it focuses its measurement pulses almost

230
00:12:03.720 --> 00:12:06.200
<v Speaker 3>entirely right there on that sensitive edge.

231
00:12:06.279 --> 00:12:08.919
<v Speaker 2>I love that. It's like trying to monitor the edge

232
00:12:08.960 --> 00:12:11.600
<v Speaker 2>of a cliff in the dark. Yes, instead of mapping

233
00:12:11.639 --> 00:12:13.960
<v Speaker 2>the entire mountain from the base to the peak over

234
00:12:14.000 --> 00:12:17.519
<v Speaker 2>and over again, you find the cliff edge once. Then

235
00:12:17.559 --> 00:12:19.600
<v Speaker 2>you just take tiny steps back and forth exactly. If

236
00:12:19.600 --> 00:12:22.639
<v Speaker 2>your foot hits air, you know the edge eroded inward.

237
00:12:22.720 --> 00:12:25.320
<v Speaker 2>If you hit rock, you know it shifted outward. You're

238
00:12:25.360 --> 00:12:27.840
<v Speaker 2>only probing the exact boundary of the change.

239
00:12:27.879 --> 00:12:29.960
<v Speaker 3>And because they're only probing the boundary, they don't need

240
00:12:30.000 --> 00:12:32.120
<v Speaker 3>thousands of data points to find the new curve. They

241
00:12:32.120 --> 00:12:33.000
<v Speaker 3>only need a handful.

242
00:12:33.159 --> 00:12:34.159
<v Speaker 2>That is so smart.

243
00:12:34.480 --> 00:12:38.440
<v Speaker 3>They programmed a field programmable gate array, an ultra fast

244
00:12:38.519 --> 00:12:42.960
<v Speaker 3>FPGA processor to capture the readout data, run a Bayesian

245
00:12:43.080 --> 00:12:46.240
<v Speaker 3>estimation algorithm to calculate the new t one time, and

246
00:12:46.399 --> 00:12:49.799
<v Speaker 3>instantly adjust the weight time for the very next pulse. Wow,

247
00:12:49.960 --> 00:12:53.720
<v Speaker 3>all of that computation and physical adjustment happens entirely on

248
00:12:53.759 --> 00:12:54.279
<v Speaker 3>the fly.

249
00:12:54.399 --> 00:12:57.159
<v Speaker 2>Real time adaptive tracking. So instead of a one second

250
00:12:57.279 --> 00:13:02.159
<v Speaker 2>long exposure photograph, you're taking razor sharp snapshots every ten milliseconds. Yep,

251
00:13:02.440 --> 00:13:06.000
<v Speaker 2>you're watching the quantum system live in high definition as

252
00:13:06.039 --> 00:13:07.519
<v Speaker 2>it breeds and fluctuates.

253
00:13:07.840 --> 00:13:11.159
<v Speaker 3>It really represents a fundamental paradigm shift in quantum observation.

254
00:13:11.960 --> 00:13:15.840
<v Speaker 3>For the first time, researchers can watch the relaxation rate dynamically.

255
00:13:16.519 --> 00:13:18.879
<v Speaker 3>They can observe the peaks and valleys of the quibbet

256
00:13:18.960 --> 00:13:23.320
<v Speaker 3>stability almost perfectly syncd with the actual environmental changes happening

257
00:13:23.360 --> 00:13:24.960
<v Speaker 3>inside the chip, and that leads.

258
00:13:24.799 --> 00:13:27.799
<v Speaker 2>To the ultimate So what of this whole conversation, Now

259
00:13:27.840 --> 00:13:30.519
<v Speaker 2>that we have this ten millisecond high speed camera, how

260
00:13:30.559 --> 00:13:33.559
<v Speaker 2>does this actually help us build a better quantum computer. No,

261
00:13:33.679 --> 00:13:36.799
<v Speaker 2>because observing a problem is great, but it doesn't automatically

262
00:13:36.799 --> 00:13:38.720
<v Speaker 2>fix the fact that the house of cards is falling.

263
00:13:39.159 --> 00:13:42.039
<v Speaker 3>No, it doesn't fix it automatically, but it makes fixing

264
00:13:42.120 --> 00:13:46.720
<v Speaker 3>it physically possible. Observation in physics is the absolute prerequisite

265
00:13:46.759 --> 00:13:50.320
<v Speaker 3>for diagnosis. When your measurement took a full second, all

266
00:13:50.360 --> 00:13:54.600
<v Speaker 3>the different microscopic causes of decoherence blended together. But at

267
00:13:54.679 --> 00:13:58.320
<v Speaker 3>ten milliseconds, you can finally start correlating the rapid drops

268
00:13:58.320 --> 00:14:02.120
<v Speaker 3>in memory stability with SPECIFCI simultaneous physical events.

269
00:14:02.320 --> 00:14:04.799
<v Speaker 2>Oh, I see, it's about matching the time stamp of

270
00:14:04.840 --> 00:14:05.519
<v Speaker 2>the crime to.

271
00:14:05.600 --> 00:14:09.799
<v Speaker 3>The suspect exactly. If the T one relaxation rate suddenly

272
00:14:09.840 --> 00:14:13.919
<v Speaker 3>plummets at exactly millisecond forty, and your other sensors detect

273
00:14:13.919 --> 00:14:17.080
<v Speaker 3>a tiny burst of background radiation hitting the substrate at

274
00:14:17.120 --> 00:14:19.799
<v Speaker 3>exactly millisecond forty, you have a direct correlation.

275
00:14:19.919 --> 00:14:21.000
<v Speaker 2>Have caught them red handed?

276
00:14:21.120 --> 00:14:24.759
<v Speaker 3>You have? You can say definitively, this specific type of

277
00:14:24.799 --> 00:14:28.240
<v Speaker 3>cosmic ray interaction creates a phone on vibration that breaks

278
00:14:28.240 --> 00:14:30.279
<v Speaker 3>the Cooper pairs and kills our kubit memory.

279
00:14:30.399 --> 00:14:32.720
<v Speaker 2>And once you know exactly what is killing the memory,

280
00:14:32.879 --> 00:14:34.879
<v Speaker 2>you can start writing the ultimate hardware path.

281
00:14:34.960 --> 00:14:38.759
<v Speaker 3>You can engineer targeted solutions. You can design novel shielding

282
00:14:38.799 --> 00:14:43.240
<v Speaker 3>materials that specifically absorb that exact frequency of radiation. You

283
00:14:43.279 --> 00:14:47.879
<v Speaker 3>can design microcircuit topographies that isolate the superconducting loops from

284
00:14:47.919 --> 00:14:51.080
<v Speaker 3>the atomic defects in the silicon. You can even write

285
00:14:51.080 --> 00:14:53.200
<v Speaker 3>more efficient error correcting codes.

286
00:14:53.039 --> 00:14:56.279
<v Speaker 2>Because the software will actually know the precise frequency and

287
00:14:56.360 --> 00:14:58.080
<v Speaker 2>nature of the errors it's trying to.

288
00:14:58.120 --> 00:15:02.799
<v Speaker 3>Correct precisely makes quantum engineering out of the dark ages

289
00:15:02.840 --> 00:15:05.200
<v Speaker 3>of trial and error and brings it into an era

290
00:15:05.320 --> 00:15:06.759
<v Speaker 3>of precision diagnostics.

291
00:15:07.039 --> 00:15:09.679
<v Speaker 2>It's incredible to think about the journey here. We went

292
00:15:09.720 --> 00:15:14.240
<v Speaker 2>from superconducting quibits that just randomly hemorrhage their memory while

293
00:15:14.279 --> 00:15:17.240
<v Speaker 2>scientists stared helplessly through a one second.

294
00:15:16.919 --> 00:15:19.000
<v Speaker 3>Blur, a very frustrating blur, to.

295
00:15:19.080 --> 00:15:22.399
<v Speaker 2>The teams at NTNU and the Neils Bohr Institute leveraging

296
00:15:22.519 --> 00:15:26.200
<v Speaker 2>active feedback loops and smart algorithms to slice that window

297
00:15:26.240 --> 00:15:29.480
<v Speaker 2>down to ten milliseconds. They didn't just build a better sensor,

298
00:15:29.879 --> 00:15:33.200
<v Speaker 2>they completely reimagine how to ask the quantum system a question.

299
00:15:33.440 --> 00:15:36.759
<v Speaker 3>It's a profound reminder that in deep tech, a measurement

300
00:15:36.840 --> 00:15:39.600
<v Speaker 3>breakthrough is often the match that lights the fuse for

301
00:15:39.639 --> 00:15:43.600
<v Speaker 3>an engineering revolution. Absolutely without stable memory, quantum computing just

302
00:15:43.639 --> 00:15:47.360
<v Speaker 3>remains a fascinating lab experiment. By giving engineers the ability

303
00:15:47.399 --> 00:15:51.759
<v Speaker 3>to track and diagnose instability in real time, this breakthrough

304
00:15:51.840 --> 00:15:56.039
<v Speaker 3>brings us significantly closer to reality altering computational.

305
00:15:55.360 --> 00:15:59.120
<v Speaker 2>Power, and that leaves us with a truly mind expanding

306
00:15:59.159 --> 00:16:01.039
<v Speaker 2>concept upon As we wrap up today.

307
00:16:01.120 --> 00:16:01.840
<v Speaker 3>Okay, let's hear it.

308
00:16:02.279 --> 00:16:06.200
<v Speaker 2>We've talked entirely about humans using this ten millisecond tracking

309
00:16:06.240 --> 00:16:09.519
<v Speaker 2>to engineer better hardware. But what if we integrate this

310
00:16:09.600 --> 00:16:13.480
<v Speaker 2>algorithm directly into the quantum processor's own operating system?

311
00:16:13.639 --> 00:16:14.720
<v Speaker 3>Oh wow? Right?

312
00:16:15.240 --> 00:16:17.519
<v Speaker 2>If a machine can now track its own memory de

313
00:16:17.600 --> 00:16:20.639
<v Speaker 2>ka in ten milliseconds, how long will it be before

314
00:16:20.679 --> 00:16:22.879
<v Speaker 2>it uses that real time data to predict its own

315
00:16:22.879 --> 00:16:23.879
<v Speaker 2>information loss?

316
00:16:24.120 --> 00:16:25.399
<v Speaker 3>That is fascinating.

317
00:16:25.519 --> 00:16:29.039
<v Speaker 2>Could a future quantum computer sense a microscopic disturbance coming,

318
00:16:29.200 --> 00:16:32.519
<v Speaker 2>realize its t one is plummeting, and actively course correct

319
00:16:32.600 --> 00:16:35.960
<v Speaker 2>or reroute the calculation before the memory even disappears.

320
00:16:35.519 --> 00:16:36.559
<v Speaker 3>Self correcting on the fly.

321
00:16:36.799 --> 00:16:39.559
<v Speaker 2>Exactly are we looking at the dawn of self healing

322
00:16:39.639 --> 00:16:40.600
<v Speaker 2>quantum computers?

323
00:16:40.879 --> 00:16:44.080
<v Speaker 3>That is the frontier we are racing toward. The microscopic

324
00:16:44.120 --> 00:16:47.840
<v Speaker 3>world is full of chaotic noise, but as we learn

325
00:16:47.919 --> 00:16:49.759
<v Speaker 3>to see it clearly, we can learn to master it.

326
00:16:50.519 --> 00:16:54.120
<v Speaker 3>I encourage everyone listening to keep questioning the invisible mechanisms

327
00:16:54.360 --> 00:16:55.679
<v Speaker 3>that power our reality.

328
00:16:55.840 --> 00:16:58.440
<v Speaker 2>Thank you so much for exploring the quantum realm with

329
00:16:58.519 --> 00:17:02.000
<v Speaker 2>us today. Keep wondering, keep asking those hard questions, and

330
00:17:02.120 --> 00:17:03.399
<v Speaker 2>we will see you next time.