WEBVTT

1
00:00:00.080 --> 00:00:02.080
<v Speaker 1>Welcome to the deep dive. We're here to go cut

2
00:00:02.080 --> 00:00:04.160
<v Speaker 1>through the noise and get you the insights that matter.

3
00:00:04.480 --> 00:00:07.440
<v Speaker 1>And today we're tackling a really big one quantum computing.

4
00:00:07.759 --> 00:00:10.199
<v Speaker 1>It's a term you hear a lot. Sounds very sci fi,

5
00:00:10.439 --> 00:00:14.679
<v Speaker 1>very powerful, but honestly it can also feel pretty mysterious,

6
00:00:14.720 --> 00:00:17.800
<v Speaker 1>kind of complex. So our mission for this deep dive

7
00:00:17.960 --> 00:00:21.719
<v Speaker 1>is simple, pull back that curtain. We want to explore

8
00:00:21.760 --> 00:00:25.679
<v Speaker 1>what quantum computing really is, what problems can it actually solve,

9
00:00:26.160 --> 00:00:28.960
<v Speaker 1>and maybe most importantly from any of you, how do

10
00:00:29.000 --> 00:00:32.719
<v Speaker 1>we even start programming these machines. We're leaning heavily on

11
00:00:33.159 --> 00:00:37.560
<v Speaker 1>quantum programming and depth solving problems with Q sharking criskeet today.

12
00:00:37.759 --> 00:00:39.159
<v Speaker 1>It gives nice practical angle.

13
00:00:39.240 --> 00:00:41.840
<v Speaker 2>Absolutely, I'm looking at it now, is well, it's crucial.

14
00:00:41.960 --> 00:00:44.920
<v Speaker 2>Understanding the basics helps us see the real potential, you know,

15
00:00:45.280 --> 00:00:49.079
<v Speaker 2>help sort the genuine breakthroughs from just the hype. It's

16
00:00:49.079 --> 00:00:50.520
<v Speaker 2>about where this is actually heading.

17
00:00:50.600 --> 00:00:53.679
<v Speaker 1>Okay, let's jump right in that the big promise quantum computers.

18
00:00:53.719 --> 00:00:57.200
<v Speaker 1>Are they going to replace our laptops, our phones? Is

19
00:00:57.200 --> 00:00:58.679
<v Speaker 1>that the future we're talking about?

20
00:00:58.920 --> 00:01:02.280
<v Speaker 2>Ah? Yeah, that's the classic question and probably the biggest misconception.

21
00:01:02.880 --> 00:01:06.920
<v Speaker 2>The short answer is no, definitely not. Quantum computing isn't

22
00:01:06.959 --> 00:01:08.760
<v Speaker 2>designed to take over everyday.

23
00:01:08.359 --> 00:01:10.879
<v Speaker 1>Tasks, really not even faster.

24
00:01:11.040 --> 00:01:14.760
<v Speaker 2>Nope. In fact, for things like email, web browsing, even

25
00:01:14.799 --> 00:01:18.480
<v Speaker 2>a lot of standard data processing, a quantum computer would

26
00:01:18.560 --> 00:01:21.120
<v Speaker 2>likely be much much slower slower.

27
00:01:21.280 --> 00:01:22.560
<v Speaker 1>Okay, that's counterintuitive.

28
00:01:22.680 --> 00:01:25.120
<v Speaker 2>It is they're just built differently. Their power isn't in

29
00:01:25.159 --> 00:01:30.120
<v Speaker 2>general speed, it's entackling very specific, highly specialized problems. Think

30
00:01:30.120 --> 00:01:33.920
<v Speaker 2>about problems where classical supercomputers, even the biggest ones, just

31
00:01:34.000 --> 00:01:34.640
<v Speaker 2>hit a wall.

32
00:01:34.680 --> 00:01:35.840
<v Speaker 1>Like what kinds of problems?

33
00:01:35.879 --> 00:01:39.000
<v Speaker 2>Well, material science is a huge one. Simulating exactly how

34
00:01:39.040 --> 00:01:43.840
<v Speaker 2>molecules behave at the quantum level, or simulating other quantum systems.

35
00:01:44.079 --> 00:01:47.079
<v Speaker 2>These are problems where the underlying physics is quantum.

36
00:01:47.239 --> 00:01:50.239
<v Speaker 1>So classical computers struggle because they're trying to imitate something

37
00:01:50.280 --> 00:01:51.239
<v Speaker 1>fundamentally different.

38
00:01:51.480 --> 00:01:56.079
<v Speaker 2>Precisely, the complexity just explodes for classical machines. But if

39
00:01:56.120 --> 00:02:00.640
<v Speaker 2>we could simulate new materials accurately, think about the possibilities

40
00:02:01.040 --> 00:02:05.560
<v Speaker 2>new catalysts for energy, better batteries, maybe room temperature superconductors,

41
00:02:05.640 --> 00:02:09.639
<v Speaker 2>new drugs, it could genuinely revolutionize industries.

42
00:02:09.919 --> 00:02:12.719
<v Speaker 1>Okay, so it's not a faster everything machine. It's more

43
00:02:12.759 --> 00:02:16.919
<v Speaker 1>like a specialized scientific instrument, maybe a super microscope for

44
00:02:16.960 --> 00:02:18.080
<v Speaker 1>certain kinds of problems.

45
00:02:18.159 --> 00:02:20.199
<v Speaker 2>That's a great way to put it. A very powerful,

46
00:02:20.360 --> 00:02:21.560
<v Speaker 2>very specific tool.

47
00:02:21.719 --> 00:02:24.719
<v Speaker 1>So if it's this specialized tool, yeah, what does it

48
00:02:24.719 --> 00:02:26.919
<v Speaker 1>actually look like and how do we interact with it?

49
00:02:26.960 --> 00:02:29.199
<v Speaker 1>Are we talking glowing boxes? Well?

50
00:02:29.240 --> 00:02:32.360
<v Speaker 2>The reality right now is interesting. We're in what John

51
00:02:32.400 --> 00:02:37.879
<v Speaker 2>Preskill called the noisy intermediate scale quantum era NIQ for sure.

52
00:02:37.719 --> 00:02:40.080
<v Speaker 1>A nice que Okay, what does that mean? In plain English?

53
00:02:40.120 --> 00:02:42.240
<v Speaker 2>It means the devices we have now are big enough

54
00:02:42.280 --> 00:02:45.840
<v Speaker 2>that simulating them perfectly on a classical computer is basically impossible.

55
00:02:45.879 --> 00:02:49.159
<v Speaker 2>So they are quantum in a meaningful way, but there's

56
00:02:49.199 --> 00:02:52.560
<v Speaker 2>still relatively small in terms of quid account. And crucially,

57
00:02:52.719 --> 00:02:55.319
<v Speaker 2>they're noisy. They make errors a lot noisy.

58
00:02:55.360 --> 00:02:56.680
<v Speaker 1>That doesn't sound good for computation.

59
00:02:56.960 --> 00:02:59.560
<v Speaker 2>It's the main challenge. It limits the complexity and length

60
00:02:59.560 --> 00:03:03.080
<v Speaker 2>of taculs we can run reliably, and the hardware itself

61
00:03:03.159 --> 00:03:06.759
<v Speaker 2>it's quite diverse. You've got superconducting circuits cooled down to

62
00:03:06.840 --> 00:03:10.960
<v Speaker 2>near absolute zero. There are trapped ions using lasers to

63
00:03:11.080 --> 00:03:15.919
<v Speaker 2>hold and manipulate individual atoms. Also neutral atoms, photonics, lots

64
00:03:15.960 --> 00:03:18.560
<v Speaker 2>of different approaches being researched, huge engineering efforts.

65
00:03:18.599 --> 00:03:22.520
<v Speaker 1>Wow. Okay, so the hardware is complex, developing and noisy.

66
00:03:23.120 --> 00:03:27.639
<v Speaker 1>That sounds challenging to program for Do I need a

67
00:03:27.680 --> 00:03:31.199
<v Speaker 1>PhD in quantum physics just to write Hello World on

68
00:03:31.199 --> 00:03:31.680
<v Speaker 1>one of these?

69
00:03:32.000 --> 00:03:35.000
<v Speaker 2>Hah? Thankfully No, that's another common worry, But it's not

70
00:03:35.039 --> 00:03:38.039
<v Speaker 2>necessarily the case. You can actually learn quantum computing through

71
00:03:38.120 --> 00:03:41.120
<v Speaker 2>quantum programming. Focus on the problem solving aspects first.

72
00:03:41.159 --> 00:03:42.919
<v Speaker 1>How does that work? Isn't the physics kind of.

73
00:03:42.879 --> 00:03:46.360
<v Speaker 2>Essential it underlies everything? Yes, But there's a quantum software stack,

74
00:03:46.479 --> 00:03:49.039
<v Speaker 2>much like the classical one we're used to. It provides

75
00:03:49.120 --> 00:03:52.280
<v Speaker 2>layers of abstraction hardware control the bottom than middle layers

76
00:03:52.280 --> 00:03:55.759
<v Speaker 2>for optimization, error handling, and then the tools. Programmers use

77
00:03:55.960 --> 00:03:59.639
<v Speaker 2>quantum languages like q shag or quiscuit, their compilers, libraries,

78
00:03:59.680 --> 00:04:02.080
<v Speaker 2>and really importantly, quantum simulators.

79
00:04:02.479 --> 00:04:03.120
<v Speaker 1>Simulators.

80
00:04:03.199 --> 00:04:06.599
<v Speaker 2>Yeah, like emulators exactly. They run on your classical computer

81
00:04:06.680 --> 00:04:09.520
<v Speaker 2>and mimic a quantum computer. We'll come back to why

82
00:04:09.520 --> 00:04:12.080
<v Speaker 2>they're so vital. But the point is you can start

83
00:04:12.120 --> 00:04:15.879
<v Speaker 2>programming without mastering all the deep quantum mechanics. First, what

84
00:04:15.919 --> 00:04:19.279
<v Speaker 2>you do need is some comfort with basic linear algebra.

85
00:04:19.399 --> 00:04:23.680
<v Speaker 1>Linear algebra, okay, vectors, matrices, complex numbers, that's.

86
00:04:23.480 --> 00:04:26.920
<v Speaker 2>The language, yes, because quantum states are represented by vectors

87
00:04:26.920 --> 00:04:30.560
<v Speaker 2>and operations are matrices. If you have that foundation, you

88
00:04:30.560 --> 00:04:32.319
<v Speaker 2>can start building algorithms that.

89
00:04:32.279 --> 00:04:36.360
<v Speaker 1>Actually sounds manageable, more like learning a new programming paradigm

90
00:04:36.639 --> 00:04:38.720
<v Speaker 1>than becoming a physicist overnight. It is.

91
00:04:38.839 --> 00:04:39.560
<v Speaker 2>Think of it that way.

92
00:04:39.639 --> 00:04:42.680
<v Speaker 1>Okay, so if we're programming, let's imagine we're sitting down

93
00:04:42.680 --> 00:04:46.000
<v Speaker 1>to write some quantum code. What are the absolute basic

94
00:04:46.040 --> 00:04:48.879
<v Speaker 1>things we can do? The fundamental operations. Yeah, like the

95
00:04:49.040 --> 00:04:52.120
<v Speaker 1>quantum equivalent of setting a variable or doing an addition.

96
00:04:52.360 --> 00:04:55.160
<v Speaker 2>Right. You can break down almost any quantum algorithm into

97
00:04:55.160 --> 00:04:59.600
<v Speaker 2>three fundamental stages. First, state preparation. You start with your quivots,

98
00:04:59.720 --> 00:05:02.720
<v Speaker 2>usually all in a default state, typically the zero state,

99
00:05:03.160 --> 00:05:05.439
<v Speaker 2>analogous to a classical bit being off.

100
00:05:05.600 --> 00:05:07.560
<v Speaker 1>Okay, starting point all zeros.

101
00:05:07.920 --> 00:05:09.839
<v Speaker 2>Then you need to transform them into the state you

102
00:05:09.879 --> 00:05:13.759
<v Speaker 2>actually want for your computation. Often this involves creating a

103
00:05:13.839 --> 00:05:15.360
<v Speaker 2>superposition that's.

104
00:05:15.160 --> 00:05:18.000
<v Speaker 1>The zero in one at the same time thing sort of.

105
00:05:18.079 --> 00:05:21.199
<v Speaker 2>Yeah, it's a combination, a blend of zero and one

106
00:05:21.240 --> 00:05:24.199
<v Speaker 2>are each with a certain weight or amplitude for a

107
00:05:24.199 --> 00:05:27.079
<v Speaker 2>single quibit. A key tool is the right theta gate.

108
00:05:27.480 --> 00:05:29.360
<v Speaker 2>Think of it like a knob you turn to set

109
00:05:29.399 --> 00:05:31.680
<v Speaker 2>the precise mix of zero and one R.

110
00:05:32.079 --> 00:05:34.160
<v Speaker 1>Got it. Prepare the initial mix. What's next?

111
00:05:34.360 --> 00:05:38.519
<v Speaker 2>Second unitary transformations We usually just call them gates. These

112
00:05:38.519 --> 00:05:41.639
<v Speaker 2>are the quantum logic operations. They act on the quibbits

113
00:05:41.720 --> 00:05:44.720
<v Speaker 2>to change their states, maybe entangle them with other cribits.

114
00:05:44.879 --> 00:05:48.199
<v Speaker 2>So like A and D or are not gates in

115
00:05:48.279 --> 00:05:52.639
<v Speaker 2>classical computing analogous, yes, but with a crucial difference. All

116
00:05:52.720 --> 00:05:54.480
<v Speaker 2>quantum gates must be reversible.

117
00:05:54.600 --> 00:05:57.360
<v Speaker 1>Reversible meaning you can always undo them exactly.

118
00:05:57.600 --> 00:05:59.800
<v Speaker 2>If you know the output state and the gay used,

119
00:05:59.839 --> 00:06:02.879
<v Speaker 2>you can perfectly determine the input state. This isn't true

120
00:06:02.879 --> 00:06:05.399
<v Speaker 2>for many classical gates, like A and D. You can't

121
00:06:05.399 --> 00:06:08.360
<v Speaker 2>tell from the output zero whether the inputs were zero, one, zero,

122
00:06:08.399 --> 00:06:09.800
<v Speaker 2>one zero or zero zero.

123
00:06:10.079 --> 00:06:12.800
<v Speaker 1>Huh. Why is reversibility so important?

124
00:06:13.040 --> 00:06:17.240
<v Speaker 2>It's fundamental to quantum mechanics. Information can't just be destroyed.

125
00:06:17.639 --> 00:06:21.560
<v Speaker 2>It means we have to design algorithms differently, carefully preserving

126
00:06:21.639 --> 00:06:24.720
<v Speaker 2>information instead of discarding intermediate results.

127
00:06:24.800 --> 00:06:29.079
<v Speaker 1>Okay, prepare the state, apply reversible gates. What's the third step,

128
00:06:29.319 --> 00:06:34.079
<v Speaker 1>third measurement. This is the only way to get information

129
00:06:34.199 --> 00:06:35.279
<v Speaker 1>out of the quantum computer.

130
00:06:36.240 --> 00:06:38.040
<v Speaker 2>To read the result makes sense.

131
00:06:38.040 --> 00:06:39.639
<v Speaker 1>You have to measure to see what happened.

132
00:06:39.759 --> 00:06:42.360
<v Speaker 2>But here's the quantum catch. If your quibit is in

133
00:06:42.439 --> 00:06:47.319
<v Speaker 2>a superposition, the measurement outcome is probabilistic, non deterministic.

134
00:06:47.600 --> 00:06:49.600
<v Speaker 1>WHOA, So you don't get a definite answer.

135
00:06:49.680 --> 00:06:51.800
<v Speaker 2>You get a definite answer either zero or one for

136
00:06:51.879 --> 00:06:54.720
<v Speaker 2>a single quibit, but which answer you get is determined

137
00:06:54.720 --> 00:06:58.399
<v Speaker 2>by probability. The probability of getting say zero I, is

138
00:06:58.439 --> 00:07:01.560
<v Speaker 2>related to the square of its amplitude in the superposition

139
00:07:01.639 --> 00:07:02.639
<v Speaker 2>before measurement.

140
00:07:02.879 --> 00:07:07.079
<v Speaker 1>So it's like the computation explores many possibilities via superposition,

141
00:07:07.480 --> 00:07:10.639
<v Speaker 1>but when you look you only see one randomly chosen

142
00:07:10.680 --> 00:07:11.920
<v Speaker 1>based on those possibilities.

143
00:07:11.959 --> 00:07:15.079
<v Speaker 2>That's a decent picture. Yeah, The measurement collapses the superposition

144
00:07:15.120 --> 00:07:16.959
<v Speaker 2>into a single classical outcome.

145
00:07:17.199 --> 00:07:20.560
<v Speaker 1>That sounds incredibly tricky to debug. If you run the

146
00:07:20.600 --> 00:07:23.240
<v Speaker 1>same code twice, you might get different answers just due

147
00:07:23.240 --> 00:07:24.639
<v Speaker 1>to the measurement randomness.

148
00:07:24.879 --> 00:07:27.720
<v Speaker 2>You absolutely can, especially if the final state is a

149
00:07:27.720 --> 00:07:31.600
<v Speaker 2>complex superposition. And this brings us back to the simulators

150
00:07:31.600 --> 00:07:32.160
<v Speaker 2>I mentioned.

151
00:07:32.240 --> 00:07:35.759
<v Speaker 1>Ah The classical programs that mimic quantum computers.

152
00:07:35.879 --> 00:07:39.879
<v Speaker 2>Yes, they are a programmer's best friend, especially when learning

153
00:07:40.079 --> 00:07:43.319
<v Speaker 2>why because on a simulator you can cheat.

154
00:07:43.519 --> 00:07:45.160
<v Speaker 1>Cheat how you can.

155
00:07:45.040 --> 00:07:49.399
<v Speaker 2>Pause the simulated execution and peak at the internal quantum state.

156
00:07:49.519 --> 00:07:53.879
<v Speaker 2>Look at all the amplitudes the full superposition without simulating

157
00:07:53.920 --> 00:07:54.519
<v Speaker 2>a measurement.

158
00:07:55.240 --> 00:07:57.240
<v Speaker 1>I see, you can see the probabilities before the dice

159
00:07:57.319 --> 00:07:57.879
<v Speaker 1>roll happens.

160
00:07:58.000 --> 00:08:01.160
<v Speaker 2>Exactly. You can see the full quantum state. That's a

161
00:08:01.240 --> 00:08:04.560
<v Speaker 2>luxury you absolutely do not have on real quantum hardware.

162
00:08:04.759 --> 00:08:07.839
<v Speaker 2>A real measurement is irreversible and collapses the state.

163
00:08:08.120 --> 00:08:11.759
<v Speaker 1>So simulators are crucial for development and debugging, especially for

164
00:08:11.839 --> 00:08:14.160
<v Speaker 1>smaller problems where simulation.

165
00:08:13.759 --> 00:08:17.199
<v Speaker 2>Is feasible invaluable. You test, debug, understand your algorithm on

166
00:08:17.240 --> 00:08:18.040
<v Speaker 2>a simulator first.

167
00:08:18.360 --> 00:08:22.439
<v Speaker 1>Okay, so we have our building blocks. Prepare the state, supersition,

168
00:08:22.720 --> 00:08:29.639
<v Speaker 1>transform it reversible gates, and measure probabilistic outcome. Got it?

169
00:08:30.560 --> 00:08:33.639
<v Speaker 1>How do we use these blocks to actually solve something useful,

170
00:08:34.080 --> 00:08:37.279
<v Speaker 1>maybe something faster than a classical computer. Let's connect this

171
00:08:37.320 --> 00:08:40.639
<v Speaker 1>to a real problem. You mentioned search problems earlier. How

172
00:08:40.639 --> 00:08:44.320
<v Speaker 1>could these quantum operations help with something like well, the

173
00:08:44.399 --> 00:08:46.720
<v Speaker 1>n Queen's puzzle that's a classic computer science.

174
00:08:46.480 --> 00:08:48.879
<v Speaker 2>Problem, right. The n Queen's puzzle is a great example,

175
00:08:49.360 --> 00:08:50.919
<v Speaker 2>and it leads us to one of the most famous

176
00:08:51.000 --> 00:08:53.120
<v Speaker 2>quantum algorithms, Grover's search.

177
00:08:53.279 --> 00:08:55.919
<v Speaker 1>Grover Search. I've heard that what's the core idea.

178
00:08:56.200 --> 00:08:59.639
<v Speaker 2>The core idea is providing a speed up for unstructured search.

179
00:09:00.240 --> 00:09:03.360
<v Speaker 2>Imagine you have a huge list of items and items

180
00:09:03.600 --> 00:09:05.159
<v Speaker 2>and a function that can tell you if an item

181
00:09:05.159 --> 00:09:06.559
<v Speaker 2>is the one you're looking for, but you have no

182
00:09:06.639 --> 00:09:08.840
<v Speaker 2>other information, no sorting.

183
00:09:08.679 --> 00:09:10.600
<v Speaker 1>Nothing, so you just had to check them one by one.

184
00:09:10.799 --> 00:09:13.519
<v Speaker 2>Classically, yes, on average, you'd expect to check about end

185
00:09:13.519 --> 00:09:15.960
<v Speaker 2>two items in the worst case n items. We call

186
00:09:16.000 --> 00:09:21.159
<v Speaker 2>that on complexity. Grover's algorithm, using quantum superposition and a

187
00:09:21.200 --> 00:09:25.000
<v Speaker 2>clever trick called amplitude. Amplification can find the item in

188
00:09:25.159 --> 00:09:27.840
<v Speaker 2>roughly o er in calls to that checking function square

189
00:09:27.879 --> 00:09:28.320
<v Speaker 2>root events.

190
00:09:28.320 --> 00:09:30.240
<v Speaker 1>So for a million items instead of a million checks,

191
00:09:30.320 --> 00:09:30.720
<v Speaker 1>is more.

192
00:09:30.600 --> 00:09:34.559
<v Speaker 2>Like a thousand exactly. It's a quadratic speed up. Now

193
00:09:34.639 --> 00:09:37.639
<v Speaker 2>quadratic isn't the exponential speed up that quantum computing is

194
00:09:37.639 --> 00:09:41.399
<v Speaker 2>truly famous for, like Shores algorithm for factoring, But for

195
00:09:41.519 --> 00:09:45.679
<v Speaker 2>certain problems we're iren is still a massive improvement over in.

196
00:09:45.639 --> 00:09:48.639
<v Speaker 1>Okay, so Grover's speeds up search. How do we apply

197
00:09:48.720 --> 00:09:52.440
<v Speaker 1>it to end Queen's We need that checking function first, right,

198
00:09:52.559 --> 00:09:55.080
<v Speaker 1>a function that tells us if a certain arrangement of

199
00:09:55.159 --> 00:09:56.519
<v Speaker 1>queens is a valid.

200
00:09:56.200 --> 00:09:59.919
<v Speaker 2>Solution precisely, and we need to implement that classical TI

201
00:10:00.080 --> 00:10:03.000
<v Speaker 2>checking function on a quantum computer. This brings us back

202
00:10:03.000 --> 00:10:04.279
<v Speaker 2>to reversible.

203
00:10:03.759 --> 00:10:06.519
<v Speaker 1>Computing, because all quantum gates are reversible.

204
00:10:06.679 --> 00:10:09.120
<v Speaker 2>Right. We can't just compute FX where x is the

205
00:10:09.120 --> 00:10:12.559
<v Speaker 2>board arrangement and FX is true or false if it's

206
00:10:12.600 --> 00:10:15.279
<v Speaker 2>a solution, and then discard X. We have to preserve

207
00:10:15.320 --> 00:10:15.799
<v Speaker 2>the input.

208
00:10:15.879 --> 00:10:16.559
<v Speaker 1>How do we do that?

209
00:10:16.679 --> 00:10:20.039
<v Speaker 2>Typically we use an extra enciloquibit. We transform a state

210
00:10:20.200 --> 00:10:24.080
<v Speaker 2>x zero, input x and selo zero into exocuibit. If

211
00:10:24.240 --> 00:10:27.279
<v Speaker 2>sx is true one, the incilla flips to one. If

212
00:10:27.320 --> 00:10:30.120
<v Speaker 2>false zero, it stays zero. The input x ra is

213
00:10:30.240 --> 00:10:31.600
<v Speaker 2>preserved ice.

214
00:10:31.360 --> 00:10:33.519
<v Speaker 1>So the answer is stored in the extra cubit without

215
00:10:33.559 --> 00:10:36.000
<v Speaker 1>losing the original board state. What kind of gates do that?

216
00:10:36.360 --> 00:10:40.320
<v Speaker 2>Multiquibit Gates like cnot control not T, and especially to

217
00:10:40.399 --> 00:10:43.399
<v Speaker 2>fully gates C cnot controlled, controlled not T are the

218
00:10:43.480 --> 00:10:46.879
<v Speaker 2>building blocks for creating these reversible classical functions.

219
00:10:46.480 --> 00:10:49.919
<v Speaker 1>Okay, so we can build a reversible n queen's checker. Now,

220
00:10:49.960 --> 00:10:52.039
<v Speaker 1>how do we represent the board x itself?

221
00:10:52.240 --> 00:10:54.120
<v Speaker 2>Ah, And this is where it gets really interesting. This

222
00:10:54.200 --> 00:10:57.799
<v Speaker 2>is the AHA moment for quantum algorithm design. How you

223
00:10:57.919 --> 00:11:01.279
<v Speaker 2>encode the problem matters enormously for a fency. Let's take

224
00:11:01.399 --> 00:11:04.360
<v Speaker 2>end four, a small four x four board. A naive

225
00:11:04.440 --> 00:11:08.039
<v Speaker 2>encoding might be use one quibit for each square on

226
00:11:08.080 --> 00:11:10.399
<v Speaker 2>the board one if there's a queen, zero if not.

227
00:11:10.559 --> 00:11:13.360
<v Speaker 1>Okay, that's four x four sixteen quibits.

228
00:11:12.960 --> 00:11:14.720
<v Speaker 2>Right, and the total number of possible states is two

229
00:11:14.799 --> 00:11:17.399
<v Speaker 2>hundred and sixteen, which is sixty five five hundred and

230
00:11:17.399 --> 00:11:20.360
<v Speaker 2>thirty six. Grover's algorithm would have to search through this

231
00:11:20.559 --> 00:11:23.720
<v Speaker 2>huge space. That's a lot of quibits and potentially many

232
00:11:23.720 --> 00:11:25.399
<v Speaker 2>many Grover iterations needed.

233
00:11:25.600 --> 00:11:27.360
<v Speaker 1>Seems inefficient? Is there a better way?

234
00:11:27.480 --> 00:11:30.960
<v Speaker 2>Definitely and optimize encoding leverages the rules of the puzzle.

235
00:11:31.000 --> 00:11:34.559
<v Speaker 2>We know there must be exactly one queen per row, right,

236
00:11:34.720 --> 00:11:37.399
<v Speaker 2>so instead of representing every square, we just need to

237
00:11:37.440 --> 00:11:39.679
<v Speaker 2>represent the column position of the queen in each row.

238
00:11:39.960 --> 00:11:43.360
<v Speaker 2>For n four, the columns are zero, one, two, three.

239
00:11:43.440 --> 00:11:45.440
<v Speaker 2>How many bits do you need to represent those numbers?

240
00:11:45.799 --> 00:11:49.200
<v Speaker 1>Two bits? Can represent four values zero, zero, zero, one.

241
00:11:49.360 --> 00:11:51.960
<v Speaker 2>Ten, eleven exactly, so two bits per row and there

242
00:11:51.960 --> 00:11:54.399
<v Speaker 2>are four rows. That means we only need four rows

243
00:11:54.440 --> 00:11:56.559
<v Speaker 2>two bits row. What is eight quippits total?

244
00:11:56.639 --> 00:11:59.080
<v Speaker 1>Wow, eight quibits instead of sixteen. That's a big difference.

245
00:11:59.240 --> 00:12:01.519
<v Speaker 2>And look at the search space. It's now just the

246
00:12:01.600 --> 00:12:03.759
<v Speaker 2>number of ways to place one queen in each row,

247
00:12:03.919 --> 00:12:06.000
<v Speaker 2>which is four choices for the first row, four for

248
00:12:06.039 --> 00:12:08.840
<v Speaker 2>the second, and so on. Forty four equals two hundred

249
00:12:08.879 --> 00:12:10.200
<v Speaker 2>and fifty six states, two.

250
00:12:10.159 --> 00:12:12.279
<v Speaker 1>Hundred and fifty six instead of sixty five thousand, five

251
00:12:12.399 --> 00:12:13.080
<v Speaker 1>hundred and thirty six.

252
00:12:13.120 --> 00:12:16.279
<v Speaker 2>That's massively smaller, huge difference. Now, searching two hundred and

253
00:12:16.279 --> 00:12:19.559
<v Speaker 2>fifty six states with Grover takes far fewer iterations, roughly

254
00:12:19.759 --> 00:12:22.000
<v Speaker 2>two hundred and fifty six, which is about sixteen, but

255
00:12:22.080 --> 00:12:24.639
<v Speaker 2>the exact number is around nine iterations for n four.

256
00:12:24.840 --> 00:12:27.559
<v Speaker 2>Compare that to searching sixty five thousand, five hundred and

257
00:12:27.559 --> 00:12:30.799
<v Speaker 2>thirty six states. It's potentially hundreds or thousands of iterations.

258
00:12:31.039 --> 00:12:34.480
<v Speaker 1>That really drives the point home. The cleverness isn't just

259
00:12:34.559 --> 00:12:37.960
<v Speaker 1>in the Grover algorithm itself, but in how you set

260
00:12:38.039 --> 00:12:40.240
<v Speaker 1>up the problem for the quantum computer.

261
00:12:40.759 --> 00:12:44.360
<v Speaker 2>The encoding is key, It's absolutely fundamental. Choosing the right

262
00:12:44.399 --> 00:12:47.600
<v Speaker 2>representation can be the difference between a feasible quantum solution

263
00:12:48.000 --> 00:12:49.840
<v Speaker 2>and one that's completely impractical.

264
00:12:50.200 --> 00:12:53.960
<v Speaker 1>This n Queen's example is powerful. It shows the potential

265
00:12:54.240 --> 00:12:56.919
<v Speaker 1>but also the subtlety. But it also brings us back

266
00:12:56.919 --> 00:12:59.720
<v Speaker 1>to that in ice Q thing you mentioned earlier, the noise.

267
00:13:00.559 --> 00:13:05.399
<v Speaker 1>How do these algorithms actually perform on real current noisy hardware,

268
00:13:05.679 --> 00:13:09.159
<v Speaker 1>and what's the path forward to running larger problems reliably?

269
00:13:09.399 --> 00:13:12.679
<v Speaker 2>That's the billion dollar question. Really. The NISQ devices we

270
00:13:12.720 --> 00:13:16.039
<v Speaker 2>have today, well, they struggle with noise. Error rates are

271
00:13:16.039 --> 00:13:19.480
<v Speaker 2>relatively high. Running grovers for even moderately large n or

272
00:13:19.480 --> 00:13:23.759
<v Speaker 2>more complex algorithms is extremely challenging. The errors accumulate and

273
00:13:23.799 --> 00:13:24.799
<v Speaker 2>destroy the computation.

274
00:13:25.080 --> 00:13:28.799
<v Speaker 1>So what's the solution? Just build better, less noisy hardware.

275
00:13:28.919 --> 00:13:31.480
<v Speaker 2>That's part of it, definitely, but the long term vision,

276
00:13:31.519 --> 00:13:34.720
<v Speaker 2>the ultimate goal is fault tolerant quantum computers.

277
00:13:34.360 --> 00:13:37.240
<v Speaker 1>Fault tolerant meaning they can correct their own errors.

278
00:13:37.200 --> 00:13:42.360
<v Speaker 2>Exactly using principles of quantum error correction. It's conceptually similar

279
00:13:42.360 --> 00:13:46.039
<v Speaker 2>to error correction in classical computers or communication, but much

280
00:13:46.080 --> 00:13:48.519
<v Speaker 2>more complex due to the nature of quantum states.

281
00:13:48.600 --> 00:13:49.200
<v Speaker 1>How does it work?

282
00:13:49.320 --> 00:13:52.639
<v Speaker 2>Roughly, the core idea is redundancy. You don't use just

283
00:13:52.720 --> 00:13:55.720
<v Speaker 2>one physical quibut to represent the information you care about,

284
00:13:56.159 --> 00:13:59.559
<v Speaker 2>instead a single logical quibit. The quibit your algorithm actually

285
00:13:59.600 --> 00:14:03.679
<v Speaker 2>uses is encoded across many physical quibots on the hardware.

286
00:14:03.440 --> 00:14:06.200
<v Speaker 1>So one quibot in my program might actually be say

287
00:14:06.559 --> 00:14:09.200
<v Speaker 1>ten or one hundred physical quibots working together.

288
00:14:09.159 --> 00:14:11.399
<v Speaker 2>It could be hundreds or even thousands, depending on the

289
00:14:11.440 --> 00:14:15.639
<v Speaker 2>code and the desired level of protection. This introduces enormous.

290
00:14:15.240 --> 00:14:18.279
<v Speaker 1>Overhead overhead, meaning extra resources.

291
00:14:17.759 --> 00:14:21.840
<v Speaker 2>Massive extra resources. A single logical gait operation in your

292
00:14:21.879 --> 00:14:26.320
<v Speaker 2>algorithm might require performing dozens, hundreds, maybe more operations on

293
00:14:26.360 --> 00:14:29.879
<v Speaker 2>the underlying physical quibuts to detect and correct errors while

294
00:14:29.879 --> 00:14:33.679
<v Speaker 2>preserving the quantum state, and some operations are particularly costly,

295
00:14:34.000 --> 00:14:37.200
<v Speaker 2>things like arbitrary rotations those rye or RS gates we

296
00:14:37.279 --> 00:14:41.519
<v Speaker 2>mentioned for state preparation. Implementing them tolerantly often requires something

297
00:14:41.559 --> 00:14:42.679
<v Speaker 2>called magic states.

298
00:14:43.039 --> 00:14:46.240
<v Speaker 1>Magic states sounds exotic, they are.

299
00:14:46.440 --> 00:14:50.159
<v Speaker 2>There are special, highly entangled resource states that have to

300
00:14:50.200 --> 00:14:54.480
<v Speaker 2>be prepared and consumed, adding significant complexity and cost to

301
00:14:54.519 --> 00:14:58.759
<v Speaker 2>the computation. It's like needing a very rare, expensive tatalyst

302
00:14:58.840 --> 00:14:59.960
<v Speaker 2>for certain reactions.

303
00:15:00.360 --> 00:15:03.960
<v Speaker 1>Okay, so fault tolerance is the goal, but it comes

304
00:15:04.120 --> 00:15:07.320
<v Speaker 1>at a potentially huge cost in terms of the number

305
00:15:07.360 --> 00:15:10.759
<v Speaker 1>of physical quibits and operations needed. This sounds like it

306
00:15:10.799 --> 00:15:13.399
<v Speaker 1>pushes practical quantum computing further into the future.

307
00:15:13.480 --> 00:15:16.919
<v Speaker 2>It's definitely a long term engineering challenge. We're talking potentially

308
00:15:17.000 --> 00:15:20.399
<v Speaker 2>millions of physical quibits for some applications. But it also

309
00:15:20.480 --> 00:15:23.320
<v Speaker 2>highlights why just looking at the theoretical number of steps

310
00:15:23.320 --> 00:15:26.440
<v Speaker 2>in an algorithm like oh you friend for Grover isn't

311
00:15:26.480 --> 00:15:27.279
<v Speaker 2>the full story.

312
00:15:27.480 --> 00:15:29.679
<v Speaker 1>You need to consider the physical cost of running it

313
00:15:29.759 --> 00:15:31.240
<v Speaker 1>reliably precisely.

314
00:15:31.399 --> 00:15:33.399
<v Speaker 2>And that's where another set of crucial tools comes in

315
00:15:33.600 --> 00:15:35.120
<v Speaker 2>Quantum resource estimator.

316
00:15:35.200 --> 00:15:37.240
<v Speaker 1>What do they do? Estimate the resources needed?

317
00:15:37.320 --> 00:15:40.399
<v Speaker 2>Yes, tools like the Azure Quantum Resource Estimator allow you

318
00:15:40.480 --> 00:15:43.519
<v Speaker 2>to take your quantum algorithm, specify assumptions about the future

319
00:15:43.600 --> 00:15:47.120
<v Speaker 2>fault tolerant hardware like quibet quality operation speeds, and it

320
00:15:47.240 --> 00:15:48.600
<v Speaker 2>estimates what you'd actually need.

321
00:15:48.919 --> 00:15:50.200
<v Speaker 1>What kind of estimates does it give?

322
00:15:50.519 --> 00:15:53.799
<v Speaker 2>It estimates the total number of physical creepits required and

323
00:15:53.960 --> 00:15:58.200
<v Speaker 2>the expected runtime to execute your algorithm fault tolerantly. This

324
00:15:58.279 --> 00:16:01.000
<v Speaker 2>gives a much more realistic picture of the true cost.

325
00:16:01.200 --> 00:16:03.840
<v Speaker 1>Can we apply that back to our end For Queen's example,

326
00:16:03.840 --> 00:16:07.120
<v Speaker 1>the optimized one with eight logical quivits, what would that

327
00:16:07.159 --> 00:16:08.960
<v Speaker 1>look like on a fault tolerant machine.

328
00:16:09.080 --> 00:16:11.960
<v Speaker 2>It's a great test case running the resource estimator on

329
00:16:12.000 --> 00:16:16.080
<v Speaker 2>that optimized and for grover Search, well, depending on the

330
00:16:16.120 --> 00:16:19.159
<v Speaker 2>specific assumptions about the error correction code and hardware, you

331
00:16:19.279 --> 00:16:21.720
<v Speaker 2>might be looking at something like, say three minutes of runtime,

332
00:16:21.759 --> 00:16:24.720
<v Speaker 2>but requiring around three hundred and forty thousand physical quipits.

333
00:16:25.039 --> 00:16:27.519
<v Speaker 1>Three hundred and forty thousand physical equipments for just eight

334
00:16:27.600 --> 00:16:30.360
<v Speaker 1>logical quibuts solving a tiny four by four puzzle.

335
00:16:30.440 --> 00:16:33.960
<v Speaker 2>That's the kind of overhead we're talking about for fault tolerance. Now,

336
00:16:34.000 --> 00:16:37.000
<v Speaker 2>these numbers depend heavily on the assumptions, but it illustrates

337
00:16:37.039 --> 00:16:39.879
<v Speaker 2>the scale, and this kind of analysis is vital. It

338
00:16:39.960 --> 00:16:42.759
<v Speaker 2>lets us compare the projected cost of a quantum solution

339
00:16:43.279 --> 00:16:46.519
<v Speaker 2>against the best known classical solutions for a problem, not

340
00:16:46.679 --> 00:16:50.159
<v Speaker 2>just comparing Grover's N against a naive n classical search.

341
00:16:50.559 --> 00:16:53.720
<v Speaker 2>We need realistic comparisons to see where the true quantum

342
00:16:53.759 --> 00:16:56.080
<v Speaker 2>advantage might lie for practical problems.

343
00:16:56.120 --> 00:16:59.480
<v Speaker 1>That's incredibly sobering but also clarifying. It paints a much

344
00:16:59.519 --> 00:17:01.799
<v Speaker 1>clearer pickre sure of the challenges and the path ahead.

345
00:17:02.080 --> 00:17:05.440
<v Speaker 2>Wow, we've really covered a lot of ground today, from

346
00:17:05.960 --> 00:17:09.519
<v Speaker 2>cutting through that initial hype around quantum computing replacing everything,

347
00:17:09.880 --> 00:17:14.160
<v Speaker 2>to understanding its real niche in specialized problems, peeking inside

348
00:17:14.160 --> 00:17:16.319
<v Speaker 2>the noisy hardware and getting a feel for the basic

349
00:17:16.400 --> 00:17:20.440
<v Speaker 2>programming steps, preparing states, applying gates, measuring and seeing Grover's

350
00:17:20.480 --> 00:17:23.839
<v Speaker 2>algorithm in action with En Queen's highlighting just how critical

351
00:17:23.960 --> 00:17:27.920
<v Speaker 2>that probleming coding is. Then finally facing the reality of noise,

352
00:17:28.039 --> 00:17:31.119
<v Speaker 2>fault tolerance and the massive resources needed. It's been quite

353
00:17:31.119 --> 00:17:33.839
<v Speaker 2>the journey, it really has. I think the key takeaways

354
00:17:33.839 --> 00:17:38.319
<v Speaker 2>are quantum computing is specialized, not general purpose. It holds

355
00:17:38.359 --> 00:17:42.319
<v Speaker 2>immense promise for fields like material science and drug discovery,

356
00:17:43.079 --> 00:17:46.720
<v Speaker 2>but we're still in the early noisy and ic Q days.

357
00:17:47.359 --> 00:17:50.079
<v Speaker 2>Fault tolerance through error correction is the goal, but it

358
00:17:50.119 --> 00:17:53.799
<v Speaker 2>brings significant overhead and as we saw, clever algorithm design

359
00:17:53.799 --> 00:17:56.920
<v Speaker 2>and smart problem in coding are crucial. Plus we need

360
00:17:56.960 --> 00:18:00.200
<v Speaker 2>tools like resource estimators to realistically gauge the path to

361
00:18:00.240 --> 00:18:01.279
<v Speaker 2>practical advantage.

362
00:18:01.359 --> 00:18:04.680
<v Speaker 1>Absolutely that in Queen's encoding example really stuck with me.

363
00:18:04.759 --> 00:18:08.880
<v Speaker 1>How drastically changing the representation altered the feasibility. It makes

364
00:18:08.880 --> 00:18:11.880
<v Speaker 1>you think, what other classical problems out there, Maybe problems

365
00:18:11.920 --> 00:18:14.319
<v Speaker 1>we think are just hard, could potentially unlock a quantum

366
00:18:14.359 --> 00:18:16.920
<v Speaker 1>speed up if we just found a more ingenious way

367
00:18:16.960 --> 00:18:19.559
<v Speaker 1>to encode them for a quantum computer. Something for all

368
00:18:19.599 --> 00:18:22.279
<v Speaker 1>of you to maybe ponder, what problem could you reimagine

369
00:18:22.319 --> 00:18:24.440
<v Speaker 1>for the quantum realm. That's all the time we have

370
00:18:24.480 --> 00:18:26.599
<v Speaker 1>for this deep dive. Thanks for joining us and keep

371
00:18:26.599 --> 00:18:27.079
<v Speaker 1>exploring