WEBVTT 1 00:00:00.080 --> 00:00:02.040 You know that feeling right when you need to get 2 00:00:02.080 --> 00:00:04.879 up to speed on something really complex fast, but you're 3 00:00:04.919 --> 00:00:06.519 just drowning in articles and papers. 4 00:00:06.639 --> 00:00:09.160 Yeah, totally information overload exactly. 5 00:00:09.919 --> 00:00:12.080 So we're here to give you that shortcut, that deep 6 00:00:12.160 --> 00:00:16.079 dive to get you properly informed. Today we're jumping into 7 00:00:16.120 --> 00:00:19.719 a field that's honestly reshaping how we even think about computation. 8 00:00:19.960 --> 00:00:22.839 It really is. We're talking quantum computing, and our mission 9 00:00:22.879 --> 00:00:27.039 today is to untack the core ideas, look at the 10 00:00:27.160 --> 00:00:30.839 tools making it accessible, which is amazing, and even touch 11 00:00:30.920 --> 00:00:33.159 on some algorithms that sound like science. 12 00:00:32.880 --> 00:00:35.600 Fiction, they really do. Our main guide here is the 13 00:00:35.640 --> 00:00:39.439 second edition of Learn Quantum Computing with Python and IBM 14 00:00:39.560 --> 00:00:43.439 Quantum by Robert Laredo. He's an IBM Mastered inventor, a 15 00:00:43.520 --> 00:00:46.079 quantum ambassador, just brilliant and. 16 00:00:46.039 --> 00:00:48.840 A great guy too. Donated his previous books royalties to 17 00:00:48.920 --> 00:00:52.399 Doctors Without Borders. We're also pulling in insights from reviewers 18 00:00:52.399 --> 00:00:56.119 like Haci Norlen, Mikhail Grossey, Sean Wagner, all deep in 19 00:00:56.200 --> 00:00:58.520 the IBM Quantum and caskit world. 20 00:00:58.600 --> 00:01:01.119 So whether you're prepping for that big meeting or just 21 00:01:01.200 --> 00:01:03.159 trying to figure out what all the buzz is about. 22 00:01:03.479 --> 00:01:06.760 Or maybe you're just super curious, which you should be. Huh. Yeah, 23 00:01:07.120 --> 00:01:10.319 this deep dive should give you those aha moments. We're 24 00:01:10.319 --> 00:01:15.840 aiming to show how these really abstract ideas are becoming concrete. 25 00:01:15.359 --> 00:01:18.400 Tools, right without making your head explode hopefully. 26 00:01:19.040 --> 00:01:22.920 Okay, so let's start at the beginning. Quantum mechanics sounds intimidating, 27 00:01:22.959 --> 00:01:25.000 but it's the bedrock of all this. We're talking about 28 00:01:25.200 --> 00:01:28.280 the rules governing the universe at the tiniest scales, the 29 00:01:28.319 --> 00:01:29.400 sub atomic level. 30 00:01:29.560 --> 00:01:34.519 Yeah, the foundations laid by you know, the giants Schrodinger, Plank, Heisenberg, 31 00:01:34.640 --> 00:01:38.560 Drach Einstein. Their work described a world that's just fundamentally 32 00:01:38.640 --> 00:01:42.519 weird compared to our everyday experience, and that weirdness is key, right, 33 00:01:42.560 --> 00:01:45.400 Like the double slit experiment exactly, that's a classic. You 34 00:01:45.439 --> 00:01:49.519 fire single particles like photons at two slits, you'd expect 35 00:01:49.519 --> 00:01:51.400 two lines on the screen behind it, right, Yeah, it 36 00:01:51.480 --> 00:01:53.200 makes sense, but you don't get that. You get an 37 00:01:53.239 --> 00:01:56.640 interference pattern like waves canceling out and adding up. It 38 00:01:56.719 --> 00:02:01.200 shows particles behaving like waves. This wave part duality. It's 39 00:02:01.239 --> 00:02:05.200 not just a cool fact, it's something Quantum computers actively use. 40 00:02:05.480 --> 00:02:09.280 So we all get classical bits on off zero one 41 00:02:09.680 --> 00:02:14.159 transistors doing their thing. But the quantum version, the quibit. Yeah, 42 00:02:14.199 --> 00:02:16.199 that's the heart of it. What makes it so different. 43 00:02:16.319 --> 00:02:20.479 Well, the huge difference is superposition Equibit isn't just zero 44 00:02:20.639 --> 00:02:22.759 or one. It can be zero, it can be one, 45 00:02:22.960 --> 00:02:25.439 or it can be a combination of both at the 46 00:02:25.479 --> 00:02:26.240 exact same. 47 00:02:26.039 --> 00:02:28.159 Time, both like simultaneously. 48 00:02:28.400 --> 00:02:30.319 Kind of think of it less like a hidden state 49 00:02:30.439 --> 00:02:33.680 like a spinning coin before it lands, and more like, well, 50 00:02:33.719 --> 00:02:36.759 the quantum state itself is a blend of possibilities until 51 00:02:36.840 --> 00:02:39.960 you measure it. Measuring forces it to choose zero or one. 52 00:02:40.159 --> 00:02:42.719 You mentioned a spinning coin, but you said it's not 53 00:02:42.879 --> 00:02:47.039 like that. Yeah, that the quantum outcome is truly random. 54 00:02:47.120 --> 00:02:50.280 That sounds profound. Are we saying there's no hidden instruction 55 00:02:50.360 --> 00:02:51.319 determining the outcome? 56 00:02:51.439 --> 00:02:55.240 That's exactly what standard quantum mechanics says. Yeah, it's fundamentally probabilistic, 57 00:02:55.319 --> 00:02:57.800 not just we don't know enough, but actually random. And 58 00:02:57.840 --> 00:03:01.560 that's a massive conceptual leap. It means quantum computers work 59 00:03:01.599 --> 00:03:03.240 with probabilities directly. 60 00:03:03.240 --> 00:03:05.639 Wow, Okay, that changes things. 61 00:03:05.479 --> 00:03:09.280 It does, and that leads to the second key, difference entanglement. 62 00:03:09.599 --> 00:03:12.599 This is the one Einstein called famously spooky action at 63 00:03:12.599 --> 00:03:13.120 a distance. 64 00:03:13.479 --> 00:03:16.560 Spooky action. I love that phrase. What does it actually 65 00:03:16.599 --> 00:03:17.479 mean for quibits? 66 00:03:17.719 --> 00:03:20.599 It means you can link two or more quibits together 67 00:03:20.680 --> 00:03:23.879 in a special way, so intimately linked that measuring the 68 00:03:23.879 --> 00:03:26.560 state of one quibbit instantly tells you the state of 69 00:03:26.560 --> 00:03:28.919 the other, no matter how far apart they are, instantly 70 00:03:28.960 --> 00:03:32.599 across any distance, instantly. If you have two entangled quibits 71 00:03:32.639 --> 00:03:34.879 and measure one is zero, you know the other is zero. 72 00:03:35.240 --> 00:03:38.000 If one's one, the other's one, always correlate. It's like 73 00:03:38.000 --> 00:03:41.639 they're still connected, sharing the same fate. It's a crucial resource. 74 00:03:41.919 --> 00:03:47.400 Okay, superposition entanglement. Yeah, and you mentioned interference. Going back 75 00:03:47.439 --> 00:03:49.960 to the double slit thing. How does that play into algorithms? 76 00:03:50.120 --> 00:03:53.360 Right? Interference Quantum algorithms are designed to leverage this. Think 77 00:03:53.400 --> 00:03:56.520 of it like paths leading to the wrong answer interfere 78 00:03:56.599 --> 00:03:58.800 destructively and cancel each other out. 79 00:03:59.080 --> 00:04:01.120 Ah, so they just disappear effectively. 80 00:04:01.240 --> 00:04:05.360 Yeah, while the paths leading to the right answer interfere constructively, 81 00:04:05.439 --> 00:04:08.759 reinforcing each other. So the probability of measuring the correct 82 00:04:08.759 --> 00:04:13.159 answer gets amplified. It's a way to sift through possibilities incredibly. 83 00:04:12.680 --> 00:04:17.079 Efficiently and visualizing this yeah, it's hard to picture. Are 84 00:04:17.079 --> 00:04:18.040 there tools for that? 85 00:04:18.439 --> 00:04:21.120 There are. For a single quibut, we often use the 86 00:04:21.160 --> 00:04:25.519 block sphere. It helps visualize its state, including its amplitude 87 00:04:25.519 --> 00:04:28.040 and phase, like a point on a globe. But for 88 00:04:28.240 --> 00:04:32.000 multiple entangled quibits that gets complicated. So there's a more 89 00:04:32.040 --> 00:04:34.959 powerful tool called the q sphere, which can show the 90 00:04:35.120 --> 00:04:39.959 combined state of many quibits, including those crucial phase relationships 91 00:04:40.000 --> 00:04:40.560 between them. 92 00:04:40.959 --> 00:04:44.319 This is fascinating, but maybe a bit abstract. How do 93 00:04:44.360 --> 00:04:46.439 people actually use this stuff? Where do you start? 94 00:04:46.759 --> 00:04:50.000 Great question? For many. The entry point is the IBM 95 00:04:50.120 --> 00:04:53.959 Quantum Platform or IQP, and it's got a cool history. 96 00:04:54.800 --> 00:04:58.040 IBM actually put the first quantum computer on the cloud 97 00:04:58.199 --> 00:05:01.199 free for anyone to use, back in May twenty sixteen. 98 00:05:01.360 --> 00:05:04.680 Wow, twenty sixteen. That feels like ages ago in this field, right. 99 00:05:04.839 --> 00:05:06.759 It was called the IBM Quantum Experience. 100 00:05:06.800 --> 00:05:06.959 Then. 101 00:05:07.120 --> 00:05:09.600 Now it's this much more developed platform, the ITP. 102 00:05:09.920 --> 00:05:11.040 What's actually on the platform. 103 00:05:11.079 --> 00:05:13.560 It's got a few key parts. The main platform view 104 00:05:13.639 --> 00:05:16.680 is your dashboard. You can see which quantum computers are online, 105 00:05:16.680 --> 00:05:18.920 how busy they are, check on your jobs, see how 106 00:05:19.000 --> 00:05:20.079 much computing time you've. 107 00:05:20.000 --> 00:05:22.079 Used, like a mission control kind of. Yeah. 108 00:05:22.560 --> 00:05:25.680 Then there's the documentation which is huge. Everything you need 109 00:05:25.720 --> 00:05:29.079 to set up your coding environment, understand Kuskit, which we'll 110 00:05:29.079 --> 00:05:31.560 get to its APIs. It's all there. 111 00:05:31.879 --> 00:05:33.560 And you mentioned learning resources. 112 00:05:33.639 --> 00:05:37.560 Oh yeah, the learning section is fantastic courses, tutorials from 113 00:05:37.600 --> 00:05:41.800 beginner stuff right up to advanced topics, even quantum safe cryptography. 114 00:05:42.160 --> 00:05:45.360 It really lowers the barrier to entry inside the platform. 115 00:05:45.399 --> 00:05:48.439 I heard there's something called the composer, like a drag 116 00:05:48.480 --> 00:05:49.360 and drop interface. 117 00:05:49.519 --> 00:05:52.879 That's right. The composer is brilliant for visual learners or 118 00:05:52.959 --> 00:05:56.079 just starting out. It's a graphical interface where you literally 119 00:05:56.240 --> 00:06:00.000 drag quantum gates onto quibit lines to build a circuit. 120 00:06:00.000 --> 00:06:00.519 Oh cool. 121 00:06:00.639 --> 00:06:02.959 And the best part is the real time feedback. As 122 00:06:02.959 --> 00:06:06.480 you build, you see visualizations update instantly, the state vector 123 00:06:06.560 --> 00:06:10.519 view showing amplitudes, the probabilities view predicting outcomes. Even that 124 00:06:10.600 --> 00:06:13.079 q sphere we talked about, you can pop a not 125 00:06:13.240 --> 00:06:15.680 t git on there and see the probability flip from 126 00:06:15.680 --> 00:06:19.240 one hundred percent zero to one hundred percent one right away. 127 00:06:19.160 --> 00:06:20.879 So you can really see cause and effects. 128 00:06:20.959 --> 00:06:22.439 Exactly. It makes it much more intuitive. 129 00:06:22.480 --> 00:06:24.720 Okay, but what if you want to go beyond dragging 130 00:06:24.759 --> 00:06:25.959 and dropping. You mentioned quiske It. 131 00:06:26.360 --> 00:06:30.639 Yes, kiss git. That's IBM's open source software development kit 132 00:06:30.759 --> 00:06:34.399 for quantum computing. It's built on Python, which makes it 133 00:06:34.399 --> 00:06:37.879 accessible to a huge community of developers. It's really the 134 00:06:38.040 --> 00:06:41.920 standard for working with IBM's quantum systems programmatically. 135 00:06:42.160 --> 00:06:44.120 Is it just one monolithic. 136 00:06:43.720 --> 00:06:47.600 Thing or no, It's cleverly layered. They think about different 137 00:06:47.600 --> 00:06:50.240 developer roles. At the lowest level. You have kernel developers. 138 00:06:50.240 --> 00:06:53.600 They're working right down near the hardware, figuring out the 139 00:06:53.639 --> 00:06:56.800 precise microwave pulses needed to manipulate the quibbits. 140 00:06:57.079 --> 00:06:59.920 Really deep stuff, okay, Hardware whisperers, ah yeah. 141 00:07:00.360 --> 00:07:03.600 Then you have algorithm developers. They're using quiskit to design 142 00:07:03.680 --> 00:07:07.199 the quantum circuits themselves. The algorithms. A big focus for 143 00:07:07.240 --> 00:07:10.680 them is minimizing the impact of noise, making things run 144 00:07:10.720 --> 00:07:12.560 better on today's imperfect hardware. 145 00:07:12.639 --> 00:07:15.519 Right noise. We'll definitely need to talk more about that 146 00:07:15.839 --> 00:07:17.839 and the third layer model developers. 147 00:07:17.959 --> 00:07:21.920 These are folks who might be say chemists or financial analysts. 148 00:07:22.199 --> 00:07:24.639 They want to use a quantum algorithm as part of 149 00:07:24.680 --> 00:07:29.120 a larger workflow. They use quisk it to integrate quantum 150 00:07:29.120 --> 00:07:32.879 computation into their existing applications, focusing on getting data in 151 00:07:32.959 --> 00:07:36.160 and results out, often without needing to know every single 152 00:07:36.199 --> 00:07:37.000 gate detail. 153 00:07:37.199 --> 00:07:39.240 That makes a lot of sense. Can you run quiske 154 00:07:39.319 --> 00:07:40.480 it on your own machine? 155 00:07:40.560 --> 00:07:43.560 You can. You set it up locally, usually with Anaconda 156 00:07:43.600 --> 00:07:46.519 for Python and Jupiter notebooks. It's great for simulating circuits, 157 00:07:46.800 --> 00:07:50.319 but crucially you need your IBM Quantum API token to 158 00:07:50.399 --> 00:07:53.879 actually send jobs to the real quantum computers in the cloud. 159 00:07:54.000 --> 00:07:55.279 Gotcha, And if you get stuck. 160 00:07:55.439 --> 00:07:58.800 There is a huge, very active community the quiskeuters on Slack. 161 00:07:58.839 --> 00:08:01.560 It's amazing. You can ask questions, get help, see what 162 00:08:01.639 --> 00:08:04.480 others are working on, connect with researchers. It's a really 163 00:08:04.480 --> 00:08:05.920 supportive environment. 164 00:08:05.680 --> 00:08:08.120 That's great to know. So we have quippits, we have 165 00:08:08.160 --> 00:08:11.319 platforms like IBM Quantum and tools like quisket. Now, how 166 00:08:11.360 --> 00:08:14.399 do we actually control the quibbuts? You mentioned gates in 167 00:08:14.439 --> 00:08:15.519 the composer. 168 00:08:15.240 --> 00:08:18.399 Right, Quantum logic gates. They are the fundamental operations you 169 00:08:18.439 --> 00:08:22.680 perform on quibbots. But unlike classical gates, which just flip 170 00:08:22.720 --> 00:08:26.879 bits based on simple rules, quantum gates are well, they're different. 171 00:08:26.959 --> 00:08:28.720 They perform unitary transformations. 172 00:08:28.800 --> 00:08:33.840 Unitary transformations sounds fancy. What does that mean? Practically? 173 00:08:34.000 --> 00:08:37.279 It essentially means there are rotations in the Quibbets abstract 174 00:08:37.320 --> 00:08:41.639 mathematical space. Think of the block sphere. A gate rotates 175 00:08:41.639 --> 00:08:44.440 the Quibbitts state vector to a different point on that sphere. 176 00:08:44.879 --> 00:08:49.159 And crucially, these transformations preserve the vector's length, which relates 177 00:08:49.200 --> 00:08:50.039 to probability. 178 00:08:50.440 --> 00:08:52.559 Okay, rotations, and you mentioned they're. 179 00:08:52.360 --> 00:08:56.200 Reversible, almost all of them. Yes, reversibility is a key property. 180 00:08:56.559 --> 00:08:59.080 If you apply a gate, you can almost always apply 181 00:08:59.159 --> 00:09:02.399 another related gate its complex conjugator inverse to get back 182 00:09:02.440 --> 00:09:06.240 exactly where you started. A HaTamar gate, for instance, applied twice, 183 00:09:06.360 --> 00:09:09.120 brings you back to the initial state, unlike some classical 184 00:09:09.159 --> 00:09:11.679 gates exactly think of a classical A and D gate. 185 00:09:12.039 --> 00:09:13.960 If the output is zero, you don't know if the 186 00:09:13.960 --> 00:09:18.120 inputs were zero, zero, zero, one, or ten. Information is lost. 187 00:09:18.399 --> 00:09:21.519 Quantum gates generally don't lose information like that. 188 00:09:21.960 --> 00:09:24.639 Okay, So what are some of the essential gates the 189 00:09:24.679 --> 00:09:25.519 building blocks. 190 00:09:25.919 --> 00:09:29.399 Well, for single clivits, the HaTamar gate or H gate 191 00:09:29.639 --> 00:09:32.919 is absolutely fundamental. It's the main way we create superposition. 192 00:09:33.360 --> 00:09:35.960 It takes a definite zero A state or one eight 193 00:09:36.000 --> 00:09:38.639 to state and puts it into an equal mix of both. 194 00:09:38.759 --> 00:09:40.559 The superposition workhorse. Got it. 195 00:09:40.639 --> 00:09:43.639 Then you have the polygates X, Y, and Z. These 196 00:09:43.679 --> 00:09:46.279 are basic rotations by one hundred and eighty degrees around 197 00:09:46.279 --> 00:09:49.159 the respective axes on the block sphere. The X gate 198 00:09:49.279 --> 00:09:51.960 is the quantum equivalent of the classical not T gate. 199 00:09:52.080 --> 00:09:54.720 It flips zero to one a's and vice versa. 200 00:09:54.799 --> 00:09:56.240 Simple flips what else. 201 00:09:56.440 --> 00:09:59.639 For more precise control, especially over the quibt phase, which 202 00:09:59.679 --> 00:10:03.159 is cre interference, you have phase gates, the ES gate, 203 00:10:03.320 --> 00:10:06.039 the T gate, and their inversus s dagger T tigger. 204 00:10:06.200 --> 00:10:09.360 They rotate the state around the z axis by specific. 205 00:10:08.840 --> 00:10:10.840 Amounts, fine tuning the phase exactly. 206 00:10:10.960 --> 00:10:12.960 And then there's the universal U gate. It's the most 207 00:10:13.039 --> 00:10:16.519 general single quibut gate. With just three angle parameters, it 208 00:10:16.519 --> 00:10:20.639 can perform any possible rotation, any single quibbit operation you 209 00:10:20.679 --> 00:10:21.320 can imagine. 210 00:10:21.320 --> 00:10:24.200 Wow, Okay, so that covers single quibts. But the real 211 00:10:24.279 --> 00:10:26.240 power comes from interactions, right. 212 00:10:26.200 --> 00:10:31.080 Entanglement precisely, yeah, and that requires multiquibit gates. The absolute 213 00:10:31.120 --> 00:10:34.240 cornerstone here is the controlled knot gate or. 214 00:10:34.279 --> 00:10:36.720 Cnot controlled not. What does it do? 215 00:10:36.960 --> 00:10:39.840 It acts on two quibbits, a control quibbit and a 216 00:10:39.879 --> 00:10:43.440 target quibt. If the control clibit is in the zero state, 217 00:10:43.519 --> 00:10:45.799 it does nothing to the target. But if the control 218 00:10:45.879 --> 00:10:48.240 quibit is in the one A state, it applies an 219 00:10:48.440 --> 00:10:51.399 X gate a not gate to the target quibt flipping it. 220 00:10:51.639 --> 00:10:54.440 Ah. So one quibt state controls an operation on another. 221 00:10:54.559 --> 00:10:57.759 Yes, and this is the primary way we create entanglement 222 00:10:57.799 --> 00:10:59.639 between equipbitts. It's fundamental. 223 00:10:59.720 --> 00:11:01.759 What are the multiquibit gates are important? 224 00:11:01.879 --> 00:11:04.679 Well, there's the Tafoli gate or CCX. It's like a 225 00:11:04.720 --> 00:11:08.480 cnot but with two control quibots. The target quibit only 226 00:11:08.480 --> 00:11:10.600 flips if both control quibots are one. 227 00:11:10.480 --> 00:11:12.080 A okay, double control. 228 00:11:12.240 --> 00:11:15.360 It's significant because the Toafoli gate is actually a universal 229 00:11:15.480 --> 00:11:18.960 classical gate. You can build any classical circuit using just 230 00:11:19.039 --> 00:11:22.519 to fullness in quantum. It shows how complex operations are 231 00:11:22.519 --> 00:11:28.399 built up, often requiring decomposition into many simpler gates like hadamards, cnots. 232 00:11:27.840 --> 00:11:29.840 And T gates interesting any others. 233 00:11:30.159 --> 00:11:32.840 The swapgate is sometimes useful. It just swaps the states 234 00:11:32.879 --> 00:11:34.960 of two quibuts. Does exactly what it says on the tin. 235 00:11:35.159 --> 00:11:39.000 Okay, So you said almost all gates are reversible. What's 236 00:11:39.039 --> 00:11:41.200 the exception where does the information get lost? 237 00:11:41.559 --> 00:11:45.320 The exception is measurement. When you measure a quiboit to 238 00:11:45.360 --> 00:11:47.559 find out if it's a zero or a one, the 239 00:11:47.679 --> 00:11:52.399 quantum state collapses. All that complex superposition information, the amplitudes, 240 00:11:52.440 --> 00:11:55.320 the phase. It's gone. You just get a classical bit 241 00:11:55.480 --> 00:11:58.360 zero or one based on the probabilities dictated by the 242 00:11:58.360 --> 00:12:02.360 state before measurement. It's the one irreversible step that makes sense. 243 00:12:02.399 --> 00:12:04.399 You have to look eventually, and looking. 244 00:12:04.320 --> 00:12:07.639 Changes things exactly. And you know, a really beautiful simple 245 00:12:07.679 --> 00:12:10.159 circuit that uses these basic gates is the one that 246 00:12:10.200 --> 00:12:11.200 creates the Bell state. 247 00:12:11.279 --> 00:12:11.879 Bell states. 248 00:12:11.919 --> 00:12:15.639 They're the four fundamental states of maximum entanglement. For two quibits. 249 00:12:16.000 --> 00:12:18.440 For example, the A plus A state is an equal 250 00:12:18.440 --> 00:12:21.960 superposition of zero, zero and eleven air. Both quibits are 251 00:12:22.039 --> 00:12:24.679 zero or both are one perfectly correlated. 252 00:12:24.759 --> 00:12:25.519 And how do you make that. 253 00:12:25.679 --> 00:12:28.480 It's surprisingly simple. You start with two quibits and the 254 00:12:28.559 --> 00:12:31.159 zero zero state, apply a Hadamar gate to the first 255 00:12:31.200 --> 00:12:34.279 quibut to put it in superposition. Then apply a Cnot 256 00:12:34.440 --> 00:12:36.440 gate with the first quibut as control and the second 257 00:12:36.519 --> 00:12:39.320 is target boom bell state maximally entangled. 258 00:12:39.600 --> 00:12:42.399 Wow, just two gates for that spooky action. 259 00:12:42.919 --> 00:12:47.840 Yep, it's foundational for so many quantum algorithms and protocols. 260 00:12:47.919 --> 00:12:50.440 Okay, so we've got the theory, the gates, the tools. 261 00:12:51.039 --> 00:12:53.960 But running this on actual hardware, I imagine it's not 262 00:12:54.080 --> 00:12:57.120 quite as clean as the simulations. You mentioned Noise earlier. 263 00:12:57.279 --> 00:13:00.639 Oh yeah, noise is the big, big, big challenge in 264 00:13:00.679 --> 00:13:04.480 the current era of quantum computing. These quantum states are 265 00:13:04.480 --> 00:13:08.159 incredibly fragile. But first, let's talk about how you run 266 00:13:08.240 --> 00:13:11.720 things efficiently on real hardware. That's where quisket run time 267 00:13:11.759 --> 00:13:12.159 comes in. 268 00:13:12.279 --> 00:13:13.639 Run time, what's that about. 269 00:13:13.799 --> 00:13:17.519 It's IBM service designed to execute quantum circuits more effectively 270 00:13:17.559 --> 00:13:20.679 on their cloud systems. It's about getting better performance and 271 00:13:20.720 --> 00:13:24.440 managing the whole process better than just sending individual isolated jobs. 272 00:13:24.519 --> 00:13:25.279 How does it do that? 273 00:13:25.679 --> 00:13:28.960 One key feature is sessions. Imagine you have an algorithm 274 00:13:28.960 --> 00:13:31.320 that needs to run a quantum circuit, then do some 275 00:13:31.399 --> 00:13:35.840 classical processing, then run another related circuit, maybe iteratively. 276 00:13:35.480 --> 00:13:37.120 Right, like machine learning or optimization. 277 00:13:37.519 --> 00:13:40.240 Exactly if you submitted each circuit as a separate job 278 00:13:40.440 --> 00:13:42.039 you keep going back to the end of the queue. 279 00:13:42.080 --> 00:13:44.679 For the quantum computer, it would take forever and the 280 00:13:44.720 --> 00:13:48.360 system's state might change between your runs. Sessions guarantee you 281 00:13:48.399 --> 00:13:51.600 dedicated access to a quantum processor for a block of 282 00:13:51.639 --> 00:13:56.679 related computations. It's huge for reducing weight times and ensuring consistency. 283 00:13:57.000 --> 00:14:00.240 That sounds essential for complex tasks really is. 284 00:14:00.360 --> 00:14:03.559 And within run time you interact using primitives. These are 285 00:14:03.600 --> 00:14:08.639 like standardized interfaces for common quantum tasks. They abstract away 286 00:14:08.720 --> 00:14:11.200 some of the hardware variation, like what kind of tasks. 287 00:14:11.240 --> 00:14:13.720 The main ones currently are the sampler and the estimator. 288 00:14:14.000 --> 00:14:17.000 The sampler primitive, for example, is designed for algorithms where 289 00:14:17.080 --> 00:14:20.879 you primarily care about the probability distribution of the measurement outcomes. 290 00:14:20.919 --> 00:14:23.240 You run the search up many times and it gives 291 00:14:23.279 --> 00:14:26.360 you back the counts or quasi probabilities for each result. 292 00:14:26.759 --> 00:14:29.559 Okay, so run time and primitives help manage the execution, 293 00:14:29.919 --> 00:14:31.919 But what about the noise itself? What are we actually 294 00:14:31.960 --> 00:14:32.679 fighting against? 295 00:14:32.919 --> 00:14:35.399 Right the noise? There are a few main culprits. The 296 00:14:35.440 --> 00:14:38.360 big one is decoherence. This is the general loss of 297 00:14:38.399 --> 00:14:40.559 that delicate quantumness. 298 00:14:40.000 --> 00:14:41.639 Losing the quantum magic. 299 00:14:41.360 --> 00:14:44.120 Pretty much It breaks down into two main types. There's 300 00:14:44.200 --> 00:14:47.080 T one time, or relaxation time. That's a long it 301 00:14:47.080 --> 00:14:49.519 takes for a quibit in an excited state like one, 302 00:14:49.600 --> 00:14:52.879 you to naturally decay back down to its ground state zero, 303 00:14:53.519 --> 00:14:55.360 like a hot cup of coffee. 304 00:14:54.919 --> 00:14:56.679 Cooling down, okay, losing energy. 305 00:14:56.879 --> 00:15:00.200 Then there's T two time, or dephasing time. This is 306 00:15:00.240 --> 00:15:02.919 about losing the phase coherence, the precise alignment of the 307 00:15:02.960 --> 00:15:06.799 quantum wave. It's caused by random fluctuations in the Quibbitts environment. 308 00:15:07.039 --> 00:15:09.679 Think of it like trying to keep a group of 309 00:15:09.720 --> 00:15:13.399 spinners perfectly synchronized. When there's random jittering, they drift apart. 310 00:15:13.840 --> 00:15:16.360 T two is usually much shorter than T one, making 311 00:15:16.360 --> 00:15:18.080 phase control really tricky. 312 00:15:18.320 --> 00:15:20.960 So energy loss and phase drift what else? 313 00:15:21.080 --> 00:15:24.840 You also have gate errors. The physical pulses used to 314 00:15:24.879 --> 00:15:28.440 implement quantum gits aren't perfectly precise, so rotation might be 315 00:15:28.440 --> 00:15:33.000 slightly off or introduce unwanted phase shifts and finally, readout errors. 316 00:15:33.279 --> 00:15:36.279 Even measuring the quibet's final state isn't perfect. Sometimes you 317 00:15:36.279 --> 00:15:38.080 measure a zero when it was actually a one, or 318 00:15:38.200 --> 00:15:38.759 vice versa. 319 00:15:38.840 --> 00:15:40.480 Well, it sounds like a mindfield. How do you even 320 00:15:40.519 --> 00:15:42.759 know how bad the noise is on a particular machine. 321 00:15:42.960 --> 00:15:45.960 Quisket actually lets you characterize the noise. You can run 322 00:15:46.120 --> 00:15:50.399 calibration experiments to measure T one T two gate errors, 323 00:15:50.480 --> 00:15:53.320 readout errors for a specific device, and then build a 324 00:15:53.320 --> 00:15:55.960 noise model based on that data. You can use this 325 00:15:56.120 --> 00:15:59.000 model in simulations to predict how noise will affect your 326 00:15:59.039 --> 00:16:01.120 algorithm before running it on the real thing. 327 00:16:01.360 --> 00:16:05.080 That's useful. So knowing the enemy, how do you fight it? 328 00:16:05.360 --> 00:16:09.600 Two main strategies, error suppression and aer mitigation suppression. First, 329 00:16:09.879 --> 00:16:13.000 air suppression techniques try to actively reduce the effect of 330 00:16:13.039 --> 00:16:18.000 noise during the computation itself. A key example is dynamical decoupling. 331 00:16:18.559 --> 00:16:20.799 During times when equip it is supposed to be idle 332 00:16:20.840 --> 00:16:25.320 waiting for other operations, it's vulnerable to dephasing. Dynamical decoupling 333 00:16:25.360 --> 00:16:29.600 inserts sequences of carefully timed pulses, often simple excates that 334 00:16:29.639 --> 00:16:32.960 effectively refocus the corbet's phase, canceling out some of the 335 00:16:33.000 --> 00:16:34.919 slow drift caused by noise. 336 00:16:34.840 --> 00:16:36.600 Like constantly nudging it back on track. 337 00:16:36.799 --> 00:16:39.000 Kind of yeah. It actively fights the diffusion of the 338 00:16:39.080 --> 00:16:39.759 quantum state. 339 00:16:39.879 --> 00:16:42.200 Okay, and error mitigation, how is that different? 340 00:16:42.399 --> 00:16:46.159 Aerror. Mitigation doesn't necessarily prevent the errors from happening. Instead, 341 00:16:46.320 --> 00:16:50.039 it assumes errors will happen, runs the computation anyway, often 342 00:16:50.120 --> 00:16:54.279 multiple times with variations, and then uses classical post processing 343 00:16:54.320 --> 00:16:58.320 techniques to estimate what the ideal noise free result would 344 00:16:58.320 --> 00:16:58.600 have been. 345 00:16:58.879 --> 00:17:01.519 AH, so you clean up the results afterwards using. 346 00:17:01.320 --> 00:17:06.400 Statistics essentially yes techniques like zero noise extrapolation ZNE, where 347 00:17:06.400 --> 00:17:09.160 you deliberately amplify the noise and extrapolate back to the 348 00:17:09.279 --> 00:17:14.039 zero noise point, or probabilistic error cancelation PECK, which learns 349 00:17:14.039 --> 00:17:17.240 the errors and tries to invert them statistically. Quis GET 350 00:17:17.279 --> 00:17:20.759 run Time incorporates these into different resilience levels you can choose. 351 00:17:21.200 --> 00:17:24.279 They use extra classical computation, but are crucial for getting 352 00:17:24.359 --> 00:17:28.160 useful answers from today's noisy intermediate scale quantum and ice 353 00:17:28.240 --> 00:17:30.200 Q devices and ice Q that's. 354 00:17:30.079 --> 00:17:34.039 The current era, right noisy, not quite fault tolerant yet exactly. 355 00:17:34.279 --> 00:17:36.640 We don't have enough quibits or low enough error rates 356 00:17:36.640 --> 00:17:39.799 for full quantum error correction yet, so suppression and mitigation 357 00:17:39.839 --> 00:17:40.880 are absolutely vital. 358 00:17:41.319 --> 00:17:45.119 This battle against noise really highlights the engineering challenge, but 359 00:17:45.240 --> 00:17:49.680 it's leading somewhere exciting the idea of quantum advantage right. 360 00:17:50.240 --> 00:17:52.839 Quantum advantage is the point where a quantum computer can 361 00:17:52.880 --> 00:17:56.720 solve a specific problem faster or more accurately than the 362 00:17:56.759 --> 00:18:00.400 best possible classical computer, regardless of how big or powerful 363 00:18:00.440 --> 00:18:01.559 that classical computer is. 364 00:18:01.680 --> 00:18:03.480 And are we seeing signs of that we are? 365 00:18:04.279 --> 00:18:07.920 IBM Quantum had a significant result in twenty twenty three 366 00:18:08.200 --> 00:18:11.119 using their one hundred and twenty seven quo bit Eagle processor. 367 00:18:11.519 --> 00:18:15.319 They demonstrated calculations related to material science that were complex 368 00:18:15.400 --> 00:18:18.599 enough to be beyond exact simulation by even the most 369 00:18:18.640 --> 00:18:22.839 powerful classical supercomputers using brute force methods. It was a 370 00:18:22.920 --> 00:18:27.119 major step showing potential for quantum utility even before fault tolerance. 371 00:18:27.279 --> 00:18:30.359 That's huge. So what kinds of algorithms actually deliver this 372 00:18:30.440 --> 00:18:31.880 speed up? Can we look at some example? 373 00:18:32.000 --> 00:18:35.440 Absolutely, some of the earliest foundational algorithms really showed the 374 00:18:35.440 --> 00:18:37.880 potential for speed up, even if the problems they solved 375 00:18:37.920 --> 00:18:39.319 were a bit academic. 376 00:18:39.640 --> 00:18:41.599 Like Deutsch's algorithm okay, wud that do? 377 00:18:41.799 --> 00:18:43.759 It solves a very simple problem. You have a function 378 00:18:43.799 --> 00:18:45.640 that takes one bit in and gives one bit out. 379 00:18:45.799 --> 00:18:48.640 Is the function constant always output zero or always outputs 380 00:18:48.680 --> 00:18:50.920 one or balanced output zero for one input one for 381 00:18:51.000 --> 00:18:51.240 the other. 382 00:18:51.680 --> 00:18:53.759 Classically, you'd have to test both inputs. 383 00:18:53.880 --> 00:18:58.160 Right, two checks exactly two queries to the function. Quantumly, 384 00:18:58.279 --> 00:19:01.920 Deutsch's algorithm solves it with just one query. It uses 385 00:19:01.960 --> 00:19:05.720 superposition and interference to check both possibilities at once. 386 00:19:05.680 --> 00:19:08.680 One query instead of two. Okay, maybe not world changing, 387 00:19:08.720 --> 00:19:09.799 but it proved the concept. 388 00:19:09.960 --> 00:19:13.279 It was revolutionary at the time. Then came Deutsch Jozza. 389 00:19:13.440 --> 00:19:17.440 It generalizes the idea to functions with ended inputs. Determining 390 00:19:17.480 --> 00:19:20.680 if that kind of function is constant or balanced Classically 391 00:19:21.160 --> 00:19:24.160 could take potentially many queries up to two to the 392 00:19:24.200 --> 00:19:26.400 power of n one plus one in the worst. 393 00:19:26.119 --> 00:19:27.319 Case, exponentially harder. 394 00:19:27.480 --> 00:19:30.799 Right, quantumly still just one query. 395 00:19:30.599 --> 00:19:32.279 One query. That's astonishing. 396 00:19:32.400 --> 00:19:35.799 It really demonstrates the power of quantum parallelism. And then 397 00:19:35.839 --> 00:19:39.880 there's Bernstein Vaserani. It finds a hidden secret string used 398 00:19:39.880 --> 00:19:42.680 in a specific type of function. Classically, you need end 399 00:19:42.720 --> 00:19:45.359 queries where n is the length of the string. Quantumly, 400 00:19:45.519 --> 00:19:46.680 again just one query. 401 00:19:46.799 --> 00:19:50.559