YouTube2Text

Majorana 2 Explained: The road to 1000x improvement — Transcript

by Microsoft · 1,379 words · 275 segments · language en · Watch on YouTube

Full transcript

  1. 0:00If you think about the way
  2. 0:01we process information,
  3. 0:03it fundamentally hasn't
  4. 0:04changed since the abacus.
  5. 0:062,400 years of calculating by
  6. 0:08moving ones and zeros around.
  7. 0:10Quantum computing changes that.
  8. 0:13We've worked
  9. 0:14on topological quantum
  10. 0:15computing for 20 years.
  11. 0:17And we've known that
  12. 0:19it can work, in theory.
  13. 0:20There was a period in which
  14. 0:22we were
  15. 0:22completely lost.
  16. 0:23We didn't know which way was up.
  17. 0:25We did get to a point
  18. 0:27when things just
  19. 0:28started to click into place,
  20. 0:29from knowing what you want to do
  21. 0:31to actually starting to build it.
  22. 0:33When what was only theory is now real.
  23. 0:37We went from Majorana 1 and almost
  24. 0:39completely reinvented it in Majorana 2.
  25. 0:42We've seen a thousand x improvement
  26. 0:43in the stability of our quantum computer.
  27. 0:53The last year has been quite a wild ride
  28. 0:56as we've gone from Majorana 1 to Majorana 2.
  29. 0:59We changed the material stack.
  30. 1:00We changed the design.
  31. 1:02We changed
  32. 1:02the way we were doing the measurements.
  33. 1:04Somehow, they all came together
  34. 1:06at the same time and all landed.
  35. 1:09The superconductor in our device was
  36. 1:11aluminum previously,
  37. 1:12and the semiconductor was
  38. 1:13indium arsenide.
  39. 1:14We changed the
  40. 1:15material stack significantly.
  41. 1:16We replaced aluminum with lead,
  42. 1:18and we introduced antimony
  43. 1:20into the semiconductor.
  44. 1:21That gave us a new material with
  45. 1:24properties that were
  46. 1:24even better and led to this
  47. 1:261000x improvement in that lifetime.
  48. 1:28And one big thing was the breakthroughs
  49. 1:30we made on the qubits,
  50. 1:33the bigger gaps,
  51. 1:34the longer lifetimes,
  52. 1:36thanks to lead.
  53. 1:37When you do an
  54. 1:37operation every microsecond
  55. 1:39and your qubit lives
  56. 1:40for about a millisecond,
  57. 1:42then it means there's one error every
  58. 1:45thousand operations.
  59. 1:47When it is
  60. 1:48a lifetime of tens of seconds
  61. 1:50and one operation takes a microsecond,
  62. 1:52that means you can do tens
  63. 1:54of millions of operations
  64. 1:55before an error.
  65. 1:56And now we're seeing it
  66. 1:57on 20-second timescale.
  67. 1:58And sometimes it's
  68. 1:58like minutes in between flips.
  69. 2:00And it takes the quantum world and puts
  70. 2:02it in units of like minutes, you know.
  71. 2:05It's kind of mind-blowing to
  72. 2:06have it in like a human scale.
  73. 2:08The larger that scale is,
  74. 2:09the more robust your qubit,
  75. 2:11the better its performance.
  76. 2:12And that dependence,
  77. 2:14that functional form, is exponential.
  78. 2:16It wasn't that we made a thousand x
  79. 2:18improvement in the underlying properties.
  80. 2:20It was
  81. 2:20improvements by a
  82. 2:21multiple of a few, two to three.
  83. 2:23But those multiples of on the order of
  84. 2:24two to three led to a thousand-fold
  85. 2:27improvement in this
  86. 2:28measure of performance.
  87. 2:31We've worked on
  88. 2:32topological quantum computing
  89. 2:34for 20 years.
  90. 2:35But for the last 20 years,
  91. 2:37that has been theory.
  92. 2:39What we see now is that
  93. 2:40lead was a wow moment.
  94. 2:42Lead was when we really saw
  95. 2:44what we theorized actually works.
  96. 2:48The quantum computer is
  97. 2:49much more than the qubits,
  98. 2:50just like a classical
  99. 2:51computer is more than
  100. 2:53the transistors.
  101. 2:54We have to control the qubits.
  102. 2:56We have to be able
  103. 2:57to measure them.
  104. 2:58We have to orchestrate them,
  105. 3:00and we have to be able
  106. 3:01to run our programs on them.
  107. 3:03So, we know how to build
  108. 3:05programs on quantum computers.
  109. 3:06We know how to put together this sequence
  110. 3:08of quantum operations,
  111. 3:10often called gates,
  112. 3:11to build out the circuits that
  113. 3:13implement an algorithm or
  114. 3:14allow us to run an application.
  115. 3:16We also have some fantastic tools
  116. 3:19in the form of
  117. 3:20the Quantum Development Kit
  118. 3:21and the Quantum Resource Estimator.
  119. 3:24So, we created the
  120. 3:25Quantum Resource Estimator
  121. 3:26to allow people to
  122. 3:27explore the trade-offs between
  123. 3:30a particular application
  124. 3:31or circuit that you need to run,
  125. 3:32the algorithms that you might
  126. 3:34use to implement that application,
  127. 3:36and the architectural considerations
  128. 3:38about the system it's going to run on.
  129. 3:41The quantum computer works on the principle
  130. 3:42of quantum mechanics.
  131. 3:44Quantum mechanics is the theory that
  132. 3:46describes how nature works at
  133. 3:48the smallest scales.
  134. 3:50And if I want to calculate the properties
  135. 3:52of a material, of a molecule, if I want
  136. 3:54to understand the chemical reaction,
  137. 3:56then I need to understand
  138. 3:58the behavior of the electrons
  139. 4:00in this molecule.
  140. 4:02And doing that exactly on a classical
  141. 4:03computer is exponentially hard.
  142. 4:06AI has helped us to increase
  143. 4:07the scale and speed
  144. 4:09with which we discover new materials
  145. 4:11or explore the details
  146. 4:13of chemical reactions.
  147. 4:14Our AI models right now,
  148. 4:16they're based on classical data,
  149. 4:18but when the system gets too big,
  150. 4:19they have to start making approximations.
  151. 4:22I put approximate data in,
  152. 4:23I get approximate answers out.
  153. 4:25Quantum changes all that.
  154. 4:27Quantum allows us to put the exact,
  155. 4:29ground-truth answer into the AI model
  156. 4:32and get first-time-right predictions
  157. 4:34out the other side.
  158. 4:35But those calculations will still take
  159. 4:38potentially hours or days.
  160. 4:40What we just want to do is we want to
  161. 4:43train an AI model to predict the results
  162. 4:45of the calculations.
  163. 4:47Then it will just take seconds and not
  164. 4:49hours to get the results.
  165. 4:52What that will enable is that your Copilot
  166. 4:54can not just write an essay,
  167. 4:58create an image, create a movie,
  168. 5:00but it can also then
  169. 5:02create materials and molecules
  170. 5:03because it understands quantum physics.
  171. 5:07That will totally change
  172. 5:09the way we make things.
  173. 5:12So, there are questions in chemistry,
  174. 5:14material science, physics
  175. 5:16that we just can't answer because
  176. 5:18we don't have the computing power.
  177. 5:20No classical computer on the planet
  178. 5:22can solve it.
  179. 5:23If we can unlock the answers,
  180. 5:25we will have a new path to discovering
  181. 5:27brand new materials.
  182. 5:28For example,
  183. 5:29superconductors that operate at
  184. 5:31room temperature
  185. 5:32or enable levitating
  186. 5:34magnetic transportation.
  187. 5:35Quantum computers are going to help us
  188. 5:37answer questions we just
  189. 5:38couldn't answer before.
  190. 5:40Skepticism is a valid
  191. 5:41and important part of science.
  192. 5:42I'm glad people are questioning and
  193. 5:44looking seriously at the work, right?
  194. 5:45I don't expect everyone to just, like, take
  195. 5:47everything at face value.
  196. 5:48You know, there's many things that are
  197. 5:50proprietary, that we can't share
  198. 5:51all the details.
  199. 5:52There have been people that have
  200. 5:53read our papers, looked at our stuff.
  201. 5:55Especially when they do it
  202. 5:56in good faith, where they're really
  203. 5:57just trying to get to the truth,
  204. 5:58that's always a good thing.
  205. 6:00Everyone who's looked
  206. 6:01very carefully
  207. 6:02and seriously at our results
  208. 6:03has seen that they are not only solid
  209. 6:06but really actually quite revolutionary.
  210. 6:08And, you know,
  211. 6:09if people are still skeptical,
  212. 6:10all I can say is, you know, there's more
  213. 6:12where that came from.
  214. 6:13And
  215. 6:13maybe we haven't knocked their socks off
  216. 6:14just yet. Pretty sure the next one will.
  217. 6:17I think we've just learned
  218. 6:19a lot along this journey.
  219. 6:20We were six to 12 months early,
  220. 6:21I think, in the material change we made.
  221. 6:24We've incorporated the
  222. 6:25use of AI-based tools.
  223. 6:26We've tightened the collaboration across
  224. 6:28various areas of the team.
  225. 6:29And very importantly,
  226. 6:31on the algorithms and application side,
  227. 6:32we've brought in some of those
  228. 6:34key applications that
  229. 6:35have enabled us to
  230. 6:37bring in those timelines.
  231. 6:38Right now
  232. 6:38we're targeting 2029
  233. 6:40for quantum computers that can
  234. 6:44solve some problems
  235. 6:45that haven't been
  236. 6:46solved before by other means.
  237. 6:47In the future, we'll build Majorana 3, 4,
  238. 6:49and 5, and the
  239. 6:50challenge now is scaling up.
  240. 6:53Just tremendous to think that we're a few
  241. 6:55years away from being able to start to
  242. 6:57use the types of scaled, fault-tolerant
  243. 7:00quantum computers that will unlock a
  244. 7:02whole new generation of applications in
  245. 7:04chemistry, material science, and beyond.
  246. 7:08I've been a part of this program since it
  247. 7:10was not even quite a program yet.
  248. 7:12Yeah, since the very beginning.
  249. 7:13I think I'm like employee two of
  250. 7:14the quantum program, basically.
  251. 7:17I've been blessed to work with some
  252. 7:18incredibly dedicated and brilliant
  253. 7:20people, which I find honestly humbling.
  254. 7:23We have been able to
  255. 7:25make devices and uncover properties
  256. 7:27that I wouldn't have
  257. 7:28dreamed of 20 years ago.
  258. 7:30And we're doing computation with it now.
  259. 7:32And that, to me, it's just blowing me away.
  260. 7:35So, at my core, I'm a technology nerd.
  261. 7:37I love working with new technologies.
  262. 7:40It's amazing to be part of an effort
  263. 7:42that's building a new form
  264. 7:44of quantum computing.
  265. 7:46There is
  266. 7:47nothing more transformative
  267. 7:48than quantum computing.
  268. 7:50It got to the point in my career
  269. 7:52when I said I need a quantum computer to
  270. 7:54do the research that I want to do.
  271. 7:57And that's why I joined Microsoft,
  272. 7:58to actually make it real
  273. 7:59and accelerate the path
  274. 8:01to getting it.
  275. 8:02There's much to discover.

About this transcript

This page contains the full transcript of Majorana 2 Explained: The road to 1000x improvement by Microsoft, generated from the public captions YouTube serves with the video. The transcript has 1,379 words across 275 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.