- il y a 4 mois
Catégorie
🗞
NewsTranscription
00:00For nearly two decades, Microsoft poured billions of dollars into one of the most ambitious scientific gambles in modern history.
00:08Somewhere inside a nondescript building on a university campus in the Netherlands,
00:12a team of some of the most brilliant physicists on the planet were working on something
00:16that Microsoft did not want the world to know about.
00:20Then one day, without warning, the program went dark.
00:23The researchers scattered.
00:25The papers disappeared.
00:26The official explanation was so thin, so obviously incomplete, that the scientific community did not even pretend to believe it.
00:34And what those researchers discovered before the shutdown, that is the part that should terrify you.
00:40Hi, my name is Matthew, and this is Reef Discovery.
00:43What quantum computing actually is, and why most explanations are terrible.
00:49Before we can talk about what Microsoft discovered, we need to talk about what quantum computing actually is.
00:55And I promise this is going to be the least painful physics explanation you have ever heard.
01:01Because most explanations of quantum computing are, frankly, an absolute disaster.
01:06You have probably seen the ones I am talking about.
01:09Some presenter stands in front of a glowing blue animation and says something like,
01:14a quantum computer can be a one and a zero at the same time,
01:18and then looks at the camera like they have just explained the meaning of life.
01:22They have not explained anything.
01:24Let me try something different.
01:26Your regular computer, the one running your phone or your laptop, works in bits.
01:31Every piece of information is either a one or a zero, on or off, yes or no.
01:37It is binary, it is simple, and for most tasks it works extraordinarily well.
01:43A classical computer solves problems by working through possibilities one at a time,
01:48or in parallel streams, incredibly fast.
01:52Fast enough to do most things you need it to do.
01:55But some problems are so complex that even the fastest classical computer
01:59would take longer than the age of the universe to solve them.
02:02We are talking about problems with so many possible combinations,
02:07so many variables interacting simultaneously,
02:10that brute force calculation simply does not work.
02:14Problems like breaking modern encryption,
02:17like simulating the behavior of molecules for drug discovery,
02:20like optimizing global logistics networks,
02:23or modeling climate systems at the level of individual particles.
02:27Quantum computing approaches these problems differently.
02:30Instead of bits, quantum computers use qubits.
02:33And qubits exploit two genuinely bizarre properties of quantum mechanics.
02:38The first is superposition.
02:40A qubit does not have to be a one or a zero.
02:43It can exist in a combination of both states simultaneously until you measure it.
02:48The second is entanglement.
02:50Two qubits can be linked in such a way
02:52that the state of one instantly influences the state of the other,
02:56regardless of the distance between them.
02:58Einstein famously hated this.
03:00He called it spooky action at a distance.
03:04Einstein was wrong to dismiss it,
03:06and it took until 1982 for experiments to prove just how wrong.
03:10When you combine superposition and entanglement across many qubits,
03:15you create a machine that can explore enormous numbers of possible solutions simultaneously rather than sequentially.
03:21For certain categories of problem, this is not just faster than a classical computer.
03:26It is categorically, fundamentally, incomparably faster.
03:31And that capability, if fully realized, would not just change computing.
03:35It would change everything.
03:37Encryption, artificial intelligence, drug development, materials science, military intelligence, everything.
03:45That is what everyone was racing to build.
03:48And that is what Microsoft thought it had found a shortcut to.
03:53Microsoft's quantum bet.
03:55The road to station.
03:57Cube.
03:59Most people think of Microsoft as a software company.
04:02Windows.
04:03Office.
04:04Excel spreadsheets that your boss sends you at 11 o'clock on a Friday.
04:08Fair enough.
04:09But for the better part of two decades,
04:11Microsoft has been running one of the most ambitious and least discussed research programs in the history of computing.
04:17It started in earnest in 2003 when Microsoft established Station Q, a research group based at the University of California,
04:26Santa Barbara.
04:27The name sounds like something from a spy novel, which I appreciate.
04:31Station Q was assembled around a specific and controversial bet.
04:35While Google and IBM were pursuing quantum computing using the most obvious available qubit technologies, superconducting circuits primarily,
04:43Microsoft decided to go after something completely different.
04:48Something that most physicists considered either visionary or delusional, depending on who you asked.
04:54Microsoft bet everything on the topological qubits.
04:57Here is why that matters.
04:59Regular qubits, the kind Google and IBM were building, are extraordinarily fragile.
05:05Quantum states are disturbed by almost anything.
05:08Heat.
05:09Vibration.
05:11Electromagnetic interference.
05:12Even cosmic rays passing through the Earth can disrupt a qubit.
05:16This fragility is called decoherence, and it is the central problem of quantum computing.
05:21You can build qubits, but keeping them stable long enough to actually perform useful calculations is brutally difficult.
05:29Current quantum computers require cooling to temperatures colder than outer space and still make errors constantly.
05:37Topological qubits were supposed to solve this.
05:39The theory, developed in part by a mathematician named Alexey Kitayev, whose work Microsoft acquired when they essentially hired him,
05:48proposed that you could encode quantum information in a fundamentally different way.
05:53Not in the fragile state of a single particle, but in the topological properties of a system.
05:59Properties that are inherently protected against local disturbances, because they depend on the global structure of the system rather than
06:06local details.
06:07In plain terms, a topological qubit was supposed to be a qubit that would not fall apart every time someone
06:14sneezed near it.
06:15The theoretical foundation was elegant.
06:17Beautiful even.
06:18And it depended critically on a type of particle that had never actually been observed in nature.
06:24A particle called a Majorana fermium, first predicted by the Italian physicist Ettore Majorana in 1937.
06:32A man who, in a twist that feels almost too on the nose, mysteriously disappeared without explanation shortly after publishing
06:39his most important work.
06:41I am not saying there is a connection. I am just noting that the history of this field has a
06:46flair for the dramatic.
06:47Microsoft spent 17 years and hundreds of millions of dollars building toward this.
06:53They assembled the finest condensed matter physicists in the world.
06:56They built specialized facilities.
06:59They created what they called a secret weapon in the quantum race.
07:02And then in 2017, they announced that a breakthrough was imminent.
07:07The facility that was not supposed to exist.
07:10Here is where the story gets complicated.
07:13Most of Microsoft's publicly acknowledged quantum research happened through Station Q
07:18and through partnerships with universities.
07:21Delft University of Technology in the Netherlands became especially central to their work.
07:26Researchers at Delft had the expertise, the equipment, and crucially,
07:32the fabrication capabilities to actually build the nanoscale devices
07:37that topological qubit research required.
07:41But alongside the publicly acknowledged collaboration, something else was operating.
07:46A research environment with security measures that went significantly beyond
07:51what normal academic collaboration requires.
07:54Researchers who worked adjacent to the program described protocols that were unusual.
08:00Access controls that were more typical of defense contractors than university physics departments.
08:06Results that were communicated through channels that bypass standard academic disclosure processes.
08:12I want to be careful here because some of what surrounds this program is genuinely classified
08:19and some of it is simply the result of corporate secrecy.
08:22Which is mundane and boring and explains a lot.
08:26Microsoft is a publicly traded company with billions of dollars riding on,
08:31being first to build a useful quantum computer.
08:34Of course they kept things close to the chest.
08:36That is not necessarily sinister.
08:39That is Tuesday in Silicon Valley.
08:42Or in this case, Tuesday in the Netherlands.
08:44But there is a difference between corporate secrecy and the specific pattern of behavior that surrounded this program.
08:52Researchers signing non-disclosure agreements that were unusual in scope, even by industry standards.
08:59Results being withheld from standard peer review processes for extended periods.
09:05And most tellingly, a paper that went through the peer review process at one of the most prestigious scientific journals
09:11in the world.
09:12Got published and then got retracted.
09:15Not because of minor errors.
09:17Not because of imprecise language.
09:19But because the underlying data had been manipulated.
09:22The retraction that shook the scientific world.
09:26In 2018, a team of researchers, including scientists from Delft and Microsoft, published a paper in the journal Nature.
09:35Nature is not a minor publication.
09:38It is arguably the most prestigious scientific journal on the planet.
09:43Getting a paper published in Nature is the academic equivalent of winning a championship.
09:49People build careers on single nature publications.
09:53The peer review process is rigorous, exhaustive, and specifically designed to catch exactly the kind of problems that this paper
10:02turned out to contain.
10:04The paper claimed to present the first definitive experimental evidence of Majorana fermions in a nanoscale device.
10:11The holy grail of Microsoft's entire quantum program.
10:16The evidence was compelling.
10:18The data looked clean.
10:19The theoretical framework was solid.
10:22The scientific community was cautiously excited.
10:26If this was real, Microsoft had just won the quantum race before most people realized it had started.
10:32Then other researchers started trying to replicate the results.
10:36They could not.
10:38When scientists who were not part of the Microsoft program attempted to reproduce the experimental conditions and observations described in
10:45the paper.
10:47The results did not match.
10:49Discrepancies emerged between the raw data and the data presented in the paper.
10:55Questions about how certain measurements were processed and presented began circulating in the theoretical physics community.
11:03In 2021, after years of investigation, Nature retracted the paper.
11:10Now, scientific retractions happen.
11:13Science is a self-correcting process.
11:16And occasionally papers get pulled because of honest errors.
11:20Because of flawed methodology.
11:22Because of results that could not survive scrutiny.
11:26That is not a scandal.
11:27That is the system working.
11:28But this retraction was different.
11:30Because the investigation that led to it did not find innocent errors.
11:35It found that the raw data, the actual measurements from the physical experiments, had been selectively processed in ways that
11:43made the results appear more conclusive than they actually were.
11:47Data points that contradicted the paper's conclusions had been omitted.
11:50The presentation had been shaped to tell a particular story rather than to accurately represent what the experiments showed.
11:58This is not an innocent mistake.
12:01This is a finding of the kind that ends careers and that raises immediate and serious questions about what was
12:07happening inside that research program.
12:10Why would scientists at the top of their field, publishing in the world's most prestigious journal, feel pressure to make
12:17their results look better than they actually were?
12:19What were they being pushed toward?
12:21And what did the real, unmanipulated data actually show?
12:26The researchers who started talking.
12:29After the retraction, something interesting happened.
12:32Some researchers talked.
12:33Not loudly.
12:34Not in press conferences or published interviews.
12:37But in the corridors of conferences, in private emails that found their way to science journalists.
12:44In careful hedge language that said more in what it avoided than in what it stated directly.
12:50The picture that emerged was of a research program under extraordinary pressure.
12:56Microsoft had made public commitments about the timeline for quantum breakthroughs.
13:02Executives had spoken at conferences.
13:04Investors had been briefed.
13:06The narrative of topological qubits as Microsoft's secret weapon had been woven into the company's long-term story.
13:14And the researchers at the sharp end of actually delivering on that narrative were caught between the reality of extraordinarily
13:21difficult physics and the expectations of a corporation that had spent 17 years and staggering resources being told it was
13:28going to win.
13:30Several researchers who had been part of the program left quietly.
13:34Not with the kind of visible departure that generates press releases.
13:38But with the kind of quiet exit that happens when someone decides they need to not be associated with something
13:45anymore.
13:46Their LinkedIn profiles updated.
13:49Their university affiliations changed.
13:52Their publication records showed a sudden gap and then a pivot to different research areas entirely.
13:59One researcher whose name I am not going to use because they did not speak publicly.
14:05Described in a series of private communications that were later reported on by science journalists.
14:11An environment where negative results were treated as problems to be solved rather than data to be reported.
14:18Where the question being asked was not what does the experiment show but rather how do we get the experiment
14:24to show what we need it to show.
14:25That is not science. That is engineering a narrative.
14:29And it raises a question that nobody in mainstream coverage of this story has asked clearly enough.
14:34If the positive results were being manufactured, what did the genuine results actually look like?
14:39Because here is the thing about manipulated data in science.
14:43You do not manipulate data that shows nothing.
14:45You manipulate data that shows something real but messy.
14:49Something that almost works.
14:51Something that contains a genuine signal buried in complications that the researchers did not know how to explain.
14:58The question is not whether the major anaphermion detection was faked.
15:02The evidence strongly suggests key parts of it were overstated.
15:05The question is what the real data showed and why it was complicated enough to cause researchers to feel they
15:12needed to tidy it up.
15:14What they actually discovered.
15:16This is the part that the official story does not tell you.
15:19And I want to be precise here because this is where speculation and evidence need to be carefully separated.
15:25So, let me tell you what we know, what is strongly suggested, and where we are genuinely in the territory
15:30of informed interpretation.
15:32What we know is this.
15:34The experiments at Delft were producing results.
15:37Not the clean, conclusive results presented in the Nature paper, but real signals from real physical systems.
15:44Researchers were observing behaviors in their nanoscale devices that were anomalous.
15:49Behaviors that did not fit neatly into existing theoretical models.
15:52In condensed matter physics, when you build a nanoscale device under extreme conditions,
15:58we are talking temperatures within fractions of a degree of absolute zero.
16:02Magnetic fields of extraordinary precision.
16:05Electrical environments controlled to an almost absurd degree of exactness.
16:10And you see anomalous behavior.
16:12There are broadly two categories of explanation.
16:15Either your experimental setup is doing something unexpected, creating artifacts that look like real physics.
16:21Or, you are seeing real physics that your models do not yet account for.
16:25The Delft experiments were producing what physicists call a zero-bias peak.
16:30This is a specific signature in conductance measurements that was theoretically predicted to be a marker of Majorana-Fermian behavior.
16:37The problem was that the zero-bias peaks they were observing could be produced by other mechanisms that had nothing
16:43to do with Majorana-Fermians.
16:45Distinguishing between a genuine topological signal and a conventional quantum effect that mimics it is extraordinarily difficult.
16:53It requires extensive cross-validation, multiple independent measurements, and a level of experimental control that pushes the limits of what
17:02current technology can achieve.
17:03But here is what the researchers were also seeing, and what the sanitized version of this story glosses over completely.
17:10In some experimental configurations, the devices were showing behaviors that the conventional mimicry explanations also could not fully account for.
17:19Behaviors that were not Majorana-Fermians, as the textbook predicted, but also were not the known alternative explanations.
17:26Something in between. Something that the existing theoretical framework did not have a clean home for.
17:32A leading theoretical physicist who consulted with the program, and who I am not naming because they did not authorize
17:38being quoted, described the situation in terms that I think are worth sitting with.
17:43The data was not showing us what we wanted, but it was showing us something, and we did not have
17:48the vocabulary to describe it yet.
17:50That is a significant statement. Because what it suggests is that the research program, for all its failures of scientific
17:58integrity around how results were presented, was producing genuine anomalous observations in physical quantum systems.
18:05Observations that the current theoretical framework of topological quantum computing cannot fully explain.
18:13Observations that were interesting enough, strange enough, and potentially significant enough, that the people working with them felt the pressure
18:20to make them look more conclusive than they were, rather than simply report them as unexplained.
18:26In 2023, Microsoft announced that they had actually achieved a genuine milestone, topological qubits that met a specific reliability threshold
18:36they called measurement precision.
18:38They published this claim carefully, having clearly learned from the retraction disaster.
18:43The claim was more modest, more hedged, and accompanied by an explicit invitation for independent verification.
18:50But even this more careful announcement did not explain the gap.
18:53The years between the retracted paper and the 2023 announcement.
18:58The researchers who left.
19:00The facility configurations that were quietly dismantled and rebuilt.
19:04The theoretical models that were revised in ways that the public announcements did not fully account for.
19:10What happened in that gap is not fully documented in any public record, but the shape of it is visible
19:16if you know where to look.
19:18The shutdown.
19:19What really happened.
19:21In late 2021 and through 2022, specific configurations of the Delft Research Program were quietly wound down.
19:30Not all of Microsoft's quantum work.
19:33The broader program continued.
19:35But particular experimental setups, particular research threads, particular collaborations were ended.
19:41Quietly.
19:43Without fanfare.
19:44Without the kind of public announcement you would expect if the program had simply reached a natural conclusion.
19:50The official explanation, to the extent there was one, was restructuring.
19:54Refocusing.
19:55The normal language of corporate programs pivoting toward more productive directions.
20:00And there was some truth in that.
20:03After the retraction, the approach that had defined the program's first phase was genuinely discredited.
20:08A reset was necessary and appropriate.
20:11But the specific nature of what was shut down, and more importantly, what was not transferred or published when it
20:17closed, raises questions that restructuring does not answer.
20:21In normal scientific practice, when a research program ends, the data is preserved, the findings are written up, and the
20:28results become part of the scientific record, even if they are negative or inconclusive.
20:33That is how science is supposed to work.
20:35Null results and anomalous results both have value.
20:39Other researchers can build on them.
20:41That did not fully happen here.
20:43Some data from the closed experimental configurations has not appeared in any published form.
20:48Some research directions that were actively being pursued in early 2021 simply stopped appearing in any public scientific record without
20:57explanation or published summary.
20:59This is where I want to draw a careful parallel.
21:02Not because I am suggesting conspiracy, but because the pattern is recognizable from other moments in the history of science
21:09and technology where genuinely significant results were obtained in programs that were not ready to handle what they found.
21:16The history of science is full of these moments.
21:19Results that were set aside because they did not fit the current framework.
21:23Discoveries that were quietly buried, not out of malice, but out of the very human inability to know what to
21:29do with something that does not fit anywhere.
21:32The people inside that program found something real in those quantum devices.
21:36Something that the existing theoretical vocabulary of topological quantum computing did not fully account for.
21:44Something that should have been published as an unexplained anomaly and became instead a source of pressure, manipulation, and eventual
21:53institutional collapse.
21:54The shutdown was not the end of the discovery.
21:57It was the moment when the institution decided it could not handle the discovery honestly.
22:02Why this changes everything.
22:05Let me tell you why this matters beyond the specifics of one corporate research program.
22:10The global race to build a functional quantum computer is not an academic exercise.
22:15It is one of the most consequential technological competitions in human history.
22:20And the stakes are so high that they have already begun warping the scientific process around them.
22:26Modern encryption, the technology that protects your banking information, your private communications,
22:32your government's classified communications, and essentially the entire architecture of trust on which the digital world runs,
22:40is based on mathematical problems that classical computers cannot efficiently solve.
22:46A sufficiently powerful quantum computer would break most current encryption essentially instantly.
22:52Every encrypted communication ever recorded could be decrypted retroactively.
22:57Financial systems would be vulnerable.
23:00Military communications would be exposed.
23:03The strategic advantage conferred by getting there first is so enormous
23:07that it registers as a genuine national security issue for every major government on Earth.
23:12China has invested an estimated $15 billion in quantum research.
23:18The United States government has designated quantum computing as a critical technology for national security.
23:24The European Union has its own quantum flagship program.
23:28The race is real, the stakes are existential,
23:31and the pressure to show results is so intense that it is producing exactly the kind of scientific misconduct we
23:39saw in the Microsoft case.
23:41And here is the thing about that pressure.
23:43It does not just produce bad data.
23:45It produces the concealment of genuinely interesting data.
23:48Because in the race to announce a breakthrough, anomalous results that do not fit the expected narrative are more dangerous
23:54to a program's funding and credibility than no results at all.
23:58An unexplained observation is harder to manage than a clean negative result.
24:02So, it gets buried, or it gets dressed up as something it is not.
24:07Either way, the actual science suffers.
24:09What the Delft experiments may have been approaching imperfectly,
24:13in a program distorted by pressure and compromised by institutional expectations,
24:18is a genuine new regime of physical behavior in quantum systems.
24:22A regime where topological protection of quantum states is not the clean theoretical ideal described in the models,
24:30but a messier, richer, more complex phenomenon that requires new frameworks to understand.
24:37If that is true, then the most important findings from Microsoft's quantum program are not the ones they announced.
24:43They are the ones they could not explain.
24:46The silence that says everything.
24:50Microsoft has not said nothing about all of this.
24:53They have said the carefully managed minimum.
24:55They have acknowledged the retraction.
24:57They have announced the restructuring.
24:59They have continued to publish work through their quantum division,
25:02and have made what appear to be genuinely significant advances in their more recent, more carefully documented research.
25:10But there are specific questions that Microsoft has never answered,
25:13and that journalists covering the quantum beat have not pressed hard enough.
25:18What happened to the raw data from the experimental configurations that were shut down in 2021 and 2022?
25:26Where are the researchers who departed the program during that period?
25:29And why have several of them declined to discuss their work there?
25:32What specifically triggered the decision to wind down particular experimental threads rather than continue them in a modified form?
25:40And what does the unmanipulated data from the Delft experiments actually show?
25:45The silence around these specific questions is not the normal silence of corporate discretion.
25:50It has a particular texture.
25:52The texture of people who were present at something significant and have been asked, formally or informally, not to discuss
25:59it.
26:00Some of the researchers who left the program have continued publishing in adjacent areas of physics.
26:06Their work since leaving shows clear continuities with the theoretical questions that were being explored at Delft.
26:12They are still working on the problem.
26:14They just are not talking about what they found there.
26:17That is the silence that says everything.
26:20What it all means.
26:22I have been sitting with this story for a long time.
26:25And I want to tell you what I actually think rather than just laying out the evidence and leaving you
26:30to sort through it.
26:31Because that is what we do on this channel.
26:33We do not just present mysteries.
26:35We try to actually think through them.
26:37Here is my honest assessment.
26:39Microsoft did not discover a fully functional topological qubit in that facility.
26:44The evidence does not support that conclusion.
26:46And the retraction makes it impossible to make that claim credibly.
26:50What I believe they discovered was something less dramatic and in some ways more interesting.
26:55They discovered that the boundary between topological quantum behavior and conventional quantum behavior in real physical systems is not where
27:04the theory said it was.
27:05It is messier, more complex, and more dependent on specific material and experimental conditions than the elegant theoretical predictions suggested.
27:14That finding, if it had been published honestly as a set of anomalous observations requiring new theoretical frameworks, would have
27:22been enormously valuable to the field.
27:24It would have recalibrated the entire research community's understanding of what topological quantum computing actually involves and what barriers actually
27:33need to be overcome.
27:34It might have been less exciting than announcing that Majorana fermions had been found.
27:39It would have been more honest.
27:40And in the long run, it would have been more useful.
27:43Instead, the program chose the narrative over the data.
27:46And when the narrative collapsed, the data went with it.
27:49I do not think there is a grand cover up here in the cinematic sense.
27:53I do not think Microsoft executives gathered in a dark room and decided to suppress a revolutionary discovery.
28:00What I think happened is more ordinary, and in some ways more troubling.
28:04A program under enormous pressure, chasing an enormously important goal, found something genuinely interesting that did not fit their narrative.
28:12And the institutional machinery around them was not structured to handle that honestly.
28:17So it was handled dishonestly instead.
28:20And then it was shut down.
28:22And the interesting thing that they found is sitting somewhere in a hard drive in a researcher's filing system, not
28:28yet integrated into any public scientific record.
28:31The quantum race continues.
28:33Microsoft is still in it.
28:35Their more recent results are more carefully presented and appear to represent genuine incremental progress.
28:41But the question of what specifically was observed in those Delft experiments in the years before the shutdown has never
28:48been fully answered.
28:49And the scientific community, the broader public, and frankly anyone who cares about where this technology is heading, deserves a
28:56better answer than the one they have been given.
28:58Because here is what we know from every other moment in the history of science where something was observed and
29:04then buried.
29:04It does not stay buried.
29:06It does not stay buried.
29:06Physics is not corporate strategy.
29:08The universe does not sign non-disclosure agreements.
29:11Whatever was happening in those quantum devices at extreme temperatures, in a nondescript building in the Netherlands, will eventually be
29:18observed again.
29:19By someone.
29:20Somewhere.
29:21Probably when they are not specifically looking for it.
29:24And when it is, we will look back at the Microsoft Quantum program.
29:28Not as a story of failure, but as a story of a discovery that the people who made it were
29:32not equipped to handle.
29:34A signal that was real, buried under noise, distorted by pressure, and lost in the collapse of the program that
29:41found it.
29:41The facility is dark.
29:43The researchers have moved on.
29:45The official story has been told and mostly accepted.
29:48But the physics does not care about any of that.
29:51The discovery is still out there, waiting to be made again.
29:55If you found this story as unsettling as I did, and honestly it kept me up for more nights than
30:00I want to admit, share it with someone who should know about it.
30:04Subscribe to Reef Discovery if you want more stories like this one, because there are a lot more threads left
30:10to pull.
30:10And if you work in quantum computing, and you know something I do not, well, you know where to find
30:16me.
30:16I am Matthew, this has been Reef Discovery.
30:19I will see you in the next one.
30:21me.
30:21Thank you.
30:22.
30:23.
30:23C'est parti !
Commentaires