- 2 weeks ago
In the world of quantum physics, space and time seem almost irrelevant. Could we soon build unimaginably fast computers? Plus: How can we detect deepfakes better? And just how smart are robotic dogs?
Category
🤖
TechTranscript
00:06We've all been there, impatiently waiting as a computer struggles to keep up.
00:11And AI applications have taken demands for computing power to a whole new level.
00:17Last year, Europe's fastest supercomputer went online in Germany.
00:21It can perform more than a quintillion calculations every second.
00:26Yet that's not even leading the pack.
00:28China's latest model is thought to be twice as quick,
00:32and future computers could soon grow faster than ever before
00:35as they tap into the power of the quantum world.
00:42How that works and what it means for the security of your data,
00:46coming up on the show.
00:47Welcome to Tomorrow Today.
00:52In his lab at Oxford University,
00:54David Nadlinger actually just wanted to show us where the quantum computer sits
01:00among all the lasers and optical setups.
01:03But now the whole system has been thrown out of whack.
01:08Precision is in the micrometer range, and something has shifted slightly.
01:13The optical fiber has a diameter of just a few micrometers,
01:16and we try to map this atom exactly onto the fiber.
01:20So you're trying to aim?
01:23Exactly.
01:23But this is a single atom, very little light.
01:27The image forms somewhere here in the middle.
01:30But we have almost no way to make it brighter, to strengthen the signal.
01:38The basement labs at the elite university may look pretty cluttered,
01:43but they could be the birthplace of a key breakthrough in quantum computing performance,
01:47led by David Nadlinger and his team.
01:52In the first part of the quantum era, our understanding was used to build lasers and transistors,
01:58all this modern technology.
02:00Now we're using those tools to manipulate individual atoms,
02:04to get them to store information.
02:06Is the atom in one state or another?
02:08Is it a zero or a one, or both at the same time?
02:13Both at once. That's the special part, when everything is possible at the same time.
02:22To solve a maze, for example, a quantum computer could calculate all the possible paths at once,
02:29rather than trying one after another like a conventional computer.
02:33And that makes a huge difference, with every additional bit of information,
02:38here the atoms, computing power doubles.
02:42It grows exponentially, and that's the promise of quantum physics,
02:46to provide access to this computational power.
02:52Exponential? Here's what that means.
02:55Two quantum bits, or qubits, give four times the computing power.
03:00Ten qubits mean 1,024 times the power.
03:04And the calculations quickly head into the unfathomable.
03:10300 classical bits, we reached that milestone 50 years ago.
03:16But 300 quantum bits would correspond to around 10 to the 90th possible states.
03:25That's more than the number of atoms in the universe.
03:30To break today's standard encryptions, estimates suggest a quantum computer would need around a million cubits.
03:41About a million, because you need to correct for errors.
03:45You may have a million cubits, but that only gives you a few thousand perfect ones.
03:51As an experimental physicist, I'm always dealing with errors.
03:55But is there any doubt we'll reach that?
03:57It's an engineering challenge.
04:00From a physics standpoint, no one doubts that it's possible.
04:06What does that mean for bank data, medical records, or intelligence services?
04:10How can information be securely transmitted in the future?
04:14Andrew Shields' answer?
04:16With quantum physics.
04:18His team has built a system not based on complex mathematics, but on the strange rules of the quantum world.
04:26We can use new techniques which are not susceptible to being broken by a quantum computer.
04:33So quantum communication gives us one way to do that.
04:38The thing is that in the world of quantum physics, simply observing particles changes them.
04:44That principle is now being used in practice.
04:48In London, a quantum secure network is being tested for the first time.
04:53In it, companies exchange secure keys via quantum encryption, though range and data rates remain limited.
05:02Andrew demonstrates the system.
05:05Instead of sending simple pulses of light through optical fibers, they carry individual photons that are set to a particular
05:13quantum state.
05:14If an eavesdropper tries to eavesdrop on the encoded single photons en route, that act actually changes their encoding.
05:23And the change in the encoding can be sensed by the system as errors in the shared bit sequence.
05:30When an error rate rises, the receiver detects eavesdropping and stops the transmission.
05:39Observation changes the information.
05:42That aspect is unique to the quantum world.
05:46It's as though a book were not really a book, independent of us and unchanging,
05:51but instead was altered by the act of reading it.
05:58Quantum encryption is coming, but so are quantum computers.
06:02The race is on.
06:04Back in Oxford, the system is up and running again.
06:07It could become the turbo motor at the heart of a future, large-scale quantum computer.
06:14Experiments with lasers and mirrors bundle light and send it to the machine's core.
06:21That's a simple microchip holding a single strontium atom, which reaches a state of excitement when struck by the laser.
06:30The atom emits single photons, and we couple them in a standard optical fiber.
06:38That fiber can then be connected to another quantum computer module.
06:45Two simple one-qubit systems connected by a strand of optical fiber.
06:52The birth of a system with quadruple, two times two, the computing power.
07:01But it only becomes truly useful when we have one atom here and another there storing my quantum information,
07:09and they're entangled.
07:12Entanglement.
07:13It's perhaps the strangest phenomenon in quantum physics,
07:17one that can be illustrated with socks.
07:22Imagine it like this.
07:25Two people represent the atoms in our quantum computers.
07:29Their socks are the light photons for exchanging information.
07:34The color of those socks represents a quantum property.
07:38Their polarization.
07:41Who is wearing which color doesn't matter.
07:44At least until at the quantum level, David and Andrew become entangled.
07:49That means the two are no longer independent, but form a single system.
07:54The quantum property, the color of the socks, is now inseparably linked.
07:59And now, as crazy as it sounds, David and Andrew can move as far apart as they want, yet remain
08:05entangled.
08:06But as soon as an atom is measured, the sock colors are determined again.
08:11At exactly the same moment, and always in opposition to one another.
08:15So even from the other side of the world, Andrew would know that if his sock is green, then David's
08:21is red.
08:26Does this really happen?
08:28David has written a program to prove that his two quantum computers really are entangled.
08:37If we press the button now, we should be able to generate entanglement again between the two modules.
08:44The monitors display images of the atoms in the two quantum computers.
08:50Completely independently of one another, the researchers measure their polarization, the color of the socks, so to speak.
09:01It goes fast. We create entanglement, measure, and then do that over and over.
09:07It looks completely random. But when we compare the two sides, the results match up perfectly.
09:17And that's what happens. The results show that independently of one another, the two systems produce identical outcomes.
09:26Measuring one determines the result of the other.
09:33000011. So far, no errors.
09:38Moments when even as a physicist, you're still amazed.
09:42A first small network. But in a few years, many quantum computers could be connected, unlocking enormous computing power,
09:52and sending current encryption methods into obsolescence.
10:02Did you get all that? If not, don't worry. You're not alone.
10:07Even Albert Einstein found quantum physics strange and unsettling.
10:11Because it just doesn't seem to make sense.
10:14We'll get back to quantum computers in a little while.
10:17But first, let's talk about cats.
10:20Or rather, the millions of images of them found online.
10:23In an age of AI, can you tell which cats are real and which are fake?
10:28A new technology could help us spot the difference.
10:35This is a real photo of a real cat.
10:39And this is an image created by AI.
10:42Without human help, how can we tell which is real and which is fake?
10:49That question is being explored at the Swiss Federal Institute of Technology in Basel.
10:55There, researchers have developed a chip that records a picture's authenticity at the moment it's taken.
11:02Felix Franke worked on the project.
11:06The chip is a sensor. It could, for instance, be part of a camera sensor that captures light.
11:11And directly within the sensor, a watermark is generated that later guarantees we'll be able to verify the image is
11:17real, not artificial.
11:20Here's how the technology works.
11:23Every time the camera takes a photo, a special chip inside the sensor creates an invisible stamp.
11:30A kind of certificate of authenticity.
11:34With this, you can determine whether an image truly comes from, say, a smartphone camera or another physical sensor,
11:40and not from an artificial source.
11:43The chip is still a prototype, but the technology could eventually be built into any kind of sensor or camera.
11:52Social networks like Instagram could then automatically check whether content is genuine,
11:58and label manipulated images accordingly.
12:03What needs to happen is that society and perhaps policymakers need to exert pressure to create incentives for companies to
12:10implement this technology,
12:11because, ultimately, it will have to be integrated into camera sensor chips to become truly usable.
12:18At Zurich-based Pixel Vision, deepfake detection is already in use.
12:27The company develops identity verification systems for customers like banks or telecoms firms.
12:34Their systems have to be able to distinguish real images from fake ones,
12:39because fraud has grown common in identity checks involving portrait photos.
12:46Here are some deepfakes.
12:49These people don't actually exist.
12:52They're 100% artificially generated images.
12:56Some look unnatural, but others are already so good that it's hard to tell whether or not they're real.
13:04Together with the Swiss-based IDIAP Research Institute, Pixel Vision developed an AI that can detect manipulated images.
13:16Our software can identify specific patterns that occurred during the generation of artificial images.
13:24We've trained our machine learning models to identify these deepfakes.
13:32Even though AI can generate highly realistic images, it leaves behind telltale traces.
13:41We detect patterns created during the generation process.
13:46For example, certain pixel arrangements or extremely subtle color transitions that the human eye can't see, but machines can.
14:00The software is trained to recognize different types of manipulation.
14:05A common one is the face swap, where one visage is replaced with another.
14:11It can also identify completely synthetic faces, people who don't exist at all, as well as real faces that have
14:18been altered.
14:19The software can detect all of that.
14:24It works for now, but because criminals are constantly developing new ways to create deepfakes, the software has to be
14:31continuously retrained.
14:33That's the only way to ensure it'll be able to keep identifying manipulated images in the future.
14:43It sounds a little crazy.
14:46We keep developing AI systems to detect the increasingly convincing fakes created by other AI systems.
14:53But those systems can do so much more.
14:56And in the future, in combination with robotics, AI could save lives.
15:01In fact, AI-powered robot dogs are already helping us identify dangers before they become threats.
15:13Robots are growing more and more agile.
15:16In today's workplaces, especially in dangerous environments, they're increasingly able to support humans by taking on tasks that require more
15:25independence.
15:26Researcher Soren Pirck tracks development in the world of robots.
15:33Robotics has been a field of research for a long time.
15:36What we're seeing now is that the hardware is slowly becoming robust and versatile enough to be widely deployed.
15:43That's a breakthrough.
15:44The hardware is finally able to perform many different tasks.
15:49Not far from Kiel on Germany's Baltic coast, this robotic dog will soon be on patrol.
15:56Testing convinced a local utility company that robots like these can improve security.
16:02It moves with open ears and eyes through the country's most modern gas-fired power plant.
16:09However, the robot dog is not meant to replace humans entirely.
16:14There are tasks that are less attractive due to shift work, and also tasks in sensitive environments where we deal
16:21with chemicals that can be dangerous.
16:23That's where we see the potential to combine the work of machines, robots, and humans.
16:32The robot dog can search for puddles in places where there shouldn't be any, and also carry a range of
16:39devices to distant destinations.
16:42Even long staircases don't keep it from bringing its sensors exactly where they're needed.
16:48And with rapid advances in software, robots like these are also increasingly savvy.
16:58Many people know ChatGPT as a large language model, which means it can evaluate and generate language.
17:06But what's exciting here is what we're now seeing in robotics, the development of vision-language models, and even vision
17:16-language action models.
17:17These systems can interpret images, understand scenarios, and define any actions for a robot accordingly.
17:27Like real dogs that can sense danger at an early stage, the robot dog in Kiel's coastal power plant can
17:36detect weaknesses in technical processes, long before a component fails.
17:44For example, using acoustic measurements, we detected an air pressure leak that the human ear couldn't hear.
17:52It shows how modern sensors can far surpass human senses.
17:59Engineers were so impressed that they decided to invest about 200,000 euros to give the robot a permanent home
18:07in Kiel.
18:08For now, it just raises alerts. When something goes wrong, it still doesn't act autonomously.
18:17Researchers at Kiel University are pushing the envelope more.
18:21They want to deploy robots autonomously in tasks like firefighting to reduce potential risk to human life.
18:31In areas like wildfire response, robotics is still in its infancy.
18:37We don't yet have robotic systems that can operate consistently in forest environments.
18:43And when it comes to getting AI to evaluate wildfire situations, there's a further difficulty, which is that we still
18:50lack sufficient data.
18:55So this test robot is not yet ready to cope with a forest that's on fire.
19:01Despite its agility, without the right software, it couldn't effectively combat flames in complex scenarios.
19:08And speaking of combat, robots are also being trained for that.
19:13Development has been accelerated by the war in Ukraine.
19:19This isn't purely military technology. It depends on how operators use it.
19:26On the one hand, it can carry small loads of ammunition.
19:31But it can also assist in clearing mines or scanning terrain for hidden explosives.
19:39The question remains, who gets to choose when such technology saves lives and when it takes them?
19:48Societies need to be debating these questions.
19:51The responsibility can't be placed on the shoulders of just one actor.
19:56It includes developers, hardware manufacturers, users and society as a whole.
20:04In the end, these remain machines that simply carry out the tasks that humans programmed them to do.
20:16Now back to quantum computers.
20:18With their extraordinary power and characteristics, they could one day break encryption methods that are currently considered unbreakable.
20:27What can we do to protect ourselves from this kind of quantum digital attack?
20:33Here more than anywhere else, money is king.
20:37Canary Wharf in London's Financial District.
20:40An estimated $3.8 trillion in foreign exchange trades are processed here every day.
20:47More than a third of the total volume globally.
20:53The rapid development of quantum computing has shaken the sector.
20:58It's one of the biggest risks we've ever seen in terms of threat to cryptography.
21:02Europe's largest bank has 220,000 employees and rakes in $25 billion in profit annually.
21:13A lot of the systems that protect customers' data and the payments and that keep it secure
21:20and is the bedrock of security in the financial services system, that is at risk from emerging quantum computing capability.
21:28A lot of money could be stolen?
21:30A lot of money can be stolen.
21:32A lot of disruption could be had.
21:34It will undermine, you know, the very foundation of trust in the financial services system.
21:40And if a quantum computer is used to break cryptography, it could result in, you know,
21:45a catastrophic loss of confidence, not just in banks, but in the financial system as a whole.
21:50Hackers in countries like China and Russia attack data networks worldwide.
21:55At the same time, the Chinese government invests more than any other in quantum computing.
22:01Secure encryption has become a geopolitical necessity.
22:04Banks are top targets for cyber criminals.
22:07And, you know, other areas that are going to be under particular kind of threat is national security
22:14and government type of data, health and energy and infrastructure.
22:19So it's a pan-industry problem.
22:22Germany's Bundesbank, for example, assumes more than 5,000 cyber attacks occur per minute.
22:29Firewalls, antivirus programs, and autonomous defense software can help.
22:34But you can't fully monitor the entire network all the time.
22:38Bank data, digital signatures, medical information, genetic data.
22:43Without secure networks, modern life harbors big risks.
22:47The ability to encrypt data securely is vital.
22:51Current encryption methods are based on math problems that are easy to solve in one direction,
22:57but practically impossible in the other.
22:59Take what's called prime factorization.
23:03With small numbers, it's simple.
23:0515 is the product of 3 times 5.
23:1077 is the product of 7 times 11.
23:16But the larger the product, the more difficult it is to find the prime factors that produced it.
23:22So already you can see that it's quite easy to go from these two numbers,
23:28multiply these two numbers and get the large number.
23:31But really quite difficult to go back in reverse.
23:35And now I have one more integer which I'd like to show you in golf.
23:41And maybe you can factorize this into its prime factors for me.
23:46What do you think?
23:47Would you like to hazard a guess?
23:49How many digits are there?
23:50So this is actually a 637-digital decimal number, I should say,
23:58which corresponds to 2,048 binary digits.
24:03And this is the basis of the RSA that we use today.
24:07And in fact, it's been calculated that it would take the world's largest supercomputer,
24:13El Capitan, something like 100 million years to find the prime factors of this number.
24:20If we had a quantum computer, on the other hand, with 4,000 logical qubits,
24:25it could solve the same problem in minutes or in hours.
24:31That's the threat quantum computers pose to current security systems.
24:36Change is coming.
24:41So when will it happen?
24:44Quantum computers are still mostly in labs.
24:47But at the University of Oxford, ongoing research could significantly accelerate their development.
24:57Quantum computers represent a completely new way of computing, one that we know is far more powerful than conventional systems.
25:07The next big milestone in the coming years will be solving a problem that produces a useful answer to something
25:14we didn't know before.
25:16Like if they're using one, for instance, to simulate a molecule or superconductor.
25:21Breaking encryption could still take about 10 years, I think.
25:25That's what I'd say if I were working at an embassy and triggered a diplomatic incident.
25:32And if you were sending an encrypted diplomatic message involving spies?
25:37If I were in that business, I'd be very concerned.
25:41No question, quantum computing has the power to disrupt.
25:50That's all for now from the fascinating world of quantum physics and technology.
25:55But make sure to join us again next time on Tomorrow Today.
25:59See you then.
26:10See you next time.
26:11See you then.
26:11See you then.
Comments