Hartmut Neven from Google Quantum AI explains how quantum computers work through superposition and demonstrates live hardware with over 100 qubits. Google has hit the first two milestones of a six-step plan to build error-corrected machines, with near-term applications in drug design, battery research, and molecular sensing.
Questions this video answers
Hartmut Neven; he leads Google Quantum AI.
Since 2012.
Over 100 qubits.
Six milestones total; the team has achieved the first two.
Their chip performed a computation the then‑fastest supercomputer would have needed 10,000 years to do (2019 result).
Chapters
I'm Hartmut. I lead Google Quantum AI. I've been working on quantum computing since 2012. And let me tell you why this is so intriguing. Today's computers, like your laptop or a server at a Google data center, operate on the binary logic of zeros and ones. A quantum computer like this one replaces the binary logic with the laws of quantum physics. That gives it more powerful operations, allowing it to perform certain computations with way fewer steps. So where does this superpower come from? Quantum computing is the first technology that takes the idea serious that we live in a multiverse. It can be seen as as farming out computations to parallel universes. Let me explain. In quantum physics, the key mathematical object to describe many worlds is called superposition. To understand what it is, let's look at this simple system. You just need three bits to describe it. Each coin is a two state system. Heads or tails, zero or one. We look at a start state. If I were to know which forces act on the system, then I can predict its trajectory and future states. This is how we reason in classical physics and also in everyday life. But if you were to treat this as a quantum system, then it can branch into many configurations simultaneously. And we have to keep track of all those trajectories, interfere them to make an accurate prediction of what states we are going to see in the future. So the equations of quantum mechanics tell us that at any time, any object, myself or the world at large exists in a superposition of many configurations. Intriguingly, look around in this room, we are forming a configuration too. And the equations of quantum physics would suggest that we sit in different arrangements in different worlds. This superpower can be applied to computation. Picture a search task by envisioning a very tall closet with a million drawers. I place an item in one of the drawers. How many drawers do you have to open to find the item? In average, it will be half a million. But if you had access to a quantum algorithm, it would only be 1000 steps to find the item. How in the world can this be? Indeed, it cannot be in a single word. So here you see good example how quantum computing can attain an advantage by performing computations in parallel worlds. Let me show you how to use a quantum computer in practice. So here you see a program in CIRC or Python based programming language to express quantum algorithms. It looks like sheet music. Each line represents a qubit and each representation box represents an operation. When I hit return, then it gets transmitted to our data center in Santa Barbara. Here you actually see a live feed of one of our Machines, Actually, our most powerful quantum computers now have over 100 qubits. There the operations get translated into waveforms, electrical pulses that control the qubits. You see how the waveforms change as I change the circuit. So this is a simple 2 qubit circuit performing quantum search. For the programmers among you, please note how I find one item in a database of four by only doing a single call to the database. This is something you could not do on an ordinary call computer. So what can you do with quantum computers today? We have prepared interesting quantum states and studied their properties. This has led to dozens of publications in high impact journals like Nature or Science. Actually, I like to think of it as creating little pieces of magic. For example, one state we prepared can be sort of as spawning a tiny traversable wormhole. We can use it to learn about the physics of wormholes. We can throw a qubit in and see how it reappears on the other side. We made time crystals. That's a cool word, isn't it? Like, who doesn't want to have a time crystal as an earring? Time crystals have amazing physical properties. They change periodically in time when without ever exchanging energy with the environment. They're the closest to a perpetual mobile that the laws of physics allow you to get. Or a final example, non abelian anyons. This is a mouthful, but these are systems that change the overall properties when exchanging two identical parts. Something humans have never seen before. Because envision a little house made of Lego bricks and envision swapping two bricks that look identical. In everyday life, you would not notice a difference. But quantum physicists had predicted that systems can exist that exchange or change their properties when you exchange two identical parts. To date, nobody has performed a practical application that can only be done on a quantum computer, despite what you may have read in the press. But today I'm excited to tell you that we are completing the design of an algorithm that may lead to first commercial applications. This quantum algorithm performs signal processing to enable new ways to detect and analyze molecules using using nuclear electronic spin spectroscopy. In times this may lead to exciting consumer applications. Envision a device akin to an electronic nose in your phone or a smartwatch. Wouldn't it be awesome if your phone could warn you that you step into a room with dangerous viruses? Or if your smartwatch could detect free radicals in your bloodstream and tell you it's time to drink your acai juice? Or warn you of allergens in food or many other truly helpful use cases to unlock more applications, we will need to build a large error Corrected quantum computer Here you see our roadmap. How to build a computer with a million physical qubits. It consists of six milestones and we achieved already the first two. Prior to 2019, nobody had shown a beyond classical computation on a quantum computer. We were the first to demonstrate it. Our chip could perform a computation that the then faster supercomputer would have needed 10,000 years to do. But recently we repeated this experiment and now frontier today's top supercomputer would need 1 billion years to perform this computation. This dramatic growth in compute power corroborates Nevin's law, which says that the power of quantum computers will grow at a double exponential rate. In 2023, we achieved a second milestone. We demonstrated again for the first time that quantum error correction is a scalable technology. Error correction sounds boring, but it's crucial. Today our two qubit operations have an error rate of one in a thousand. That means that in every thousand steps or so, the quantum computer will crash. To improve this, we combine many physical qubits to a logical qubit to reduce the error rate to one in a billion or even less. We are about halfway through our roadmap and we are optimistic. So we will complete it before the end of this decade. We have done analytical and numerical studies to predict which algorithms will be impactful on such a large quantum computer. A class of applications we like, and we call Feynman's killer app, is the simulation of systems where quantum effects are important. This is relevant for designing more effective, more targeted medicines. Specifically, we have worked with a pharmaceutical company on algorithms to describe cytochrome P450. This group of enzymes metabolizes about 75% of the drugs we take. Or the design of lighter, faster charging batteries that can hold a larger charge for electric cars or even electric airplanes. Or to hasten the design of fusion reactors to help with climate change. Arguably humanity's most urgent challenge. A recent result is a novel algorithm that delivers significant speed up for optimization is a big deal because optimization problems are ubiquitous in engineering, finance or machine learning. A way to think about this result is in the future when an AI will play chess or go against the quantum AI. The quantum AI will win. This result shows that quantum computers will become a must have capability to serve foundational computational tasks. I'm also very interested in the intersection of physics and neurobiology. Quantum information science may enable us to answer one of humanity's deepest what creates conscious experience? An attractive conjecture is that consciousness is how we experience the emergence of a single classical world. Out of the many the multiverse is composed of. With academic collaborators have started a program to experimentally test this conjecture using methods of quantum neurobiology. If our conjecture is correct, this would allow us to expand human consciousness in space, time and complexity. In conclusion, we are making steady progress towards building the world's first useful quantum computer and applying its enormous power to important challenges. A quantum computer will be a gift to future generations, giving them a new tool to solve problems that today are unsolvable. Thank you.
