Danny and Derek welcome back Kevin Klyman, researcher at Harvard University’s Belfer Center, for a discussion of the technology arms race between the US/EU and China. They discuss China’s role as the leading hi-tech manufacturer in the world, de-risking and export controls, the CHIPS and Science Act, the dynamic with Taiwan and its semiconductor industry, the advent of AI and quantum computing, and more.
Thank you so much for this discussion, I found it very helpful. I couldn’t find enough discussions or interviews with Kevin on other platforms so I hope you will do a few more segments on this topic of Semiconductors or technology in general with him. Many thanks
Great episode as usual, but a pretty poor and inaccurate description of what quantum computers are, how they work, and what they can (and can't) do. If you ever want to talk to a physicist, you can give me a call ;)
Can you briefly say how the description of what they can and can't do was inaccurate? Is the Wikipedia opening paragraph better? It says QC leverages wave-particle duality, also mentions physical sims as use.
Sure, though some has been said in other comments now. The first thing as mentioned elsewhere is how they work: it's not important that you can put more information per bit, it's that the mechanism for how it arrives to the answer is different. Any kind of computer you can think of as a physical system (like a Rube Goldberg machine) that has been designed and set up such that its final configuration encodes the answer to the problem you want to solve. "Classical" computers have a dictionary of moves (pieces of the Rube Goldberg machine) they can make, while quantum computers have a different dictionary of moves, some of which would take a very long time to emulate with a classical computer. The different dictionary of moves means that they can arrive at answers to certain problems provably much faster than classical computers can. A simple example of a "quantum" algorithm is using scattering light to compute a Fourier transform, which on a classical computer takes an amount of time proportional to n log(n) for a vector of size n, but if the problem is encoded in a physical object and you add "send light through the object and measure the scattered intensity" into your set of moves, you can do it in linear time.
The second main issue is that "certain problems" are actually *very* specific certain problems. Quantum computers are *not* simply faster, nor will they allow the fast computation of arbitrary things that are hard for classical computers to solve. For instance, there is no known quantum algorithm for solving generic NP-hard problems like the traveling salesman, and a quantum computer should be as bad or worse as a classical computer at problems like these unless an algorithm is discovered. Shor's algorithm, a quantum algorithm for factoring large integers and that (if implemented on a large and stable quantum computer) would allow modern internet security based on public key encryption to be bypassed, is a very specific and fragile example: similar hard combinatorics problems don't have known fast quantum algorithms! Many news articles will mention as an aside that quantum computers would revolutionize drug development by making the simulation of molecules more efficient (this idea comes from statements by Feynman and corresponds with what you mention about "physical sims"), but recently a bunch of important people in the field have pointed out that it isn't certain they will actually outperform classical methods even for this! (see https://arxiv.org/abs/2208.02199) A lot of work has gone into "quantum cryptography" as well, which use properties of the way that quantum systems evolve to "guarantee" that no one can decrypt or snoop on your message. This is all plausible, but it is a lot of attention spent on a sort of non-problem: very few attacks on modern "classical cryptography" involve intercepting and decrypting messages in transit. Instead, they usually take the form of attacks on the device that displays the decrypted message (think Pegasus, which doesn't break the Signal protocol but can read your screen). In short, the "use case" proposition for quantum computers isn't super clear, beyond inconveniencing people who design web security for a while.
The description of quantum computing was _painfully_ inaccurate. It's not "just a different way of building a computer." It's a completely different _kind_ of computation, making use of superposition to arrive at answers with far less work than is required in conventional computing; if/when we can build sufficiently large quantum computers, they will solve problems that couldn't be solved by a conventional computer before the heat-death of the universe (and not just breaking encryption).
Qubits are not synonymous with atoms; in fact, atoms aren't even listed among the many physical implementations on wikipedia: https://en.wikipedia.org/wiki/Qubit
When I hear something so completely incorrect about a subject I know well enough to spot it, I have to wonder how much nonsense I've absorbed from that source on subjects I don't know enough about to fact-check without considerable effort.
I really liked this episode. People talk a lot about the influence of huge corporations on international relations (or the other way around) and I think its really interesting to hear about that the dynamics of that for a specific market.
Thank you so much for this discussion, I found it very helpful. I couldn’t find enough discussions or interviews with Kevin on other platforms so I hope you will do a few more segments on this topic of Semiconductors or technology in general with him. Many thanks
I agree with Mali, this was a super fascinating discussion and I found Kevins points and answers super easy to follow and clear
Great episode as usual, but a pretty poor and inaccurate description of what quantum computers are, how they work, and what they can (and can't) do. If you ever want to talk to a physicist, you can give me a call ;)
Can you briefly say how the description of what they can and can't do was inaccurate? Is the Wikipedia opening paragraph better? It says QC leverages wave-particle duality, also mentions physical sims as use.
Sure, though some has been said in other comments now. The first thing as mentioned elsewhere is how they work: it's not important that you can put more information per bit, it's that the mechanism for how it arrives to the answer is different. Any kind of computer you can think of as a physical system (like a Rube Goldberg machine) that has been designed and set up such that its final configuration encodes the answer to the problem you want to solve. "Classical" computers have a dictionary of moves (pieces of the Rube Goldberg machine) they can make, while quantum computers have a different dictionary of moves, some of which would take a very long time to emulate with a classical computer. The different dictionary of moves means that they can arrive at answers to certain problems provably much faster than classical computers can. A simple example of a "quantum" algorithm is using scattering light to compute a Fourier transform, which on a classical computer takes an amount of time proportional to n log(n) for a vector of size n, but if the problem is encoded in a physical object and you add "send light through the object and measure the scattered intensity" into your set of moves, you can do it in linear time.
The second main issue is that "certain problems" are actually *very* specific certain problems. Quantum computers are *not* simply faster, nor will they allow the fast computation of arbitrary things that are hard for classical computers to solve. For instance, there is no known quantum algorithm for solving generic NP-hard problems like the traveling salesman, and a quantum computer should be as bad or worse as a classical computer at problems like these unless an algorithm is discovered. Shor's algorithm, a quantum algorithm for factoring large integers and that (if implemented on a large and stable quantum computer) would allow modern internet security based on public key encryption to be bypassed, is a very specific and fragile example: similar hard combinatorics problems don't have known fast quantum algorithms! Many news articles will mention as an aside that quantum computers would revolutionize drug development by making the simulation of molecules more efficient (this idea comes from statements by Feynman and corresponds with what you mention about "physical sims"), but recently a bunch of important people in the field have pointed out that it isn't certain they will actually outperform classical methods even for this! (see https://arxiv.org/abs/2208.02199) A lot of work has gone into "quantum cryptography" as well, which use properties of the way that quantum systems evolve to "guarantee" that no one can decrypt or snoop on your message. This is all plausible, but it is a lot of attention spent on a sort of non-problem: very few attacks on modern "classical cryptography" involve intercepting and decrypting messages in transit. Instead, they usually take the form of attacks on the device that displays the decrypted message (think Pegasus, which doesn't break the Signal protocol but can read your screen). In short, the "use case" proposition for quantum computers isn't super clear, beyond inconveniencing people who design web security for a while.
The description of quantum computing was _painfully_ inaccurate. It's not "just a different way of building a computer." It's a completely different _kind_ of computation, making use of superposition to arrive at answers with far less work than is required in conventional computing; if/when we can build sufficiently large quantum computers, they will solve problems that couldn't be solved by a conventional computer before the heat-death of the universe (and not just breaking encryption).
Qubits are not synonymous with atoms; in fact, atoms aren't even listed among the many physical implementations on wikipedia: https://en.wikipedia.org/wiki/Qubit
When I hear something so completely incorrect about a subject I know well enough to spot it, I have to wonder how much nonsense I've absorbed from that source on subjects I don't know enough about to fact-check without considerable effort.
I really liked this episode. People talk a lot about the influence of huge corporations on international relations (or the other way around) and I think its really interesting to hear about that the dynamics of that for a specific market.