AI Summary
The video discusses the current state of quantum computing, highlighting a mismatch between industry optimism and actual progress. It reviews recent claims of quantum advantage, historical predictions that failed to materialize, and offers a critical perspective on the commercial viability of quantum computers in the near term.
Chapters
Quantum computing is highly anticipated, but progress has not met industry expectations. Despite this, optimism remains high, creating a mismatch that the video aims to explain.
A new claim of quantum advantage was made, but the calculation involved producing a random distribution of values with no known practical use, similar to previous demonstrations.
Philosopher Amit Hagar notes that within 18 months, all quantum advantage demonstrations have been reproduced on conventional computers, except for one random sampling case.
Hagar argues that demonstrating a hardware can run a circuit is just 'competent diagnostic engineering,' and the press has conflated this with computational advantage, which has been productive for funding but not for actual progress.
Five years ago, many expected small, noisy quantum computers (NISQ) to have commercial use. Predictions from Google (2017), IBM (2018), IONQ (2021), and Kipu Quantum (2023) all failed to materialize.
Boston Consulting Group predicted $2-5 billion in value by 2024, but actual value to end users was nil. The only profitable quantum application so far is forecasting profitable quantum applications.
The author strongly doubts NISQ computers will have commercial uses, citing John Preskill, who coined NISQ but never claimed commercial value. The author attributes the hype to academia leaking into industry without correction.
The author clarifies they are not saying quantum computers will never be useful, but it will take a long time, likely beyond this decade, and uses will be niche, not a billion-dollar market.
Demis Hassabis argues that quantum processes in nature likely have regularities that can be modeled with AI on conventional computers, a point the author thinks scientists aren't taking seriously enough.
The video concludes that quantum computing's commercial promise has been overhyped, with no practical applications yet. The author remains skeptical of near-term breakthroughs, suggesting that AI might model quantum phenomena better than quantum computers themselves.
Mentioned in this Video
💡 Key Takeaways
Quantum Advantage Claims Reproduced
Provides concrete evidence that quantum advantage claims have been consistently debunked, undermining hype.
00:58Hagar's Critique of Quantum Advantage
Offers a sharp academic perspective that challenges the validity of quantum advantage as a meaningful milestone.
01:24Only Profitable Quantum Application
Highlights the irony that the only profitable quantum application is forecasting profitable applications, illustrating the hype cycle.
03:12AI vs. Quantum for Modeling Nature
Presents a contrarian view that AI might outperform quantum computers in modeling quantum phenomena, a key debate.
04:58Full Transcript
[00:00] Quantum computing has been one of the most, if not the most anticipated new technology on the horizon, but things haven't been going as many in the industry thought they would. Nevertheless,
[00:14] they're more optimistic today than ever. What explains this weird mismatch? Today, I have an explanation for this and also clarification on my opinion about quantum computing. Just last week,
[00:26] we saw another claim of quantum advantage. This time from IBM and researchers at the University of Chicago. Quantum advantage means that a quantum computer did a calculation that would have taken
[00:40] unfeasibly long on a conventional computer. It doesn't necessarily mean that the calculation had any particular use. Indeed, in this new example, as well as in many previous ones, the calculation is to produce a random distribution of values. It's got no known use. There have been other
[00:58] previous claims that quantum advantage has been achieved. But as the philosopher Amit Hagar just nicely summarized in a new paper, within 18 months of any of these demonstrations, they've all been
[01:11] reproduced on a conventional computer or at least been shown to be reproducible. The one exception he names was one of those random sampling cases. Hagar is blunt with his judgment. He writes,
[01:24] "A demonstration that a hardware can run a circuit and that the circuit produces an observable matching unknown answer is a demonstration of competent diagnostic engineering. The press cycle of the past several years has conflated such results with computational advantage and
[01:41] the conflation has been productive of funding without being productive of the meta fact under investigation." What Hagar is referring to is that 5 years ago, many thought that quantum
[01:55] computers would already have a commercial use even when they're small and noisy. This has been referred to as Noisy Intermediate Scale Quantum computing, NISQ for short. For example,
[02:07] in 2017, researchers from Google claimed that we contend that short-term returns are possible with the small devices that will emerge within the next 5 years. Didn't happen. In 2018,
[02:20] researchers from IBM said that in the near term, NISQ computers have the best potential to deliver business advantage. One of them also said that IBM sees the next few years as the dawn of the
[02:32] commercial quantum era. Didn't happen. The quantum computing firm IONQ said in its 2021 annual report during the NISQ computing era, we expect quantum machine learning to be the first solution
[02:44] to transition into broadly available applications. Didn't happen. Enrique Solano, one of the founders of the German startup Kipu Quantum said that we work to reach a NISQ advantage for industry
[02:57] use cases within 18 to 36 months. that was on August 11th, 2023, almost exactly 36 months ago, didn't happen. These claims from researchers then spilled over into marketing analysis. The
[03:12] Boston Consulting Group, for example, predicted in 2019 that we anticipate value for end users in these sectors to reach a relatively modest 2 to 5 billion by 2024. The value to end users turned
[03:25] out to be even more modest, namely nil. To date, the only profitable quantum application has been forecasting profitable quantum applications. In my mind, the idea that NISQ
[03:37] computers will have any commercial uses was obviously [ __ ] And I mean 11 out of 10 on the [ __ ] meter. If you need evidence of this other than my word, I'll pretend to be deeply insulted, but refer
[03:51] you to John Preskill, one of the world's experts on quantum computing, who for all I know termed the phrase NISQ, and who famously did not claim these devices would have any commercial use. Yes, sometimes the relevant bit is what they don't say. So, what I think happened is that academia [ __ ]
[04:08] leaked into industry and consulting firms and no one was there to clean up the mess. No one except possibly Scott Aaronson who's been saying all along that quantum computers are super interesting from
[04:20] a research perspective. But commercial advantage, well, let's not talk about it. Let me be clear. I'm not saying that quantum computers will never have a commercial use. I just think it'll take
[04:32] a long time to get there. Certainly not in this decade. And the uses will be extremely niche and limited. Not a billion-dollar market. I've noticed that people call me pessimistic when it comes to
[04:46] quantum computers. I think I'm not. It's just that even my most optimistic self thinks it's idiotic that quantum computers will bring a new industrial revolution as some headlines put it. At least
[04:58] confidence has reached for tolerance. I also want to remind you that Demis Hassabis, who I've put down as one of the most intelligent people on the planet, has argued that whatever quantum processes happen in
[05:11] nature, will likely have so many regularities, you can model them with AI on a conventional computer. I think scientists aren't taking him seriously enough. If my day had any more hours,
[05:24] I try to formalize this into a formal conjecture, but the way things are going, AI will probably beat me to it. The last time I wrote code, it was Fortran. This is why I decided I need to
[05:38] urgently upgrade my coding skills. If you're in a similar situation, I have a recommendation for you. Boot.dev, who've been sponsoring this video. Boot.dev is focused on back-end development. So,
[05:50] Python, SQL, Go, the kind of stuff that actually runs the world behind the scenes. And you don't just watch lessons, you build real projects, you write code directly in the browser, you solve challenges, and you level up like it's a slightly nerdy role-playing game. They
[06:06] also have what they call the training grounds. That's basically infinite practice problem. I've been meaning for a while to set up a website for all my content, but I want to know what's going on with the data and what vulnerabilities to watch out for. And Boot.dev has been really
[06:23] eye opening. You can create an account for free, read and watch all the content and try a demo of the interactive features. If you decide to go for the full experience, use my code Sabine
[06:36] to get 25% off your first year. So go and check this out. Thanks for watching. See you tomorrow.