from Hacker News

The Google Willow Thing

by Bootvis on 12/10/24, 4:34 PM with 458 comments

  • by nsxwolf on 12/10/24, 9:53 PM

    Man, reading this makes me feel so small. Being a "software engineer" consuming APIs and updating database rows seems laughably childish compared to whatever the hell it is I just read. I can't even imagine why I should bother trying to understand it. It's completely inaccessible. Only an elite few get to touch these machines.
  • by bradleyjg on 12/10/24, 11:44 PM

    The problem it solved would take a septillion years to do on a conventional computer but no one other than a quantum researcher cares about that problem.

    How about solving a problem someone that’s not a quantum researcher would care about. Give me traveling salesmen with n=10. Or factor a 10 digit number. Something.

    Until then quantum computers are in the same category as commercial fusion. Long on “breakthroughs”, zero on results.

    Look at cancer researchers for a nice contrast. The annual number of “breakthrough that could cure cancer!1!” announcements have dropped to near zero while steady, real progress is being made all the time.

  • by munchler on 12/10/24, 5:44 PM

    The argument in favor of the Everettian multiverse (“where else could the computation have happened, if it wasn’t being farmed out to parallel universes?”) seems illogical to me. Aren't these parallel universes running the same computation at the same time, and thus also "farming out" part of their computations to us? If so, it's a zero-sum game, so how could there be an overall performance gain for all the universes?
  • by bambax on 12/10/24, 10:35 PM

    > Having said that, the biggest caveat to the “10^25 years” result is one to which I fear Google drew insufficient attention. Namely, for the exact same reason why (as far as anyone knows) this quantum computation would take ~10^25 years for a classical computer to simulate, it would also take ~10^25 years for a classical computer to directly verify the quantum computer’s results!!

    I don't understand that part, can someone explain? There should be plenty of problems that take a long time to solve, but are trivial to verify? Like for example factoring extremely large numbers that are the product of a few very large primes? Maybe not on the order of 10^25 years, but still?

  • by joak on 12/10/24, 10:31 PM

    The hardware is progressing, we have a issue though: we don't have algorithms to run on quantum computers. Besides Shor's algorithm, useful to break RSA, we have nothing.

    Just vague ideas like: it could be useful for quantum simulations or optimisation or maybe ...

    If tomorrow we have a full running quantum computing what would we run on it? We are in a vacuum.

    The only hope is a breakthrough in quantum algorithms. Nothing in sight, not much progress on this side.

    Oh yes, Zapata Computing, the best funded company in quantum algorithms just went under this year.

  • by ChrisArchitect on 12/10/24, 5:43 PM

    Related:

    Willow, Our Quantum Chip

    https://news.ycombinator.com/item?id=42367649

  • by imranq on 12/10/24, 11:39 PM

    Summary: Its a real result, the cool part is more qubits seem to live longer rather than shorter, bad part the results are not explicitly verifiable, only through extrapolation
  • by r33b33 on 12/11/24, 12:02 PM

    Let's talk about things that actually matter - where to invest in post-quantum world?

    I'll keep this short.

    - Google’s Willow quantum chip significantly outpaces current supercomputers, solving tasks in minutes that would otherwise take billions of years.

    - Hypothesis: Accelerating advancements in tech and AI could lead to quantum supremacy arriving sooner than the 2030s, contrary to expert predictions.

    - Legacy banking systems, being centralized, could transition faster to post-quantum-safe encryption by freezing transfers, re-checking processes, and migrating to new protocols in a controlled manner.

    - Decentralized cryptocurrencies face bigger challenges:Hard forks are difficult to coordinate across a decentralized network.

    - Transitioning to quantum-safe algorithms could lead to longer transaction signatures and significantly higher fees, eroding trust in the system.

    - If quantum computers compromise current cryptography, tangible assets (e.g., real estate, stock indices) may retain more value compared to digital assets like crypto.

    Thoughts?

  • by machina_ex_deus on 12/10/24, 6:01 PM

    Before invoking parallel universes, how about comparing the system to nature's mind-boggling number of particles in the macroscopic world? A single gram contains 10^23=2^76 particles. Google's random circuit sampling experiment used only 67 qubits, Which is still order of magnitude below 76. I wonder why, the chip had 105 qubits and the error correction experiment used 101 qubits.

    Did Google's experiment encounter problems when trying to run RCS on the full 105 qubits device?

    Before saying that the computation invoked parallel universes, first I'd like to see that the computation couldn't be explained by the state being encoded classically by the state of the particles in the system.

  • by aeternum on 12/10/24, 5:43 PM

    >it would also take ~10^25 years for a classical computer to directly verify the quantum computer’s results!!

    This claim makes little sense. There are many problems that are much easier to verify than to solve. Why isn't that approach ever used to validate these quantum computing claims?

  • by urbandw311er on 12/11/24, 12:10 AM

    > No doubt people will ask > me what this means for > superconducting qubits > versus trapped-ion or > neutral-atom or > photonic qubits

    I laughed at this. If I understood more than literally 2 words of that, then yes - no doubt I would ask about that.

  • by devit on 12/11/24, 12:21 AM

    Where's the performance on common useful tasks?

    What's the largest number it can factor using Shor's algorithm? What's the largest hash it can compute a pre-image for using Grover's algorithm?

  • by dataflow on 12/10/24, 6:54 PM

    Dumb question: can someone explain the following?

    Imagine a ball falling on the ground.

    Simulating the O(10^23) atoms in each one with a classical computer would take (say) 10^23 times the amount of work of simulating a single atom. Depending on the level of detail, that could easily take, you know, many, many years...

    We don't call the ball a supercomputer or a quantum computer just because it's so much more efficient than a classical computer here.

    I presume that's because it can't do arbitrary computation this quickly, right?

    So in what way are these quantum computers different? Can they do arbitrary computations?

  • by JKCalhoun on 12/10/24, 5:48 PM

    What does quantum computing need to move forward? Will just throwing a lot of money at the thing allow it to scale? Or are there fundamental problems blocking it that require new physics or new material sciences?
  • by nuancebydefault on 12/11/24, 10:27 AM

    >> But for anyone who wonders why I’ve been obsessing for years about the need to design efficiently verifiable near-term quantum supremacy experiments: well, this is why! We’re now deeply into the unverifiable regime that I warned about.

    Can anybody explain me why it is hard to find a problem that can be solved only by a basic quantum computer within a short timespan and can be easily verified by a normal computer? I thought there are so many algo's out there for which one direction is fast and the reverse takes ages.

  • by de6u99er on 12/11/24, 4:56 AM

    IMHO, we're still a long way from anything truly useful. The problem Google used to demonstrate quantum supremacy feels more like a glorified random number generator. Even if quantum computers can generate results faster, processing and storing the data still takes a lot of time. It’s hard not to draw a parallel with claims about "instantaneous quantum communication," where entangled particles appear to defy the speed of light — it seems impressive at first, but the practical value remains unclear.
  • by LikeBeans on 12/11/24, 8:46 AM

    In simple terms, if I understand quantum computing, and please correct me if I'm wrong, the big benefit is parallel computing at a massive scale whereas classical computing is serial in nature. If yes likely both method are useful. But a very useful use case for quantum computing is AI training to create the models. Currently consumes a lot of GPUs but QC has nice chance to impact such a use case. Did I get it right?
  • by victor22 on 12/11/24, 12:27 PM

    Will I be attacked for thinking this is at least fishy? or are they just being ultra secretive.

    They never talk about what this computer is actually doing.

  • by r33b33 on 12/11/24, 10:04 AM

    Can someone just give it to me straight: should I sell my crypto positions and move to stock indices and real estate? Yes or no?

    No nuance, just yes or no.

  • by thrance on 12/11/24, 11:10 AM

    Still no real potential applications beyond factoring large integers (of dubious use) and doing some obscure quantum physics simulations.

    QC companies are selling quantum AI and quantum finance and other hand wavy stuff. Yet, I don't see any algorithms that has a proven advantage in these domains over classical ones running on clusters of GPUs.

  • by 01HNNWZ0MV43FF on 12/10/24, 5:44 PM

    > No doubt people will ask me what this means for superconducting qubits versus trapped-ion or neutral-atom or photonic qubits,

    I only wonder what it means for cryptography.

    The letters "crypt" don't appear in the text.

  • by daft_pink on 12/10/24, 5:53 PM

    I just want to know when this thing is going to put me out of work forever with it’s insane and crazy math skills.
  • by Jasondells on 12/11/24, 12:28 PM

    While these advancements in quantum error correction are undeniably impressive, I can't help but feel a sense of déjà vu here.... reminds me of the decades of "breakthroughs" in fusion energy—always promising to change the world, but perpetually just out of reach.

    The Google benchmark with random circuit sampling is fascinating from a theoretical perspective, but it’s hard to see how this translates into solving problems that matter outside the quantum research community.

    The lack of practical applications is a glaring issue. Sure, it's cool that Willow can outperform Frontier on this obscure task, but where’s the real-world impact?

    Cryptography, optimization, drug discovery—these are the kinds of problems quantum computing needs to beat if it’s going to justify the investment. Until that happens, it feels like we’re stuck in a cycle of overpromising and underdelivering, with flashy press releases but no tangible results.

    And let’s talk about scalability. Even if Willow hits the error-correction frontier, the number of physical qubits needed to build a truly practical quantum computer seems astronomical. Millions of qubits just to factor a number?

    It’s hard to see how this scales in a way that makes economic or scientific sense. Right now, it feels like quantum computing is a field for researchers who are okay with not seeing practical outcomes in their lifetimes.

    Maybe I’m too cynical, but this smells like another example of tech marketing getting ahead of the science. Maybe we can admit that we’re still decades away from quantum computing having any real-world relevance?

  • by flkenosad on 12/10/24, 10:36 PM

    Could bitcoin miners use this to "guess" the next block?
  • by mupuff1234 on 12/10/24, 5:35 PM

    I just want to know if the stock movement is justified or not.
  • by Sergii001 on 12/11/24, 8:03 AM

    It becomes more and more interesting
  • by partloyaldemon on 12/11/24, 2:42 AM

    "except now with the PR blitz from Sundar Pichai on down"

    I definitely read this first pass and thought 'damn the CEO is hitting its own quantum supercomputer with bug reports'. that's cold. It just came out.

  • by krick on 12/11/24, 1:03 AM

    Surprisingly sparse on actual information considering all this humble-bragging (actually, not even so humble) about how he was teaching peasants to catch fish for 20 years, to be forced yet again to hand it out with his own bare hands! Well, he didn't hand out much.

    The post reiterates some facts from the original statement, which are pretty vague for most, I believe. The only useful clarification is that simulation results are indeed unverifiable, lol (as some might have suspected, but still nice to have a definitive confirmation from somebody who is supposedly an expert on this).

    Then it addresses the cringeworthy "Everettian multiverse" statement discussion. Granted, it indeed was one of the most discussed things on the previous thread, but I have honestly assumed that it's so obviously moot that one can simply ignore it. Everyone knows that at least one of top-3 threads on HN comments must be either a lame joke or some sort of bikeshedding argument.

    And that's pretty much it. "This blogger said this usual generic words, that reporter asked for an interview, but I declined, also, kudos to Google team for amazing work, it's unclear if it's any good, but it surely isn't bad!" Uh, ok, thanks for the clarification, man.

    I get it that this post was written in a hurry, but given all that "fish-handing" stuff and all these commenters in this very thread complaining about how they don't know the difference between "trapped-ion or neutral-atom" qubits (as if this distinction was the very essence of the post, which author paid much more attention to than to his responses to NYT journalists) it just doesn't deliver.

    ...So, what did I expect? Well, I didn't expect anything, but let's state the obvious. Google's benchmark was to produce some very specific (and unverifiable) random distribution (which, BTW, he kinda says, but waaay less clearly than it could have been said). Obviously, nobody cares about that. Everyone cares about when they will be able to run Shor's algorithm on Google's Chip, and factor primes and deprecate RSA into oblivion. Obviously. Some may wonder why it's not possible to do it on that Willow thing, others may suspect that it may have something to do with the fact they need a ton of "physical" qubits to emulate logical qubits because of error-correction. Also, it is widely advertised, that the very thing that is special about Willow is vastly better (and somehow "more promising to scale") error-correction. So, what people really want from a generous and skillful fisherman is obviously some critical, appropriately-speculative, ELI5-style analysis of the updated state of the art. What does Willow have, what does it need to become practical, what are some realistic projections on when we can get there. Where is all of that? Where is the fucking fish?!

  • by blast on 12/10/24, 5:44 PM

    Damn he's funny.

    For 20 years I’ve been trying to teach the world how to fish in Hilbert space, but (sigh) I suppose I’ll just hand out some more fish.

  • by Ham121 on 12/11/24, 1:01 PM

    Google Willow is an exciting step forward in technology! The potential for it to transform how we interact with and integrate AI into our daily lives is incredible. I am curious to learn more about its real-world applications and how it will improve accessibility and efficiency across various domains. Kudos to the team for pushing boundaries and innovating looking forward to seeing whats next!
  • by megamix on 12/11/24, 7:59 AM

    This is how miracles work. They are just physical laws fast forward, I like how this explains the ancient miracles. So maybe our previous generations had access to these things and it was lost along the way.