Just says in P
Recently a Twitter account started called justsaysinmice. The only thing this account does, is to repost breathless news articles about medical research breakthroughs that fail to mention that the effect in question was only observed in mice, and then add the words “IN MICE” to them. Simple concept, but it already seems to be changing the conversation about science reporting.
It occurred to me that we could do something analogous for quantum computing. While my own deep-seated aversion to Twitter prevents me from doing it myself, which of my readers is up for starting an account that just reposts one overhyped QC article after another, while appending the words “A CLASSICAL COMPUTER COULD ALSO DO THIS” to each one?
Comment #1 April 17th, 2019 at 3:51 pm
There’s no need to be so excitable about this. Surely by now everybody has read “Classical states, quantum field measurements”, https://iopscience.iop.org/article/10.1088/1402-4896/ab0c53, and understands that random fields and quantum fields are very close indeed to being empirically equivalent? When a quantum field contains a complete random field of quantum non-demolition measurements, the equivalence is exact 🙂
Comment #2 April 17th, 2019 at 9:09 pm
Something close in spirit to this already exists at https://twitter.com/BullshitQuantum (which isn’t to say there isn’t room for more quantum-hype-debunking Twitter accounts!)
Comment #3 April 17th, 2019 at 11:04 pm
I like your heuristics for dismissing bad research and research journalism 🙂 What’s your opinion on switching human efforts from backpropagation/neural nets to making a better map of the human brain (connectome)?
I mean, if there’s no optimal solution for intelligent systems, and it’s just finding good approximations (which by the No free lunch theorem are equivalent) why not reuse the information generated by evolution for billions of years, which is a lot of computational energy/low entropy already spent?
Comment #4 April 18th, 2019 at 3:09 am
Filip #3: That’s sufficiently far outside my expertise that, even if I had a strong opinion, there’d be no reason to listen to me. But I see no reason why efforts to map the human connectome and efforts to build deep nets in whatever way is found to work can’t proceed in parallel, as in practice they surely will.
Comment #5 April 18th, 2019 at 4:03 am
scott:
let’s not stop with pointing out ‘overhyped’ pop science articles; we could have full transparency in all published papers. eg.,
for experimental papers: “the results presented here have not been reproduced elsewhere.”
for theoretical papers: “the predictions of the calculations presented here have not been compared to experiment (and in some cases, are untestable).”
for papers based on research supported by funding: “nothing new or interesting is reported or discussed herein but this paper will help to enable the authors to obtain future funding (which will provide income to my university via overhead charges and the financial support of students and which will pay for trips to conferences held in nice locations.”
Comment #6 April 18th, 2019 at 4:28 am
joe #2 and Richard #5: The genius of the “IN MICE” Twitter account is that it targets a single, specific, objective factual issue—one that’s obvious once pointed out but that the popular accounts (and sometimes even the papers themselves) are hoping you don’t notice—rather than a generalized railing against all bullshit research. I wouldn’t want to lose that.
Incidentally, in quantum information, “this experiment hasn’t been replicated” and “this theory hasn’t been tested experimentally” are almost never the problem—if you think they are, then you’re going to completely miss the reality on the ground. More often the problem is something like: “this theory was tested experimentally, but it needn’t have been, because the results of the experiment were obvious and predictable, and they don’t mean anything like what the popular articles say they do.”
Comment #7 April 18th, 2019 at 5:13 am
Filip #3: You seem to misunderstand the ‘No free lunch theorem’. That theorem only applies under some theoretical conditions. Why would approximations be equivalent in practice?
Comment #8 April 18th, 2019 at 6:10 am
Incidentally, joe #2: thanks for the ref; I hadn’t seen the “Quantum Bullshit Detector”! So far it looks pretty well-calibrated, but the sample set is small and vigilante justice is a high-risk affair. Safer, surely, to tweet factual clarifications (“A CLASSICAL COMPUTER COULD ALSO DO THIS”) and leave the evaluative judgments to the reader?
Comment #9 April 18th, 2019 at 9:47 am
I don’t have a twitter account, and don’t wish to open one. Would someone be kind enough to submit the paper I mention in my #1, “Classical states, quantum field measurements”, https://iopscience.iop.org/article/10.1088/1402-4896/ab0c53, to @BullshitQuantum? The referee struggled with my presentation, but nonetheless said it was a dense, interesting and even elegant paper, so an extra log(3) bits won’t plumb that complexity much, but I’m curious which way @BullshitQuantum will jump.
Comment #10 April 18th, 2019 at 2:54 pm
the BIG difference is that the QC guys do no harm to animals.
So, let them have their fun.
Comment #11 April 19th, 2019 at 5:21 am
“A CLASSICAL COMPUTER COULD ALSO DO THIS”
This is about our preprint. We show that if we have an Oracle Hamiltonian H = -|w> is unknown (Grover problem Oracle), we can organize a projection operation to this unknown state
P = |w><w| . So, we have non-deterministic quantum computation, and we can use even mixed states as initial. However, we have a classical efficiency.
Now I think, can it be nevertheless interesting for some quantum processing (quantum verification)…
Comment #12 April 25th, 2019 at 5:25 am
[…] decide to write here something after all. The title is from a Shtetl-Optimized post. Couple days ago my e-print about quantum state “routing” in qubit trees appears. It is close […]
Comment #13 April 25th, 2019 at 4:08 pm
Rainer #10: I see you’ve never met my friend Erwin Schrödinger, or you’d know what he did with his cat…
Comment #14 July 21st, 2019 at 8:02 pm
[…] Scott Aaronson, University of Texas computer science professor and author of quantum hype-busting blog Shtetl-Optimized, said of the account: “vigilante justice is a high-risk […]
Comment #15 July 27th, 2019 at 6:54 am
[…] Scott Aaronson, College of Texas pc science professor and writer of quantum hype-busting weblog Shtetl-Optimized, mentioned of the account: “vigilante justice is a high-risk […]