I never visit Twitter normally, so when I actually click an
interesting link I always land in a 5-10 tweet chain that should
have been a short blog post.
But somehow people put up with it and rather use external web
services that automatically compile tweets into continuous text.
That's a surprisingly good tweet thread that does an excellent job of pragmatically walking the line between "quantum computing is impossible" and "real quantum computers are only 2-3 years away".
Advances are being made, but we're still very much in the primitive science experiment stage.
I feel like you can replace "quantum computing" in this rant with almost any other major gov science initiative and it would still have relevance. Lunar colonization. Nuclear fusion. Genome therapy. All 5 years away since 2000. All require major materials breakthroughs. All hyped in the glossy popular media ;)
Establishing an actual colony on the moon instead of just a temporary outpost will require a huge amount of basic scientific research. This goes way beyond just economics and politics? For example how do we keep people healthy and prevent bone mineral loss and muscle wasting in low gravity? How do we mitigate the effects of increased background radiation? Any real sustainable colony will need people working out on the surface so they can't just hide in underground tunnels.
We could easily have done that research in the 20 years after the cold war. These are not research that requires state of the art technology (except for radiation shielding which requires good materials science). But we didn't, because there was no immediate demand.
Nuclear Fusion has made significant progress, and generally speaking the joke has been "its 20 years away and always will be". Only starting 2020 did people start to estimate shorter timelines, with ITER expecting to have net positive energy between 2025 and 2035. There are several fusion startups/projects looking to beat ITER, including SPARC and Commonwealth. In 2020 a detailed peer review of the system could find no major obvious flaws. SPARC is hoping to be done by 2025 with net postiive energy probably a few years after that.
Neither ITER nor SPARC require any materials breakthroughs, but do rely on breakthroughs made over the last 30 years.
People's expectations about a faster fusion timeline were always based on funding that never manifested. See https://i.imgur.com/3vYLQmm.png
We are in the ballpark of 1976 timeline expectations given the amount of spending we committed to fusion.
ITER was supposed to achieve fusion in 2016, but that has now been delayed until 2026:
When ITER first received formal approval in 2006, it was slated to first achieve fusion in 2016, a date which has since been pushed back at least 10 years. Issues with component construction and design disagreements have been blamed for the delays.[1]
> People's expectations about a faster fusion timeline were always based on funding that never manifested.
ITER was initially budgeted €5 billion which was provided, then jumped to €15 billion, which was also provided:
The project was officially begun in 2006 with an estimated cost of €5 billion and date for the beginning of operations—or first plasma—in 2016. Those figures quickly changed to €15 billion and 2019, but confidence in those numbers has eroded over the years.[2]
And yet the timeline just for ITER to do an experimental run is still 10 years, and then:
If everything goes to plan, ITER will pave the way for another reactor, called DEMO, which will expand the technologies perfected by ITER to an industrial scale, and hopefully prove that nuclear fusion is a viable source of energy.[2]
Very unclear where these predictions and budget numbers come from. Apart from the 1976 budget, the rest just seem to be made up, and the end dates seem arbitrary.
On the point of the media. Of course some blame can be put on the researchers but science media also presents perverse incentives for them to popularize their findings by extrapolating and speculating on sensational consequences.
Even if they did, it's unlikely we will find out in the current political climate. We have to wait until 'end of pandemic' before people will rationally analyze the data.
If you assume a reasonable interpretation of the analogy, he means they’ll eventually get it right. I.e. it’ll take another N tries, where N is of course finite.
Theorem: my bathroom has achieved quantum supremacy, greater than any current industry system.
Proof:
- No current super computer can accurately (much less real time) model an hour long high fidelity video of my bathtub full of water set into motion by jiggling my foot against it.
- No current quantum computer can accelerate a problem of that size.
Let me know when quantum computing is more than triviality in expensive components.
It may be a loss for Microsoft but it is a win for the scientific method. And it is also one more score for the push to include all data with scientific papers and not just the pretty version.
The authors say that they cut the data out for "aesthetic" reasons. Personally, I don't buy it. They claimed a groundbreaking result, that obviously doesn't hold for the "ugly" data. It's possible that they deluded themselves, but from where I sit, that just counts as two more victims of fraud.
More often than not people that do a thing like this are deluding themselves. I've seen this in founders too.
Outright fraud is at least 10 times rarer than self delusion by my quick tally over 14 years of looking at companies, I don't see any reason why the academic world would be substantially different.
The range goes all the way from 'wouldn't it be nice if 'x' existed, so what does it matter if I have to fake it for a bit' to people who refuse to believe they are wrong right up until the debt collectors arrive and the whole thing falls apart. Fraud and self delusion are two completely different things and you shouldn't conflate them like that.
I would really like to understand how committing outright fraud by omitting contra data for "aesthetic" reasons and then getting caught is a win for the scientific method.
By this logic what would be a loss for the scientific method? Committing fraud and never getting caught?
It's the 'getting caught' bit that matters. In science truth wins out over the long term and that is in large part due to how the scientific method has been structured. Keep in mind that the difference between fraud and self delusion is very hard to determine and that humans suffer a lot from the second. That we also fix fraud with the same countermeasures is a win.
Personally, I think I'd enjoy the fame of being the first to prove the existence of Majorana fermions much more than whatever promotions / bonuses etc may have been on the table. One might expect a Nobel prize for such a result. This may be cynical, but my guess is that they wanted to get their stake in the ground first, and optimistically saw the misbehaving data as a hardware glitch that would get ironed out before they received any real scrutiny.
Seems a lot like a deliberate lie in order to gain funding. While Microsoft backing this means they have solid financials, they are also pulling in serious amounts of funding from state actors. To the point where in for instance Denmark they where described as one year “draining the entire state innovation fund”. Doing fund raising based on articles that where knowingly manipulated to support untruthful claims really’s should be treated like financial fraud. Though likely this will come out with just a reprimanded and no real consequences.
Locklin explains it best in "Quantum computing as a field is obvious bullshit" [1]:
"quantum computing” enthusiasts expect you to overlook the fact that they haven’t a clue as to how to build and manipulate quantum coherent forms of matter necessary to achieve quantum computation. A quantum computer capable of truly factoring the number 21 is missing in action. In fact, the factoring of the number 15 into 3 and 5 is a bit of a parlour trick, as they design the experiment while knowing the answer, thus leaving out the gates required if we didn’t know how to factor 15. The actual number of gates needed to factor a n-bit number is 72 * n^3; so for 15, it’s 4 bits, 4608 gates; not happening any time soon."
That argument is quite bullshit, no? Just because something takes a long time doesn’t mean it won’t happen. We invented the first algorithms thousands of years ago, but only constructed working computers to execute those algorithms 100 years ago.
I don't entirely disagree but, isn't all revolution research like this ? You circle around an objective, throwing random things at it with no idea how to achieve it until someone has an eureka moment and you suddenly pass it. The bigger the objective, the bigger the wall to pass, the larger the eureka needed.
If we only funded or took seriously research where the researcher actually knew the solution or had a good idea how to reach it, we wouldn't have made that kind of a tech jump in the last century, let alone in the last 50 years.
They meant, if a group of spies were pointing guns at the physicists and engineers and saying "You will build a functional quantum computer within 5 years, or we will pull the trigger," then their life expectancy will be exactly five years.
(I assume the "factoring 15 parlor trick" doesn't count as functional, since you have to punt the number of gates just to get the expected answer.)
> In 2010, I laid out an argument against quantum computing as a field based on the fact that no observable progress has taken place.
I'm a physicist by training and a quant by profession - from what I can gather, not unlike the author.
Without judging the merits of quantum computing I will just notice that arguments like "has no merit because it has made no progress" are backwards-looking. You're not assessing anything from first principles, you're just saying that nothing has happened _yet_.
In finance this reminds me of trend followers. Yes, you can make money by following the trend, but you're the last crowd to jump on the opportunity, right behind the guys who were looking ahead instead of back.
That said, such people also provide a nice balance against "we don't know everything therefore anything is possible" crackpots.
Oh wow, looking at their graph and the portion they cut out for "aesthetic reasons" this looks almost like fraud to me. It's at least hard to imagine how Kouwenhoven or anyone in his group could not have found this highly problematic. I did my PhD in a similar field (superconducting quantum computing) and if I had cut data from a graph like that I would've gotten a really strong reprimand from my supervisors. If you do a series of IV curves at equidistant points you cannot simply cut out the data you don't like, and if you do remove irrelevant data points (again, you probably shouldn't) you need to replace them with a placeholder value such as a black background to make it immediately clear that you left out some data. Just cropping your graph (which also makes most labels on the X axis invalid) is just insane, they hammer that into your head during the first undergrad lab classes already. I'm sure if they had marked the missing curves with a placeholder the reviewers would've noticed immediately and this article would probably not have been published. So from an outside perspective this doesn't look good at all for the authors.
Well, I guess this shows why it's so important to publish the raw data as supplementary material along with the article.
I remember that when I was working as an undergrad research assistant in a physics lab, a PhD student and I had plotted some data, to see if the relationship between 2 variables was linear. We had cut off many of the data points to the left and right, keeping only a small region in the middle. When we showed it to our supervisor, he came back with "of course it looks linear if you only look at a small interval, that's just Taylor expansion." He then explained to us that it's incredibly scientifically sketchy to leave out perfectly good data from one's plots. Not fooling yourself is surprisingly hard when you're wishing for a specific outcome.
On a related note to the supplementary data thing, I've recently seen a lot of people complaining online about not being able to reproduce the results of machine learning papers that don't have the code for their models included in the supplement. "Code, or it didn't happen." I guess each field has its own problems with replication of results.
52 comments
[ 3.9 ms ] story [ 111 ms ] threadIf you're interested, I made a video about these "anyons" [1]. Or, more briefly there's this [2].
[1] https://www.youtube.com/watch?v=Y0mn7UBonSs
[2] https://quantumcomputing.stackexchange.com/questions/2030/wh...
But somehow people put up with it and rather use external web services that automatically compile tweets into continuous text.
Advances are being made, but we're still very much in the primitive science experiment stage.
the other two are pipe dreams though.
Neither ITER nor SPARC require any materials breakthroughs, but do rely on breakthroughs made over the last 30 years.
People's expectations about a faster fusion timeline were always based on funding that never manifested. See https://i.imgur.com/3vYLQmm.png
We are in the ballpark of 1976 timeline expectations given the amount of spending we committed to fusion.
When ITER first received formal approval in 2006, it was slated to first achieve fusion in 2016, a date which has since been pushed back at least 10 years. Issues with component construction and design disagreements have been blamed for the delays.[1]
> People's expectations about a faster fusion timeline were always based on funding that never manifested.
ITER was initially budgeted €5 billion which was provided, then jumped to €15 billion, which was also provided:
The project was officially begun in 2006 with an estimated cost of €5 billion and date for the beginning of operations—or first plasma—in 2016. Those figures quickly changed to €15 billion and 2019, but confidence in those numbers has eroded over the years.[2]
And yet the timeline just for ITER to do an experimental run is still 10 years, and then:
If everything goes to plan, ITER will pave the way for another reactor, called DEMO, which will expand the technologies perfected by ITER to an industrial scale, and hopefully prove that nuclear fusion is a viable source of energy.[2]
> See https://i.imgur.com/3vYLQmm.png
Very unclear where these predictions and budget numbers come from. Apart from the 1976 budget, the rest just seem to be made up, and the end dates seem arbitrary.
[1] https://www.discovermagazine.com/technology/why-nuclear-fusi...
[2] https://www.sciencemag.org/news/2015/11/iter-fusion-project-...
http://i.imgur.com/3vYLQmm.png
https://old.reddit.com/r/energy/comments/5budos/fusion_is_al...
If funding had not been cut to far below "fusion never" we would have been done decades ago.
I sure hope the scientists developing the COVID-19 vaccines didn't cut out any data points "for aesthetics".
https://www.forbes.com/sites/tommybeer/2020/11/11/trump-pred...
Saying this is the death of QC is like saying Windows 1 means the GUI is a worthless concept.
Proof:
- No current super computer can accurately (much less real time) model an hour long high fidelity video of my bathtub full of water set into motion by jiggling my foot against it.
- No current quantum computer can accelerate a problem of that size.
Let me know when quantum computing is more than triviality in expensive components.
Outright fraud is at least 10 times rarer than self delusion by my quick tally over 14 years of looking at companies, I don't see any reason why the academic world would be substantially different.
The range goes all the way from 'wouldn't it be nice if 'x' existed, so what does it matter if I have to fake it for a bit' to people who refuse to believe they are wrong right up until the debt collectors arrive and the whole thing falls apart. Fraud and self delusion are two completely different things and you shouldn't conflate them like that.
By this logic what would be a loss for the scientific method? Committing fraud and never getting caught?
Meanwhile teh M1 is doing "boring optimization of binary computing"...
...sorry...quantum joke
"quantum computing” enthusiasts expect you to overlook the fact that they haven’t a clue as to how to build and manipulate quantum coherent forms of matter necessary to achieve quantum computation. A quantum computer capable of truly factoring the number 21 is missing in action. In fact, the factoring of the number 15 into 3 and 5 is a bit of a parlour trick, as they design the experiment while knowing the answer, thus leaving out the gates required if we didn’t know how to factor 15. The actual number of gates needed to factor a n-bit number is 72 * n^3; so for 15, it’s 4 bits, 4608 gates; not happening any time soon."
[1] https://scottlocklin.wordpress.com/2019/01/15/quantum-comput...
If we only funded or took seriously research where the researcher actually knew the solution or had a good idea how to reach it, we wouldn't have made that kind of a tech jump in the last century, let alone in the last 50 years.
Excuse me? They have plenty of "clues". Have a look at this for some recent work from the google group:
https://scirate.com/arxiv/2102.06132
Will it lead to quantum computers in the future remains to be seen, but "haven't a clue" is just nonsense.
They meant, if a group of spies were pointing guns at the physicists and engineers and saying "You will build a functional quantum computer within 5 years, or we will pull the trigger," then their life expectancy will be exactly five years.
(I assume the "factoring 15 parlor trick" doesn't count as functional, since you have to punt the number of gates just to get the expected answer.)
I'm a physicist by training and a quant by profession - from what I can gather, not unlike the author.
Without judging the merits of quantum computing I will just notice that arguments like "has no merit because it has made no progress" are backwards-looking. You're not assessing anything from first principles, you're just saying that nothing has happened _yet_.
In finance this reminds me of trend followers. Yes, you can make money by following the trend, but you're the last crowd to jump on the opportunity, right behind the guys who were looking ahead instead of back.
That said, such people also provide a nice balance against "we don't know everything therefore anything is possible" crackpots.
Well, I guess this shows why it's so important to publish the raw data as supplementary material along with the article.
On a related note to the supplementary data thing, I've recently seen a lot of people complaining online about not being able to reproduce the results of machine learning papers that don't have the code for their models included in the supplement. "Code, or it didn't happen." I guess each field has its own problems with replication of results.