Toshiba Touts Algorithm That's Faster Than a Supercomputer (bloomberg.com) 35
It's a tantalizing prospect for traders whose success often hinges on microseconds: a desktop PC algorithm that crunches market data faster than today's most advanced supercomputers. Japan's Toshiba says it has the technology to make such rapid-fire calculations a reality -- not quite quantum computing, but perhaps the next best thing. From a report: The claim is being met with a mix of intrigue and skepticism at financial firms in Tokyo and around the world. Toshiba's "Simulated Bifurcation Algorithm" is designed to harness the principles behind quantum computers without requiring the use of such machines, which currently have limited applications and can cost millions of dollars to build and keep near absolute zero temperature. Toshiba says its technology, which may also have uses outside finance, runs on PCs made from off-the-shelf components.
"You can just plug it into a server and run it at room temperature," Kosuke Tatsumura, a senior research scientist at Toshiba's Computer & Network Systems Laboratory, said in an interview. The Tokyo-based conglomerate, while best known for its consumer electronics and nuclear reactors, has long conducted research into advanced technologies. Toshiba has said it needs a partner to adopt the algorithm for real-world use, and financial firms have taken notice as they grapple for an edge in markets increasingly dominated by machines. Banks, brokerages and asset managers have all been experimenting with quantum computing, although viable applications are generally considered to be some time away.
"You can just plug it into a server and run it at room temperature," Kosuke Tatsumura, a senior research scientist at Toshiba's Computer & Network Systems Laboratory, said in an interview. The Tokyo-based conglomerate, while best known for its consumer electronics and nuclear reactors, has long conducted research into advanced technologies. Toshiba has said it needs a partner to adopt the algorithm for real-world use, and financial firms have taken notice as they grapple for an edge in markets increasingly dominated by machines. Banks, brokerages and asset managers have all been experimenting with quantum computing, although viable applications are generally considered to be some time away.
Why sell it? (Score:2)
Make a few billion dollars with it and THEN sell it.
Or NOT and continue making money.
Unless it's just an ego thing and they want the recognition.
Re: (Score:1)
Re: (Score:2)
Unless it's just an ego thing and they want the recognition.
Even then it does not make sense. Nothing strokes egos more than several billion dollars just look at Musk, Gates, Bezos, Zuckerberg etc.
Selling quantum computer (Score:4, Funny)
Toshiba manager: We can't make a quantum computer
Marketer: What can you come up with?
Toshiba manger: I can make a fake quantum computer
Marketer: I can sell that
And here we are.
Re: (Score:2)
Re: (Score:2)
it's not a fake quantum computer, is simulating a quantum computer.
Re: (Score:1)
Piped Piper? (Score:2)
"You can just plug it into a server and run it at room temperature," Richard Hendricks said in an interview
Re: (Score:2)
This is bad (Score:2)
Cooling the CPU running it is hard. (Score:2)
Silly article (Score:5, Informative)
Here's a link to the guy's paper:
https://advances.sciencemag.or... [sciencemag.org]
It looks like he claims to have found an easily parallelizable algorithm to find approximate solutions to minimizing an Ising model. That's basically what DWave's quantum computer does. If you try hard enough, you can cast some interesting problems as Ising models.
I'm not really sure finding arbitrage opportunities is one of them, but maybe?
Re: (Score:2)
See https://fpl2019.bsc.es/sites/d... [fpl2019.bsc.es] for FPGA optimization presentation as well.
Re: (Score:2)
How much more restrained could he have been? Interviewer says, "would this be useful for finance?" CEO says, in so many words, "well, yeah,... that, or whatever..."
Re: (Score:2)
I think finance is probably the least interesting application. Once you release a new algorithm into the arbitrage game, everyone adopts it, and you've hit the status quo. The market gets ever so slightly more efficient and absolutely nothing of substance changes.
Other combinatorial problems can have a lot more impact. Just to pick a random example, airline routing got a lot more efficient and brought prices down quite a lot.
Re: (Score:2)
If this really worked (at least for Wall Street), then no person in his right mind would publish it. It would be either sold to a Wall Street firm or used to get a Wall Street job. That it's published is an indication that it either doesn't work or that it isn't practical.
Re: (Score:2)
It's an algorithm for solving a math problem. The author does research in that field. I expect some manager came and asked him WTF is a combinatorial problem and he gave arbitrage trading as a possible application. Manager then ran with it.
Being able to do arbitrage trading a wee bit faster is really pretty much irrelevant to anyone who's not an arbitrage trader. Other applications might be more interesting.
This is a category error (Score:5, Insightful)
The comparison "algorithm faster than a super computer" is utterly nonsensical, because they are two different kinds of things.
Moreover, any algorithm "faster than a supercomputer" - could just be run ON that supercomputer to negate the statement.
It's like saying you have a bread recipe that is faster than making bread.
Re: (Score:3)
Okay, you got my attention. Tell me about this bread recipe of yours.
Re: (Score:2)
Okay, you got my attention. Tell me about this bread recipe of yours.
I'm not sure you're ready for fake-quantum bread.
What I can tell you, is that this recipe makes bread FASTER THAN MAKING BREAD.
I could say more, but you risk your sanity...
Re: (Score:2)
EXCUSE ME
The recipe for bread is faster than a pizza oven.
That is eq
Re: (Score:2)
50 thousand quatloos for the recipe!
Re: (Score:2)
Trading should be a capital crime. (Score:2, Interesting)
Somebody can make me work ... with money he got ... from shuffling goods around ... by preying on those more desperate ... with things he got from those less desperate. Despite never having worked a single second in his life.
Going so far to start out with borrowed money, paid back only with money from those schemes.
I'd say there never was a better textbook definition of a leech and a moocher, than that. But I'd be insulting the leeches (the animal).
So I, as a worker, have to compete in the market, with some
Re: (Score:2)
The money changers, those doing it correctly, investigate and invest in companies that are likely to succeed. This insures that capital gets funneled to areas where it is most useful. Do on let Dunning-Krueger convince you that you know more about money management than you actually do. It may look chaotic, inefficient, and unfair from the outside, but that is simply because you don't know what you're looking at.
Kinda like truck driving, retail (Score:2)
I totally get that, the feeling. You're out there driving a truck or stocking shelves or assembling stuff or whatever it is you do, while somebody else is working at their computer doing something about stocks or whatever, it's not entirely clear what *exactly* they do, but it doesn't look much like work to you. And they make a lot more money than you, at least at the moment. (Traders paychecks are a roller coaster). Fuck that guy, right!
I mean if you do hot shots for a living, that's something useful -
Re: (Score:2)
Ha ha, well done.
Re: (Score:2)
If there's no trading, there's no market. If there's no market, where are you going to invest your hard earned dollars? Where's your employer going to put your 401k? It wouldn't even exist. Ironically, your screed against capitalism sounds just like a Republican's rant against welfare recipients. The grass is always greener on the other side. Welfare recipients actually have rather poor health outcomes. The real hardcore traders tend to burn themselves out. They did a lot of coke in the 80s. That c
Of Toshiba and Submarine propellers (Score:3)
Toshiba has said it needs a partner to adopt the algorithm for real-world use,
Ahh, so all they have is a claim. They haven't actually used it for anything useful.
and financial firms have taken notice as they grapple for an edge in markets increasingly dominated by machines
You know, 2008 was a long time ago now. Crashes happen in cycles going from 11-17 years mostly. The next one is coming. Don't know when, but we're closer to it than were are to 2008. Sometimes being first in line is a good thing. And as a lot of people are about to find out in a hurry - sometimes it's not.
Unlikely (Score:2)
If Toshiba really found an algorithm that significantly outperformed existing arbitrage-finding algorithms, they would not be announcing. It. They would be dominating the world currency markets instead, trouncing all the existing players. It doesn't matter if they have the expertise on staff, one does not forego trillion-dollar opportunities because they don't feel like doing some hiring interviews. Toshiba has plenty of capital and employment infrastructure to make something like this happen.
Sounds right. Except it applies to Intel (Score:2)
What you said seems to make sense. However, it occurs to me that it also pretty much applies to Intel. Intel makes chips that are a bit faster for these particular mathematical operations, l than their competition does, and has been making them for decades. Intel COULD keep their very best hardware and microcode for themselves, and use it to compete in any industry - security (crypto), rendering, trading, whatever.
The chip doesn't do any trading of any assets. The chip doesn't know anything about any ma
Supercomputer suitability (Score:1)
The fact that an algorithm performs better not on a supercomputer isn't really saying much, other than that the characteristics of the algorithm are a poor match for a typical HPC system. This can occur for a lot of reasons: it may not lend itself well to a high degree of parallelism, it may involve moving data back-and-forth during the computation, the data it has to work on may be small and the amount of processing to do on one data point may be smaller than the amount of overhead incurred in getting it u