Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
AI United States China IT Technology

China Has Now Eclipsed The US in AI Research (washingtonpost.com) 97

Earlier this week, the Obama administration discussed a new strategic plan aimed at fostering the development of AI-centered technologies in the United States. What's striking about it is, the Washington Post notes, although the United States was an early leader in deep-learning research (a subset of the overall branch of AI known as machine learning), China has effectively eclipsed it in terms of the number of papers published annually on the subject (Editor's note: the link could be paywalled; alternate source). From the report: The rate of increase is remarkably steep, reflecting how quickly China's research priorities have shifted. The quality of China's research is also striking. The chart narrows the research to include only those papers that were cited at least once by other researchers, an indication that the papers were influential in the field.
This discussion has been archived. No new comments can be posted.

China Has Now Eclipsed The US in AI Research

Comments Filter:
  • by Anonymous Coward on Saturday October 15, 2016 @11:33PM (#53083659)

    Sad that this even needs to be mentioned, but it's not the quantity of papers that matter, but the *quality*.

    The "publish or perish" mentality needs to go. Being prolific is utterly meaningless if all you write is shit. In fact, it does a disservice to scientific discourse, because now people have to expend extra effort to separate the crap from the shite.
     

    • Chicken chicken chicken chicken chicken chicken chicken chicken [wired.com]
      chicken chicken chicken chicken chicken chicken chicken chicken
      chicken. Chicken chicken chicken chicken chicken chicken chicken
      chicken. Chicken, chicken chicken chicken, chicken chicken, chicken
      chicken chicken “chicken chicken” chicken “chicken chicken”
      chicken. Chicken, chicken chicken chicken chicken chicken chicken
      (chicken chicken) chicken chicken chicken chicken chicken,
      chicken chicken chicken chicken chicken chicken c

    • Re: (Score:2, Informative)

      by Anonymous Coward

      ... which is why they're looking at papers that have been cited, rather than just published by peer review.

      • by Anonymous Coward

        Having a single non-self citation is a really low bar; all you need is different research groups willing to cite each other.

        • by Anonymous Coward

          Having a single non-self citation is a really low bar; all you need is different research groups willing to cite each other.

          Webster University does things like this and it is one reason that their work is not considered "scholarly".

      • That's also not necessarily meaningful, as new published papers will cite older papers, and the authors will most often cite papers they're familiar with (or even published themselves). As a result, the more papers published by a country, the more cited papers published by that country. That, and they only looked for papers that used "deep learning" or "deep neural network", which is not necessarily a proxy for AI research in general. A proper examination would require looking at whose citing the papers, ho

      • ... which is why they're looking at papers that have been cited, rather than just published by peer review.

        That only works if you eliminate cycles from the citation graph. Otherwise, you cite me, and I will cite you. Win-win with no quality needed.

    • by jandersen ( 462034 ) on Sunday October 16, 2016 @02:23AM (#53084019)

      This is a very valid point, but fortunately the OP has some mention of this:

      The quality of China's research is also striking. The chart narrows the research to include only those papers that were cited at least once by other researchers, an indication that the papers were influential in the field.

      Let us not blind ourseves with the prejudice, that because the researchers are Chinese, it must somehow be of a poorer quality. China is simply investing much more aggressively in education than the US, so it is no surprise they are able to produce more, good research. That said, I think it is more relevant to look at this per capita; there are 300 million Americans and 1300 million Chinese, which is ~4 times as many, so until they produce >4 times as much good quality research, they are still catching up. They will get there, without a doubt, but I think there is some way to go still.

      • by Calydor ( 739835 ) on Sunday October 16, 2016 @02:31AM (#53084037)

        The thing is that there have been multiple articles on Slashdot over the years showing a systematic pattern of fraud, cheating, plagiarism etc. in Chinese research done to bolster the appearance that China was developing far faster than they actually were. There have also been articles revealing metaphorical circle-jerk problems with paper A citing paper B citing paper C citing paper A, with the researchers all being good friends of each other.

        • by AmiMoJo ( 196126 )

          The same claims are made about western papers, especially in fields like medical research and psychology.

          The west also has many low quality journals that publish papers which are nothing more than "take me off your fucking making list" 100 times. Even the slightly respectable ones publish computer generated nonsense.

          A better metric is to look at how competitive China is in the field of AI.

        • "[F]raud, cheating, plagiarism, etc." in *low-end* research, which we also have in spades in the U.S. and in the West more generally (it's really bad in a lot of the also-ran European countries). At the top end, Chinese research is every bit competitive with other players in serious global research, and they have more resources available to them, which they can apply to problems without nearly so much systemic overhead thanks to their particular governing system.

      • It's the same in my fields (applied maths/fluid dynamics/ship hydrodynamics).
        15 years ago, the standard of papers by Chinese researchers was generally very poor, but now they are
        contributing as much as 1/3 of papers in top quality, peer-reviewed journals.

        China is building about one university per week at present, and most of them have a scientific focus.
        Of course, the standards at those new universities will initially be mediocre - it takes time
        to set up labs and to put lecturers of quality in place.
        Like In

        • they see STEM as the future, the best way out of poverty and for modernising their country. OTOH, look at the erosion of science courses in many western countries

          Just look at the people studying science at those western universities. These days, a fair number of them are from Asia or the Middle East. And maybe it's a cultural thing, science isn't just unpopular here or perceived as "too hard", it's actually looked down on a little. I can remember when being an engineer in any field got you social status, and putting that title next to your name meant something. These days? Just listen to every single movie dad talking to their disappointing child: "you could ha

          • by Anonymous Coward
            I had an idea treatment for a rare deadly disease that "real" doctors had been working on for around 40 years and not only getting nowhere, the only clinical trial that showed any treatment versus placebo difference before ours, showed that the standard of care made it worse. Nevertheless, We were constantly reminded, they are the experts.

            After fighting the "experts" to get my idea tested in humans, it worked but the clinicians didn't include any of our team on either NEJM paper. The inventors (the PhDs)
      • Let us not blind ourseves with the prejudice, that because the researchers are Chinese, it must somehow be of a poorer quality.

        It's been demonstrated that Chinese papers are of an average poorer quality.

        China is simply investing much more aggressively in education than the US, so it is no surprise they are able to produce more, good research.

        Dollars spent on education do not automatically produce superior outcomes.

      • by Nemyst ( 1383049 )
        I disagree with the "quality" metric though. A single citation is all you need to be influential? Nice, then I guess just about every paper is influential. Seriously though, you need more than just one citation to be influential, and you also need to look at who cites them. Is it all Chinese authors citing themselves? Is it all authors at the same university? Are the cited papers actually relevant, or are they just ticking a box in the previous work section?

        Call me biased all you wish, but I still don't t
    • I'm not in this field but since a considerable amount of research in this area is done by companies in the US (Goggle, Apple, etc etc). who might not want to publish their results, does this mean the US is less productive or simply reached the phase where such knowledge is actively implemented?

      How do the results look if you look at granted patents?

      Just curious, as I said, not at all in this field.
    • by Anonymous Coward

      What academic groups tend to do is reference each others papers. Maybe a post-doc graduates and their past supervisors.

    • TFS

      The quality of China's research is also striking. The chart narrows the research to include only those papers that were cited at least once by other researchers, an indication that the papers were influential in the field.

      So quantity _and_ quality. With a population 4 times as the US population combined with an important growth, demographics show that China is on the verge of being number one, everywhere.

  • by fluffernutter ( 1411889 ) on Saturday October 15, 2016 @11:36PM (#53083669)
    Is China closer to self driving cars than Google is?
  • Quality Quantity (Score:5, Insightful)

    by Anonymous Coward on Saturday October 15, 2016 @11:36PM (#53083671)

    Sure they can churn out more papers.. but are they making any progress. That remains to be seen.

  • by Anonymous Coward

    The solution here should be simple: let them pay for the research and we'll just steal it. :D

  • by Anonymous Coward

    What the article completely misses is that most AI research in the US is commercial. Google/Apple/Facebook/Microsoft all have invested heavily in it, and will happily pay hundreds of millions for your deep learning company at the drop of a hat. Meanwhile China has publicly funded research and, well, Alibaba. That AI is now eminently commercializable is a further reason not to worry too much about publicly funding it a lot. Publicly funded research is best used for science that isn't profitable within either

  • As a reviewer... (Score:5, Informative)

    by Anonymous Coward on Saturday October 15, 2016 @11:57PM (#53083735)

    Oh look, something I can comment on.

    I don't research AI or ML, but I work in related-enough areas that I get called upon to referee for the top (and sometimes not-so-top) conferences/journals in the area pretty regularly.

    The quality of the Chinese papers is usually either very low or very high but only because it's based on obviously fabricated data. Among double blind submissions (for which neither reviewers nor authors know the others' identity), I've probably recommended for acceptance about 40% of American/European/Israeli/Indian papers and less than 10% of the Chinese ones. You can find out the origin when the paper finally gets published somewhere and/or uploaded to Arxiv.

    Outside of Baidu research and a small group at Tsinghua University, I have little faith in the quality of Chinese research.

    • by lorinc ( 2470890 )

      They are indeed very low. The thing is that publishing something is a mandatory part of many students curriculum in China. This means that you have many papers that are written as an exercise part of a project at BSc level. Obviously, there are little meat on those paper, and the writing style is atrocious.

      But hey, quantity is the best, right?

    • "The quality of the Chinese papers is usually either very low or very high but only because it's based on obviously fabricated data."

      That certainly was the case in many scientific fields until about 10 years. It's not as prevalent now, especially in mathematics and computer science.
      Low quality papers and research are now more the domain of developing countries like several Arabic nations, Brazil and some smaller Asian nations.

  • by 93 Escort Wagon ( 326346 ) on Saturday October 15, 2016 @11:59PM (#53083741)

    I really hate to say this, but it's not inconceivable that the Chinese might be intentionally gaming the citation-based rankings by citing each other a lot (we have seen such things happen in the west, too, albeit on a smaller scale). I'd like to see how many citations these papers get from non-Chinese sources as well.

    • by Anonymous Coward

      Nearly a decade ago fraudulent scientific papers [economist.com] had already become a serious problem for Chinese academics, including fraudulent citations (from the linked article):

      From 2002 to 2012, more than 1m Chinese papers were published in SCI journals; they ranked sixth for the number of times cited by others.
      ...
      In 2010, however, Nature had also noted rising concerns about fraud in Chinese research, reporting that in one Chinese government survey, a third of more than 6,000 scientific researchers at six leading institutions admitted to plagiarism, falsification or fabrication.

      (There are possibly mitigating factors, so read the link before drawing too strong of a conclusion)

    • I really hate to say this, but it's not inconceivable that the Chinese might be intentionally gaming the citation-based rankings by citing each other a lot (we have seen such things happen in the west, too, albeit on a smaller scale). I'd like to see how many citations these papers get from non-Chinese sources as well.

      Citations in academia are also not entirely a measure of how useful the cited paper is--it is a political act and one about increasing your reputation, expressing appreciation for someone else's work, showing you know the field, and even increasing the number of people who notice and read your paper, all depending on your particular position. Most cited works in a good paper are on-point to a significant degree, but there is a lot of room for discretion.

      So citation count can occasionally be useful, but it d

  • by Anonymous Coward

    Number of papers published and number of papers with at least 1 citation are terrible metrics. There are a large number of conferences and journals where it is trivial to get awful work published. The 1 citation metric also doesn't state that it excludes self-citations. This is on par with stating that something is true because I saw it on wikipedia.

    If the numbers hold for conferences on lists such as Conference Rankings [ualberta.ca] and they analyzed impact factor, then that might actually have some merit.

    • It also fails to account for unpublished internal/proprietary research - who knows what unreported breakthroughs have been made at IBM/Google/MS/DARPA/etc.?
  • We need to declare cyberwar on them!! :/
    • The US did that years ago. And lost.
      Still, you've got Trump and are leading China in Jeebus research, so it's not all doom and gloom.

  • by Anonymous Coward

    Just mass publishing papers doesn't mean that much. People get their PHD's in astrophysics by publishing a paper on how to make astrophysics more accessible to kindergartners. What is important is the meaningful breakthroughs, like defeating Le Sodol at Go, self driving cars, natural language interfaces, and so forth.

    If China is ahead in that, hats off to them, but so far it seems like the big US tech firms (Google, Facebook, etc...) are at the forefront.

  • Most research isn't published in white papers. White papers are propaganda and strategic mass deception mostly aimed at controlling the public and non classified realm. At best its entertainment. Everything going on in this field is classified and not published in papers. Today we for sure have a much more advanced AI and robotics program than China or Russia probably combined. Our military budget is actually greater than all major countries combined. We have more technology than the rest of the world combi

    • "Our military budget is actually greater than all major countries combined."
      Yes. Maybe the US should double the spending so it can win in Afghanistan faster.

      "We have fully mapped the brain..."

      Bullshit.

    • by SirSlud ( 67381 )

      /. is an excellent source of hilarious crackpots.

  • ...doing it, you know it's a bubble.

  • China has eclipsed the US in gaming the citation metrics. Most of the seminal papers in AI are either US or Canadian. Nearly all of the new exciting ones are from the US, Canada or the UK (DeepMind in particular is on fire right now). You can publish a ton and not make a difference. Like it or not, only maybe 2-3 dozen papers really matter each year.

  • The 1% doesn't want competition for control.
  • Oh come on, the quality measure here is just ludicrous. China is certainly on the way up, but also leads in number of bad papers and circle jerks for paper cites . That's a common tactic in boiler room publish or perish science and not just in China. Anyone can get cited once or even twice; people often cite themselves just to drive their numbers up. If you read Nature this is a common subject, as is China's magnificent drive to game the numbers.

    What matters are results, and here are some countries that

    • "What matters are results, and here are some countries that have demonstrated more talent in AI than China has so far. In no particular order, and I've definitely missed some:
      - UK
      - Canada
      - United States
      - Israel
      - Germany"

      And those nations will probably continue to lead until the next AI winter.
      (Remember the circle-jerking and horse-shit research that led to previous winters?)

  • Yes, I know it's pretty standard, but it's about as telling as the klines count in code. This is true not only of this particular field of research. It's a problem in academia in general. The whole publish-or-perish system encourages it and at some point the publication volume reaches a critical mass where extra information starts to take away from the understanding of the subject. Knowledge requires culling of unnecessary information. And the chase for higher citation count does nothing to encourage n
  • Whether those paper publications are high quality or not, it definitely shows the difference between a government with full control over its economy and one that's given all the control over to businesses. In my opinion, there are 3 primary reasons for the decline of science education in the US:
    1. Anti-intellectualism on a massive scale -- other countries shower their scientists with praise and research dollars, and we dismiss research as something those "leftist egghead liberals in the ivory tower" do. Oth

  • While I don't think that the number of papers is a good way to measure the advancement of research I am thinking that China may be in a better position than the US for AI research.
    They are more than a billion, they don't care as much about privacy as we do and, should their project be regarded as interesting by the government, they can get people working for them for free (internship, ...). And when we consider that a significant part of AI is gathering data, the sheer number can prove a significant advanta

  • Come to think of it ... what good is the Great Wall ??
  • "China now delivers low-cost and high-quality products" --Jack Ma

An age is called Dark not because the light fails to shine, but because people refuse to see it. -- James Michener, "Space"

Working...