Artificial Intelligence

Amazing New Chinese A.I.-Powered Language Model Wu Dao 2.0 Unveiled

It's ten times more powerful than the current U.S. effort.

|

Earlier this month, Chinese artificial intelligence (A.I.) researchers at the Beijing Academy of Artificial Intelligence (BAAI) unveiled Wu Dao 2.0, the world's biggest natural language processing (NLP) model. And it's a big deal.

NLP is a branch of A.I. research that aims to give computers the ability to understand text and spoken words and respond to them in much the same way human beings can.

Last year, the San Francisco–based nonprofit A.I. research laboratory OpenAI wowed the world when it released its GPT-3 (Generative Pre-trained Transformer 3) language model. GPT-3 is a 175 billion–parameter deep learning model trained on text datasets with hundreds of billions of words. A parameter is a calculation in a neural network that shapes the model's data by assigning to each chunk a greater or lesser weighting, thus providing the neural network a learned perspective on the data.

Back in November, The New York Times reported that GPT-3 "generates tweets, pens poetry, summarizes emails, answers trivia questions, translates languages and even writes its own computer programs, all with very little prompting." GPT-3, move on over. Wu Dao 2.0 is here.

Wu Dao 2.0 (Chinese for enlightenment) is ten times larger than GPT-3, using 1.75 trillion parameters to simulate conversational speech, write poems, understand pictures, and even generate recipes. In addition, as the South China Morning Post reports, Wu Dao 2.0 is multimodal, covering both Chinese and English with skills acquired by studying 4.9 terabytes of images and texts, including 1.2 terabytes each of Chinese and English texts.

"Wu Dao 2.0's mulitmodal design affords it a range of skills, including the ability to perform natural language processing, text generation, image recognition, and image generation tasks," reports VentureBeat. "It can write essays, poems, and couplets in traditional Chinese, as well as captioning images and creating nearly photorealistic artwork, given natural language descriptions." In addition, Wu Dao 2.0 can predict the 3D structures of proteins, like DeepMind's AlphaFold, and can also power "virtual idols." Just recently, BAAI researchers unveiled Hua Zhibing, China's first A.I.-powered virtual student (see image below).

In 1950, British mathematician and computer scientist Alan Turing proposed a test for intelligence in a computer, requiring that a human being should be unable to distinguish the machine from another human being by using the replies to questions asked of both.

Would Wu Dao 2.0 pass this test? Perhaps. "Wu Dao 2.0 aims to enable machines to think like humans and achieve cognitive abilities beyond the Turing test," said Jie Tang, the lead researcher behind Wu Dao 2.0. If not Wu Dao 2.0, the rapid progress being made in the development of language models makes it clear that one will soon pass the Turing test.

"The way to artificial general intelligence is big models and big computer [sic]," BAAI chair Dr. Hongjiang Zhang added in a statement. "What we are building is a power plant for the future of AI. With mega data, mega computing power, and mega models, we can transform data to fuel the AI applications of the future."

"The technology could be used as an intelligent assistant — less Siri, more Janet from The Good Place — to help people navigate life admin tasks, such as applying for a new driver's license," notes Politico Europe, adding, however, that there are darker possibilities. "Critics also say that these language models could be used as a surveillance tool, for example, to analyze social media for critical comments, or as powerful disinformation and propaganda machines."

NEXT: Study Finds Vapers Not More Likely To Get COVID-19

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. It can process 10 time the number of slave laborers!

    1. Can it set up and operate Bitcoin mining nodes at 10X the rate of your average human?

      Asking for a… you don’t need to know why I’m asking.

      1. It’s probably the bee’s knees at modeling protein folding and antigen-binding…

        1. Making money online more than 15$ just by doing simple work from home. I have received $18376 last month. Its an easy and simple job to do and its earnings are much better than regular office job and even a little child canNN do this and earns money. Everybody must try this job by just use the info
          on this page…..VISIT HERE

    2. His job was located in Lower Manhattan, and after work he usually took the subway back up to Harlem, where most of New York City’s black population lived. Occasionally,.more detail………….CLICK HERE.

  2. Ask it where the Coronavirus originated.

    1. Chinese robot: No comprende.

    2. It will spit this out:
      June 17. Spanish virologists find traces of C-19 in Barcelona wastewater from March 2019: “The levels of SARS-CoV-2 were low but were positive,” said research leader Albert Bosch.
      June 20. French virologists find SARS-CoV-2 was spreading in France in December 2019. “Early community spreading changes our knowledge of the COVID-19 epidemic”.
      Nov. 16. Italian Researchers find Coronavirus in Italy from September, 2019. “Traces of SARS-Cov-2 have been found in samples of waste water taken in Milan and Turin between September 2019 and March 2020”.

      Nov. 30. American researchers find high levels of Covid-19 antibodies in archived Red Cross blood samples throughout the USA from Dec. 2019. Serologic testing of U.S. blood donations to identify SARS-CoV-2-reactive antibodies: December 2019-January 2020.

      Feb. 25. “Analyzing Covid genomes using k-mer natural vector method, we conclude that the virus likely already existed in France, India, Netherlands, England, and USA before the Wuhan outbreak”.

      Mar. 30. Joint WHO-China Report on Jan.-Feb. China visit: “Researchers reviewed 76,000 clinical records from October to November 2019, in which were 92 possible cases of Covid-19. 67 of those had no signs of infection based on antibody tests done a year later, and all 92 were ultimately ruled out based on the clinical criteria for Covid-19”.

      June 1. WHO sends 30 Italian 2019 biological samples to Rotterdam’s Erasmus University laboratory for re-testing.

      June 8. Erasmus University concluded that the Italian samples from 2019. “Are very similar to what (Italy’s National Cancer Institute) discovered, despite some small differences. The combined results made a convincing case that the coronavirus or a similar virus was circulating in Italy months before the country’s first officially recorded case”.

    3. That’s an important scientific question.

      I’m genuinely curious why you think it’s an important question. Say it came from a lab instead of an animal market. What is actionable about that to you? What do you want to do as a result?

      1. “…I’m genuinely curious why you think it’s an important question…”

        Bullshit. To unpack this, the first pile of lefty bullshit is shitstain’s claim to be ‘genuinely curious’ about any matter of fact; bullshit!
        Shitstain is never ‘genuinely curious” about anything; shitstain’s curiosity extends to whether he can offer offer bullshit supporting his proggy agenda by distorting what’s on the table.
        —————————————
        Now, let’s look at shitstian’s second attempt at mendacity:

        “… Say it came from a lab instead of an animal market. What is actionable about that to you? What do you want to do as a result?”

        This pile of shit assumes that the question is difficult to answer, or hopes that those reading it are stupid enough to accept the bullshit on offer.
        Hey, lefty asshole! How about we hold Fauci accountable for funding (with US taxpayer money) what is looking more and more likely to be the source of those “millions of US deaths” you keep lying about?
        There are some who post here who might be more of a fucking lying piece of lefty shit than you, but they’re hard to find.
        Please ask you mom why she didn’t abort you and let us know why she was so pathetically stupid.

        1. So your answer is to burn the witch Fauci.

          Boy are you going to be surprised when you learn that Tucker Carlson isn’t the most judicious purveyor of truth in America today.

          1. Tony – I never agree with you, but here you’ve outwitted the regular “thinkers” here at reason. Well done. As for all you fucks who couldn’t be bothered with what the man made virus would mean for mankind, perhaps you should sit in silence for ten minutes to catch up with the rest of us.

      2. Tony has thought this shit out to it’s unfeasible conclusion. If indeed it was man made, then that means the conspiracy involves our media, some repubs, some dems, the entire IC community, by extension the us military, the WHO, the CDC, the CCP, the UK, possibly Russia, as well as numerous activist and charitable orgs like Soros and Gates, most of our governors, and of course, trump and obama. With a conspiracy that large, is it even really a conspiracy, or simply an end around the constitution – an unethical societal change from the ground up?

        Indeed, if it’s man made, there’s absolutely nothing we can do but play along with the ruse, lest we bring about certain death sooner.

  3. Wu Dao 2.0 (Chinese for enlightenment) is ten times larger than GPT-3, using 1.75 trillion parameters to simulate conversational speech, write poems, understand pictures, and even generate recipes.

    “With four parameters I can fit an elephant, and with five I can make him wiggle his trunk.” – John von Neumann

    “Simplicity and elegance are unpopular because they require hard work and discipline to achieve and education to be appreciated.” – Edsger Dijkstra

    “There is no greatness where there is no simplicity, goodness and truth.” – Leo Tolstoy

    1. Ten trillion parameters means it won’t fit in the typical computer. Then again, nothing is local anymore, Alexa and Siri don’t process anything on your local device they upload what you say for Amazon/Apple to decipher.

      The problem is not the Chinese. I don’t trust them further than I could throw them. But I don’t trust our own government either, regardless of which team color is in office. And this is a tool governments at all levels can use to manipulate and control their population. This is also why freedom is crucial, and a pox to all the parties that insist on imposing cultural norms.

      1. Ten trillion parameters means it won’t fit in the typical computer. Then again, nothing is local anymore, Alexa and Siri don’t process anything on your local device they upload what you say for Amazon/Apple to decipher.

        I was going for a multi-layered approach. If there’s anything that climate science has taught us in the last few decades it’s that more parameters, no matter how fast/powerfully executed, do not always a more accurate prediction make.

        And your deflection and claim are between contradictory and biased non-sequitur (or at least seemingly anyway). The US government doesn’t fund OpenAI the way the Chinese government funds the Beijing Academy of Artificial Intelligence. I absolutely don’t trust either government with either of the AIs and absolutely don’t blame the Chinese people directly for any AI (ab)use, but in the context of the article or AIs presented, the Chinese government is clearly the 500 lb. gorilla.

      2. I agree with what you are saying in essence, but Apple literally announced this week that Siri now works locally on newer devices (a feature which Alexa already had on some Android devices).

        1. Turn on airplane mode. Try to use Siri. Doesn’t work.

      3. Some serious people think that there’s an irreconcilable difference between Chinese and Western culture on the issue of individualism, going all the way back to Confucius.

        It’s worth wondering whether favoring freedom, however you define that, is to impose cultural norms on people who might be just fine under a collectivist authoritarian culture.

        I’m certainly fond of individualism and liberty as defined in the West, but a little voice inside my head wonders if I’m not actually most interested in protecting my cultural assumptions against threats.

        But as of now I think it’s still true that authoritarianism is unstable and that people grow weary of living their lives for the state or church. Otherwise they wouldn’t need so many psychologically damaging incentives to keep things in place.

        1. Some lying lefty shits keep posting as if anything they post is not a pile of lefty shit

          1. I’ve never seen a blown gasket keep going so long.

          2. Devi, say something of substance or stfu

        2. It’s worth wondering whether favoring freedom, however you define that, is to impose cultural norms on people who might be just fine under a collectivist authoritarian culture.

          Go fuck yourself

    2. Speaking of elephants. Are those Chinese elephants adorable or what? I can’t stop watching them.

  4. Yeah but can it find pictures of tank man for you.

    1. help people navigate life admin tasks

      Monster Chinese computing algorithm vs. the administrative state. Place your bets!

  5. If not Wu Dao 2.0, the rapid progress being made in the development of language models makes it clear that one will soon pass the Turing test.

    Q: “Is there a God?”

    A: “There is *now*.”

    More seriously, it’ll be interesting to see what kinds of, um, “mistakes” have to be engineered into (learned by) the machine in order for it to pass for a human.

    1. Turing test has already been passed by much lesser software. The problem is that it’s not Alan Turing administering the test. Most people today CAN”T distinguish between a good computer and a real person.

    2. I expect it will make mistakes. After all, it’s based on the inputs, which is full of them. Watch Watson on Jeopardy, and some of its mistakes.

  6. the cylons will speak mandarin.

    1. If they all look like Cylon #6, I’d start learning Mandarin.

      1. oh yes. Starbuck and Boomer were cylons too

        1. Dirk Benedict and Herbert Jefferson? I must have missed that episode.

          1. Its very convoluted.

    2. If birth rates continue the current slide, eventually the only beings left will *be* cylons, though it should be a while yet.

  7. “There are darker possibilities. “Critics also say that these language models could be used as a surveillance tool, for example, to analyze social media for critical comments, or as powerful disinformation and propaganda machines.”

    I don’t suppose Fauci’s NIAID funded this research. The NIH just funds research at Chinese virology labs that work for the People’s Liberation Army–not data labs. Fund Chinese AI? That’s too dangerous, too crazy!

    1. Broken record.

      1. Says the squawking bird.

      2. “Broken record.”

        Lying pile of lefty shit.

    2. Tripling down on the failed thing. At some point it becomes uncomfortable for everyone around you.

      1. Lying pile of lefty shit.

      2. Every question needs an answer and every story needs an archetypal villain. Nothing occurs by chance. Fauci fills the role of Lex Luthor in the Covid saga which ties into the stolen election plot. It all fits together.

  8. CT-3 is K Street’s first choice for climate propaganda, pro and con, but wait till Chinese & Australian coal flacks put Wu Dao to work as a comment bot.

    https://vvattsupwiththat.blogspot.com/2021/06/wii.html

  9. Is it though? Chinese researchers have a worse reputation for.lying about technical achievements than even American ones.

    They ‘cloned a person’s, after all.

    Then there’s their ‘train on a virtual track’ – known to the rest of us as ‘a bus’.

    And, of course, the bus that would drive over traffic.

    1. This actually. I’ll wait a month or two and see if this pans out to be real or not.

  10. Stay at home mom Kelly Richards from New York after resigning from her full time job managed to average from $6000-$8000 a month from freelancing at home…

    This is how she done it… http://jive4.com/

  11. Kinda makes you think Frank Herbert was on to something…how long before we are going to need a Butlerian Jihad to save us from the machines.

    1. You think you’re kidding?

      AI by Microsoft or Apple terrifies me, AI by the Chineese has “The End of Humanity” written on the front label as a prime feature.

      1. It is probably for the best.

        In some theories that has already happened many times anyway and we are just living through another iteration.

        1. “This has all happened before, and it will all happen again”?

          Yeah, I sort of know that one, and I’ve been known to quote it, but the geological and biochemical records sort of stand against it.

          Ima stick with “Thou shalt not build a machine in the likeness of a human mind”.

  12. “The technology could be used as an intelligent assistant . . . to help people navigate life admin tasks, such as applying for a new driver’s license.”

    Or it could just drive the car.

    1. It could also give itself the traffic ticket. Much more efficient.

    2. Who TF are these people that can’t navigate through applying for a new DL themselves? Do we even want them driving?

  13. The next step would be to ask it “what is the answer to the ultimate question of life, the universe and everything?”

    I expect an answer in about 7 1/2 million years.

    1. Meh, you want to know the answer to the question: “What is the meaning of life”?

      It’s sort of like an eigenvalue, where the answer is pretty much the question. (Apologies to the mathematicians here.)…. The meaning of life is life.

    2. The answer to your question is simple: the meaning of life is to avoid answering the meaning of life.

      You’re welcome.

      1. The answer is 42

        Everyone knows that, Gorsuch used to begin his law lectures with that question.

        Some people here are not up on nerd culture.

  14. Well when “wokness” and “equity” drive research dollars the US is going down the tubes. You have school districts destroy gifted programs because they might support the narrative the wokes are terrified of…culture matters and some cultures after a few generations have much higher performers…the US was great because it was based on ability/talent/intelligence in science and technology…but now…we hold certain groups back..but as long as we have the right percentages the elites will not feel guilty and have the narrative they want.

  15. that there are darker possibilities

    I think the word you are looking for is certainties not possibilities. It’s ludicrous to think that the most valuable thing this can do is to replicate the most boring imaginable tasks that human would prefer not to do. Don’t need much AI for that.

    No. What’s valuable is for the AI to do stuff that humans are repelled at because we, usually, have some basic ethical standards of expected behavior imprinted on us really early. Course with no constraints, the most unethical sociopaths are the ones who will figure out what that AI will do.

  16. Of course, there is nothing nefarious about this at all. The Chinese developing AI technology? I’m sure the speech processor handles speech in a totally unbiased way and the dataset is totally not curated at all.

    1. And you’ll still use it.

    2. There will come a day when an AI will identify some cell phone user using unapproved language which will be interpreted as criticizing the government. The cell phone will then be detonated a second afterwards.

  17. The story is mostly nonsense. This is not a model of how the mechanics of how humans actually produce language. Humans use an entirely set of principles to create language. See Chomsky’s Universal Grammar.

    What we have here is an algorithm that uses a massive database as a training set to create an impressive statistical approximation of human language. But it in no way ‘thinks’ about and generates language like an actual human, as the lead researcher is trying to claim.

    1. So quantity does not automatically equal quality?

  18. “”The technology could be used as an intelligent assistant — less Siri, more Janet from The Good Place — to help people navigate life admin tasks, such as applying for a new driver’s license,” notes Politico Europe, adding, however, that there are darker possibilities. “Critics also say that these language models could be used as a surveillance tool, for example, to analyze social media for critical comments, or as powerful disinformation and propaganda machines.”

    That latter seems much more likely.

  19. Well, it’s taught ‘reason’ how to kiss chinese patootie. Does it also negotiate Dementia Joe’s bribes with Hunter?

  20. Best Android Development Company in jaipur
    (https://www.searchnplays.com)

Please to post comments