it is no secret that china is working hard to surpass the rest of the world in artificial intelligence research and now for the first time they might have actually accomplished that what many people consider to be the biggest and most realistic sounding ai model gpt-3 has just been beaten almost eight-fold by chinas new artificial intelligence model which they’ve named wu dao 2 0 which china has unveiled at their giant annual ai convention which is being visited by some of the leading ai researchers from all over the world welcome to todays episode of ai news in.
this episode i will show you what this new ai has already accomplished how its performing compared to american models and finally what this means for the near future of ai research the beijing academy of artificial intelligence is an excellent venue for displaying cutting-edge technology and skill in a variety of fields the nonprofit institution promotes advances in ai theories methods systems and applications by encouraging scientists to confront issues furthermore the beijing academy of.
artificial intelligence places a special emphasis on long-term ai research the baai is an excellent venue for displaying technology and skill in a variety of fields the nonprofit institution promotes advances in ai theories methods systems and applications by encouraging scientists to confront issues furthermore the baai has a distinct focus on long-term ai technology development in china artificial intelligence is a major deal over 70 000 individuals have registered for the event and many more are tuning in to see the baai presentations thereafter many of the techniques algorithms systems and applications on display are brand new.
wu doubt 2 0 on the other hand was a huge hit at the baai outperforming openais gpt-3 in almost every manner language is defined by the encyclopedia britannica as a system of customary spoken physical or written symbols by which human beings express themselves as members of a social group and participants in its culture from this concept we may deduce that language is an essential component of.
human communication language helps us to not only exchange ideas emotions and feelings with one another but also to establish and build communities and empires to put it another way language is what makes us human professor gareth gaskell of the university of yorks department of psychology estimates that the average 20 year old knows between 27 000 and 52 000 words by the time they reach 60 that number.
has risen to between 35 000 and 56 000 as a result when we employ words in speech our brain must make a split second judgment about which words to use and in what order in this sense the brain functions as a processor that can do several tasks at once according to language experts each word we know is represented by a distinct processing unit whose sole purpose is to determine the chance of incoming speech matching that word the processing unit that represents a.
word in the brain is analogous to a pattern of activity across a group of neurons in the brain as a result when we hear the beginning of a word thousands of such units become active due to the large number of probable matches the average person can understand up to eight syllables per second the objective is to access the word stored meaning rather than to recognize it before a word is completely recognized the brain searches for a variety of probable interpretations listeners begin to record various.
alternative interpretations such as captain or capital when they hear a word fragment like cap before the complete word appears according to studies language like most things driven by artificial intelligence in the 21st century is changing to take on new forms and meanings the notion of linguistic models has recently risen to prominence in ai in essence language models analyze text data to predict word likelihood.
this means that language models analyze data using statistical and probabilistic approaches to determine the likelihood of a particular word sequence natural language processing applications such as those that create text as output frequently employ language models machine translation and question answering are two examples openai presented their version of an autoregressive language model dubbed gpt-3 in march of the same year which employs deep learning to generate human-like writing the gptn series third generation language model has a 175 billion machine learning parameter capacity gpt-3 can create news items that human.
assessors have trouble differentiating from pieces authored by humans according to a study published by openai researchers the language model can also be trained to create 100 pages of material for only a few cents in energy expenses according to the researchers microsoft leased exclusive use of the language model and its underlying code since the gpt-3 was judged so robust and powerful however just a year later another language model surpassed both gpt-3 and turing nlg in terms of inventiveness and.
innovation at the baai this model dubbed wu dao 2 0 was on display tang jia baai research academic vice president and singua university professor headed the team that created wu dao 2 0 which has been termed chinas first domestic super scale intelligent model system over 100 ai researchers from peking university singua university renmin university of china chinese academy of sciences and other institutions backed.
him up wu doubt 2 0 is the follow-up to wu dao 1 0 which was released earlier this year by the baai wu dao 2 0 is chinas answer to gpt-3 and its larger and better to begin with wu dow 2 0 unlike gpt 3 develops abilities in both chinese and english as a result of analyzing 4 9 terabytes of pictures and writings woo dao 2 0 has also formed partnerships with 22 companies including xiaomi and the video app kuai show the chinese model was trained on 1.
75 trillion parameters approximately 10 times as many as the 175 billion parameters used by gpt-3 wu dao 2 0 can also produce classic chinese style poetry answer questions write essays and create picture text in addition according to baai this language model met or exceeded state-of-the-art standards on nine criteria finally wazeebing the worlds first chinese virtual student was introduced by wu dao 2.
0 wa has the ability to study create drawings and write poems she will be able to study coding in the future wu dao 2 xeros learning capacity is in striking contrast to gpt-3s other information about how and what wu dao 2 0 was trained or still unknown making it difficult to directly compare it to gpt3 the new language model on the other hand is a testimony to chinas ai aspirations and outstanding research initiatives there’s no denying that ai innovation will accelerate in the next years and.
many of these breakthroughs will benefit a variety of different sectors dr kaifu lee one of the ai luminaries and investors who helped establish at least seven ai-powered unicorns recently spoke at the hong kong science and technology park on the potential of transformers and fine-tuning huge pre-trained models like wu dao 2 0 these models may be tailored for a wide range of sectors and applications including education finance law.
entertainment and most significantly healthcare and scientific research the use of transformers in biomedical research will almost certainly lead to new discoveries that will benefit people everywhere and despite the trade conflicts i truly hope that the governments would consider cooperating on biological research to achieve a gi they count on gpt like multimodal and multitasking models wu dao 2 0 like gpt-3 before it is undeniably a significant step toward agi however it is uncertain how much closer we will get some scientists believe that in order to achieve a gi we will require hybrid models others support embodied ai.
completely rejecting established bodyless paradigms like neural networks no one knows which road is the correct one even if bigger pre-trained models are now the logical trend we may be missing the forest for the trees and we may wind up reaching a less ambitious ceiling in the near future the only thing that is certain is that if the world must suffer from environmental degradation detrimental.
prejudices or enormous economic expenses not even achieving agi is worth it so what is your opinion on this new artificial intelligence model from china do you believe that reaching agi by simply scaling the parameter count of natural language models is possible or do you think its more complicated than that also are you glad that america isnt the sole player in the field of ai anymore please tell us your opinion in the comment section below i would love to hear what you have to say about it.
thank you for watching ai news we consistently report on the newest technologies that are shaping the future of our world wed appreciate you subscribing and watching our other videos see you around and take care.
A team of Chinese scientists have now created an artificial intelligence that is equal to or even surpasses the accomplishments of OpenAI’s 2019 release. It’s called “Wu Dao AI” and it has the ability to learn new things, speak multiple languages, and does research in medicine and other areas.
The developers of this AI model have been working relentlessly for 7 years to get to where they are now and it shows. The AI can now not only understand human speech but also has a very large vocabulary and can even carry on a coherent conversation with a human being. This model is so smart that even after 7 years of development, it is still learning and getting better. It can now answer questions that were not answered correctly by the previous version of itself.
So what about the other players?
America will not just sit there while the rest of the world takes over the technological lead in AI. They’ve got a fighting chance at winning this race too. Especially since they’re already so far ahead. They’ll likely answer with even bigger AI models very soon, and then the race to AI supremacy will continue with the rate of innovation increasing exponentially.
When the dust settles, I believe the winner of the AI war will be either USA, China or Russia. The USA will still be the world’s superpower but they may not have the luxury of being the worlds only AI superpower.