Science and technology giant iFLYTEK recently held a grand press conference in Hefei, Anhui Province, and officially released its latest iFLYTEK Spark Cognitive Model, which has aroused widespread attention and heated discussions. According to Liu Qingfeng, chairman of HKUST Xunfei, and Liu Cong, director of the research institute, the large model has surpassed ChatGPT in three major capabilities, including text generation, knowledge question and answer, and mathematical ability, and has become a leader in the industry.
In addition, Liu Qingfeng also announced the time points for the three upgrades of the large model within the year, and strives to catch up with ChatGPT as a whole by the end of October. Since the release of ChatGPT last year, cognitive large-scale model technology has continued to evolve step by step, and artificial intelligence has attracted widespread attention around the world, setting off a new upsurge.
The advent of iFLYTEK’s Xunfei Xinghuo Cognitive Model is undoubtedly a milestone in the field of artificial intelligence, and it deserves close attention from industry insiders and technology enthusiasts.
Liu Qingfeng revealed that on December 15, 2022, HKUST Xunfei officially launched a special research on the "1+N" cognitive large model. Among them, "1" refers to the large model of general cognitive intelligence, and "N" refers to the implementation of the large model in various fields such as education, office, automobile, and human-computer interaction.
Liu Qingfeng believes that when paying tribute to and learning from OpenAI, we must quickly catch up and strive to surpass, and this requires a scientific and systematic evaluation system.
HKUST Xunfei took the lead in designing a general-purpose cognitive large-scale model evaluation system through the National Key Laboratory of Cognitive Intelligence, and jointly discussed with the Chinese Academy of Sciences Artificial Intelligence Industry-University-Research Innovation Alliance and the Yangtze River Delta Artificial Intelligence Industry Chain Alliance to form 481 detailed models covering 7 categories. Sub-task type.
Liu Qingfeng
Demonstration of multiple capabilities
At the release site, Liu Cong conducted a number of demonstrations, showing the seven categories of capabilities of the Xinghuo Cognitive Big Model in the evaluation system of the General Cognitive Intelligence Big Model. Judging from the on-site demonstration, the Xinghuo cognitive model can basically complete various tasks smoothly and perform well.
Liu Cong also asked the big model of Xinghuo: "Why are you called Xinghuo?" The big model of Xinghuo replied that the name "Xinghuo" has three meanings: first, the emergence of the intelligence of the current cognitive big model makes iFlytek see more clearly The long-term goal and realization path of general artificial intelligence AGI; secondly, as early as 2001 at the "half-time meeting" of iFlytek of HKUST, the entrepreneurial team put forward the slogan of "burning the brightest torch, either the first to start a prairie fire, or the first to extinguish it" ; Finally, "Xinghuo" comes from the "Heart" constellation in the twenty-eight constellations of Chinese traditional culture. In traditional Chinese culture, it is regarded as a symbol of auspiciousness, representing strength, courage and perseverance. These three meanings also show the pursuit and firm belief of HKUST Xunfei in the field of artificial intelligence.
In the demonstration, Liu Cong demonstrated the multiple capabilities of the Xinghuo Cognitive Big Model, including text generation, language understanding, logical reasoning, mathematical capabilities, coding capabilities, and multimodal capabilities. In terms of text generation, the Xinghuo model can write welcome speeches for conferences, tell stories, write emails, and write promotional texts in English, etc., which is an excellent performance. In terms of language understanding, the Xinghuo large model has demonstrated multi-level cross-lingual language understanding capabilities, including grammar checking and element extraction.
Logical reasoning is in the same vein as mathematical ability. When asked "What is the result of adding 1 to 100?", the Spark model quickly calculated the answer as 5050. When asked "there are three kinds of flowers in the flower bed, a total of 88 flowers, among which the number of rose flowers is 4 times that of chrysanthemums, and the number of peonies is 5 times that of chrysanthemums, 2 less. The answer was quickly calculated as 43 peonies.
At the demonstration site, Liu Cong also demonstrated the ability to code, modifying code, writing code and other abilities. In terms of multimodal capabilities, the Xinghuo Big Model can generate speech, pictures, videos and other materials based on key words, and can also generate virtual people for video display. The multimodal input and expression capabilities continue to be enriched.
General artificial intelligence now "dawn"
At the results conference, Liu Qingfeng pointed out that the technological step and rapid evolution of the cognitive large model has attracted worldwide attention and heated discussions. He said that the new generation of cognitive large models learn and train through various types of data. The data comes from materials including voice, pictures, images, etc., which has become the "dawn" of general artificial intelligence, and the spark of general artificial intelligence is emerging .
Liu Qingfeng also pointed out that the current general artificial intelligence includes seven dimensions, including text generation, language understanding, knowledge question answering, logical reasoning, coding ability, mathematical ability and multimodal ability, and the improvement of these seven aspects will also affect the production and lifestyle. Bringing huge disruption, the "intelligent emergence" of cognitive large models will bring new opportunities to solve the needs of human beings. In the future, in the information industry and digital economy, everyone may start a business, and creativity and inspiration will become industrial benefits, which in turn will bring about major changes in the entire industrial ecology. This also indicates that artificial intelligence technology will play an increasingly important role in the future development.
HKUST Xunfei recently released its latest cognitive large-scale model technology, which is a conversational general intelligence emerging and has broad application prospects. HKUST Xunfei is a member of the AI "national team" and has accumulated more than ten years in the field of cognitive intelligence. From the Xunfei Ultra Brain Project in 2014 to last year’s Ultra Brain 2030 Project, iFlytek has been focusing on AI’s ability to understand knowledge, learn well, and evolve, and is committed to bringing robots into every home.
With the accumulation of various capabilities, iFlytek launched the "1+N" large model technology research in 2022, where "1" refers to the base platform for the research and development of general cognitive intelligence large model algorithms and efficient training programs, and "N" It is the application of cognitive intelligence large-scale model technology in education, medical care, human-computer interaction, office, translation and other industries.
HKUST Xunfei Xinghuo large model will be the first to launch in-depth applications in industries such as education, office, medical care, and automobiles. At the same time, it will be the first to launch products such as AI learning machines and Xunfei Hearing. The vertical industry applications released today focus on the four industries of education, office, automobile and digital employees. It is worth noting that HKUST Xunfei has not released Xunfei’s medical-specific large model, but will continue to embed years of accumulated medical capabilities into general artificial intelligence. The medical-specific large model is expected to be released during the 1024 conference.
As the AI "national team", iFLYTEK has accumulated profound technology, has been deeply involved in education, medical care, government affairs and other fields for many years, and has a deep understanding of downstream needs. It has over 50TB of industry corpus data and active applications and scenarios with over 1 billion user interactions per day. and data advantages are significant. The cognitive big model will empower the company at three levels of human-computer interaction, content generation, and personalized service, and continuously strengthen product and technical advantages. Under the wave of AI, iFLYTEK is expected to usher in a new round of development opportunities.