On June 9, at the 24th anniversary celebration of iFlytek, iFLYTEK Xunhuo Cognitive Model V1.5 was officially released. Wu Xiaoru, President of HKUST iFlytek, and Liu Cong, Dean of the Research Institute, conducted on-site real-world testing of the Xinghuo large model. After one month, the Xinghuo large model has achieved open knowledge questions and answers in terms of comprehensive capabilities. upgrade in one direction. In addition, the Xinghuo APP was released at the same time, and iFlytek further promoted the application of the Xinghuo cognitive model in education, medical and other fields, and opened the iFLYTEK Xinghuo development interface.
“Today we released the latest results of the Xinghuo large model again as scheduled, because all our technology platforms are completely independently developed, and we are very confident that the technology at each stage is controllable.” Director of HKUST Xunfei Director Liu Qingfeng said that Xunfei should not only pay tribute to and catch up with OpenAI, but also make independent innovation at the source. Currently, it has explored more potential paths and frontier intersections of general artificial intelligence in brain-like intelligence, large neural network models, and game intelligence. research opportunities.
Xunfei Xunhuo Cognitive Model V1.5 Live Demo/Photo provided by the organizer
Knowledge question and answer ability is improved by 24% compared with the original, which can solve more complex problems
“What are the composition questions for the 2023 national college entrance examination paper, and analyze the meaning it conveys.” “The college entrance examination has just ended, and your child is about to start a new starting point. Please write a heart-warming letter to your child.” University of Science and Technology News Wu Xiaoru, President of Feifei, demonstrated the response of Xunfei Xinghuo Cognitive Model V1.5 to such problems.
In Wu Xiaoru’s view, by combining the language comprehension ability and general expression ability of the large model with the search plug-in, the Spark large model solves the problems that new knowledge is difficult to update and fact-based questions and answers are easy to “make a difference”. % improvement, “The real power of the big model is not to make the generated content exactly the same as the direct search, but to use natural language understanding as the core, combined with plug-in capabilities such as class search, to solve more complex problems.”
In addition, he said that with the blessing of long-chain thinking reasoning and mathematical logic reasoning, “Xunfei Spark” has ushered in an upgrade in logic and mathematics. “There are three people on a boat, but there are two fathers and two sons, what’s going on?” To this, Xunfei Xinghuo replied, “These three people are grandfather, father and grandson,” and gave the analysis results.
In terms of multiple rounds of dialogue, when you enter “I want to write a story about a little rabbit traveling and exploring Huangshan.” After a few seconds, the Xinghuo Cognitive Model will generate a wonderful little story. Of course, further plot requirements can be put forward, such as ” The little rabbit meets a new friend Pikachu.” Through multiple rounds of dialogue and interaction on the spot, an imaginative story quickly came to life on paper.
In addition, Xunfei Xinghuo APP and applets are also officially released today. It includes pure voice dialogue, multi-modal input, multi-terminal support, multi-functional assistant and other functions. It can realize voice communication at any time, supports graphic recognition, mathematical formula recognition, and presents it in various forms on various terminals.
At the press conference, HKUST Xunfei upgraded the AI spoken language assistant released on May 6 to the Xinghuo Language Companion APP. With the blessing of the large model, the Xinghuo Language Companion APP can conduct open dialogues and situational exchanges, and practice together like a spoken English teacher. Perform real-time oral error correction. Users can also use video conversations to communicate face-to-face with the virtual human teacher.
In the medical field, for After being discharged from the hospital, the condition suddenly changed and I didn’t know what to do.Liu Cong demonstrated the various service capabilities of the post-diagnosis rehabilitation management platform: According to the automatic analysis of the patient’s health profile, the platform can intelligently generate a personalized rehabilitation plan for the patient, including focus attention, medication guidance, rehabilitation exercise, discharge follow-up, health knowledge, Patient consultation, etc., and urge patients to follow the plan.
Independent research and development of key modules Each stage of technology is controllable Open spark development interface
“On the Xunfei artificial intelligence open platform, there are currently more than 4 million developer teams. This time, the ecology brought by general artificial intelligence will be more prosperous. We will work with all ecological partners to promote the arrival of the era of large-scale models.” Liu Qingfeng explain. At the scene, he announced the opening of the development interface of the Spark Cognitive Large Model, fully opening up the seven-dimensional capabilities and 200+ assistant capabilities to developers, and supporting rapid integration of multi-terminal access and privatization deployment. Build ecology.
From the release of the Xinghuo Cognition Large Model on May 6, to the upgrade to V1.5 on June 9, according to the key milestones in the development of the large model previously announced by HKUST Xunfei, it will benchmark ChatGPT on October 24 this year: Chinese beyond, English equivalent.
“The reason why we can set such a clear goal is that we not only have a good technical accumulation and a well-established team, but also every key module is completely independently developed, and the software and hardware platforms are run on domestically-made reliable platforms. Yes. We have a complete innovation system, which is the key to iFLYTEK being able to move forward step by step and give clear milestones and time nodes. IFLYTEK is also the only one in China that currently gives a clear time for the development of large models. We hope to work with industry partners to make the dream of self-controllable and continuous evolution more real.” Liu Qingfeng said.
Liu Qingfeng said: “We are not only paying tribute to and catching up with OpenAI, but also making independent innovations at the source.” and other fields to explore more potential paths and opportunities for cutting-edge cross-research. “The new era of general artificial intelligence will be a great historical process, and it will not be achieved overnight. Therefore, we must not only have short-term ambitions for tackling key problems, but also have long-term perseverance. This is the industrial logic and scientific research development that we should follow. There are rules, so that we can truly win in the long-distance race, truly solve the social needs, and build a better world.”
Luo Yidan, Shell Finance reporter of the Beijing News
Editor Song Yuting
Proofreading Liu Baoqing Return to Sohu to see more
Editor:
Disclaimer: The opinions of this article represent only the author himself. Sohu is an information release platform, and Sohu only provides information storage space services.