On April 26, the iFLYTEK Spark large model V3.5 (hereinafter referred to as "iFLYTEK Spark") was launched in the spring. Facing the pain points of users' efficient and accurate knowledge acquisition, iFLYTEK released the industry's first large model of long text, long graphics and long voices, which can not only quickly learn massive texts, graphic materials, and conference recordings from various information sources, but also give professional and accurate answers in various industry scenarios.
iFLYTEK further upgraded the Spark voice model, debuted multi-emotional super-anthropomorphic synthesis, has the ability to express emotions, and launched a one-sentence voice reproduction function to make technology more warm.
At the same time, for enterprise application scenarios, iFLYTEK has launched the Xinghuo intelligent twins platform to help enterprises solve the last mile problem of large model landing.
Continuing to use technological progress to solve real needs, iFLYTEK Xinghuo is also being recognized by more and more users. According to Qimai data, the number of downloads of the iFLYTEK Xinghuo APP on Android has exceeded 96 million, ranking first among the domestic tool general model APP.
The first large model of long text, long graphics and text, and long voice is launched to help efficient knowledge acquisition
Why does iFLYTEK want to make a large model with long text, long graphics and long voice?
Through the iFLYTEK Xinghuo APP, it can be seen that the peak of user use is not on weekends, but at 9:30 a.m. and 3:30 p.m. on weekdays. This means that most users use iFLYTEK Xinghuo to solve work-related rigid needs. Efficient knowledge acquisition is a high concern for both users and developers.
iFLYTEK analysis found that in the process of knowledge acquisition and learning, the information that the majority of users can get is often not only ready-made long texts, but also the content of newspapers and books, PPT content of various seminars, board books on the teacher's blackboard, students' notes, as well as various meeting recordings, interviews, various online press conferences, training and education videos, etc., can these texts, pictures, voices, etc. be uploaded to iFLYTEK Xinghuo to quickly acquire knowledge?
To this end, iFLYTEK has launched the first large model that supports long text, long graphics and long voices, to solve the needs of users for obtaining multi-source information in real scenarios.
After the new upgrade of the iFLYTEK Xinghuo long text function, it has the capabilities of long document information extraction, long document knowledge Q&A, long document induction, long document text generation, etc., and has generally reached 97% of the level of the latest long text version of GPT-4 Turbo in April, while the overall level of iFLYTEK Xinghuo long text has surpassed GPT-4 Turbo in the knowledge question and answer tasks in multiple vertical fields such as banking, insurance, automobiles, and electricity.
The implementation of the long text function needs to solve the problem of efficient information processing: in the face of millions or even tens of millions of words, the long text large model consumes a lot of computing resources.
In order to solve the problem of application efficiency and accuracy of large models, Liu Qingfeng said that based on the ability of iFLYTEK Xinghuo V3.5 to understand, learn and answer long texts, iFLYTEK has carried out important model pruning and distillation, so as to launch the industry's best performance of 13 billion parameters of the large model, in the case of effect loss of only 3%, so that Xinghuo has achieved great efficiency improvement in document upload and analysis processing, the first response time of knowledge questions and answers, and text generation. The test shows that under the condition of ensuring the effect of long text, whether it is 10K, 64K, 128K token, or longer text, the performance of the Xinghuo large model is the best in the industry.
Facing complex graphic and text scenarios, iFLYTEK has launched the Spark graphic recognition large model for the first time on the basis of years of international first-class technology accumulation in the graphic recognition and formula recognition competitions.
Compared with the limitations of traditional small model line-by-line text recognition, the Xinghuo graphic recognition large model has three advantages: 1) It can directly process very complex layout analysis, which has covered 31 typical scenarios, such as books, academic papers, patents, newspapers, posters, PPT, etc., and can automatically identify and mark out 18 different layout elements, such as headers, footers, titles, paragraphs, tables, formulas, seals, etc. Handwriting, etc.; 2) Integrate the context semantics of the text for text recognition, which makes the recognition more accurate;3) In-depth optimization for professional fields such as education, finance, medical care, and scientific research, which can automatically realize professional symbol recognition in more fields.
According to the internationally published authoritative English test set, the image and text recognition effect of iFLYTEK Xinghuo exceeds that of Microsoft and Google. From the perspective of typical application scenarios, it is in the leading position in the industry in terms of recognition effect in scientific research, finance, and enterprise product technical documents.
In addition, in the face of the demand for efficient access to a wide range of audio and video information, iFLYTEK has also launched the long voice function, which combines the world's leading speech recognition and translation technology to realize one-click reading of meeting recordings and learning videos, and realize efficient knowledge acquisition in audio and video scenarios.
Released the contract assistant and upgraded the AI learning machine to solve the real needs with technological progress
The upgrade of iFLYTEK Xinghuo's long text, long graphics and text, and long voice capabilities further promote the implementation of large models in various scenarios. Liu Qingfeng focused on the application of iFLYTEK Xinghuo in bidding, contract, education and other scenarios.
In the bidding scenario, with iFLYTEK Xinghuo's leading text comprehension, logical reasoning and mathematical capabilities, iFLYTEK and the National Energy Materials Corporation have cooperated in an intelligent unmanned review system in the enterprise procurement scenario, which has been recommended as a typical case on the website of the State-owned Assets Supervision and Administration Commission. According to reports, more than 57,000 orders have been reviewed in the National Energy Group, with an accuracy rate of 97%. This time, the ability to superimpose the upgraded long text and long graphics can make the bid evaluation more convenient, efficient and accurate.
In our daily life, we often encounter a variety of contracts when buying and selling goods, decorating houses or buying car insurance, what should we do if we don't understand the risks? iFLYTEK has launched the Spark Contract Assistant, which can conduct risk review, contract comparison, summary and contract generation of our contracts, quickly identify potential risk loopholes, and become a "legal assistant" in your pocket.
In the education scenario, iFLYTEK has further upgraded the iFLYTEK AI learning machine product, which not only makes the correction of essays and science more accurate, but also makes the intelligent supplementary learning more targeted and efficient; At the same time, it improves children's willingness and ability to take the initiative to ask questions.
The smart blackboard has also been upgraded again, equipped with long text and long voice capabilities, so that the efficiency of actual transcription can be improved, and the ability to sort out chapters can be improved. The second is the Spark teacher assistant, after integrating the ability of long texts, the high-quality teaching and auxiliary content can be integrated, and the teacher can directly integrate the content of the teaching assistant in the process of lesson preparation, so as to further enrich the resources of lesson preparation and improve the efficiency of lesson preparation.
In addition, Xinghuo scientific research assistant has been applied in the Chinese Academy of Sciences, Sanya Yazhou Bay Science and Technology City, Beijing University of Posts and Telecommunications, Harbin Institute of Technology and other institutions and universities. With the upgrade of multimodal capabilities, iFLYTEK Xinghuo Scientific Research Assistant has also further improved the effect of paper Q&A, review generation, experimental interpretation, etc., making the analyzed academic materials more abundant, and further empowering the scientific research work of universities and research institutes.
It can "resonate emotionally" and "reproduce the sound of a sentence"
In the era of the Internet of Everything, more realistic AI voice interaction is needed. At the launch of iFLYTEK Xinghuo V3.5 at the beginning of the year, iFLYTEK launched the super anthropomorphic dialogue function, and the voice of AI is more natural and realistic, with an anthropomorphism of 83%, which is widely welcomed by users. Whether it is speech intelligibility, fluency or expressiveness, the effect exceeds that of OpenAI and Microsoft.
This time, iFLYTEK released multi-emotional super-anthropomorphic synthesis, which further improved the perceptibility of emotional expression, and the perceptibility of emotional expressions such as happiness, apology, comfort, coquettishness, and confusion reached more than 85%, and the AI voice was more vivid and real.
The 2024 model of Haobo HT is the first in the industry to be equipped with iFLYTEK's super-anthropomorphic synthesis technology, and has been officially launched globally on April 25.
In addition to hyper-anthropomorphic dialogue, iFLYTEK has also launched the "one-sentence voice replication" function, which allows you to customize your AI assistant voice in one sentence. For example, imitating children's voices, reading books and newspapers to grandparents every day, and imitating our voices to tell children stories when we are on a business trip. This feature can make the world a warmer place.
Liu Qingfeng said that iFLYTEK has always been an industry leader in personalized speech synthesis, and has now advanced to one-sentence voice replication. At that time, iFLYTEK AI needed to go to Taiwan to record Lin Chiling's voice for a week, and later it took a day to imitate Guo Degang's voice, and then it took 5 minutes of recording, and now it can be imitated in one sentence. You can experience it on the iFLYTEK Xinghuo APP.
Released the Xinghuo intelligent twins platform to inject new quality productivity into the enterprise
Since its release on January 30 this year, iFLYTEK Xinghuo V3.5, as the first large-scale model of nationwide computing power training, has been widely welcomed by partners and developers in various industries. In the past three months, iFLYTEK has added 550,000 real-name certified developers, more than half of whom are from enterprises.
For enterprises, how to efficiently acquire and learn knowledge is also a pain point, and iFLYTEK gave the answer to the intelligent twin, and launched a new intelligent twin platform for enterprise scenarios.
The process of building an agent mainly involves task understanding, external information source connection, internal IT system connection, and in-depth integration of private domain knowledge, and finally outputs the answer according to the execution results of each task, so that the construction of the agent can be finally completed.
Liu Qingfeng said that on the Xunfei Xinghuo intelligent twins platform, for the user's input, first of all, based on the very smart base ability of the Xunfei Xinghuo large model, it will automatically realize the accurate understanding and task planning of the user's input. Secondly, after analyzing the relevant tasks and corresponding tools, iFLYTEK Xinghuo has also built a system of external information sources including weather, flights, enterprise checks, etc. Finally, through the private domain knowledge integration mechanism, the intelligent twins platform can easily realize the integration of the enterprise's industry and the enterprise's private domain knowledge, so as to achieve more accurate professional understanding and knowledge Q&A.
In addition, the Xinghuo intelligent twins platform can also realize the creation of new agents and the collaboration of multiple agents by dragging and dropping. The Spark agent platform can reach the last mile of large-scale model application enterprises in an agile manner.
According to Liu Qingfeng, iFLYTEK will release the iFLYTEK Spark model V4.0 on June 27 to further liberate productivity and release imagination.
At this year's National People's Congress and the National People's Congress, the "artificial intelligence +" action was carried out to accelerate the development of new quality productivity for the first time in the "Government Work Report". The knowledge management revolution brought about by large models is being staged, and both enterprises and individuals can stand on the shoulders of artificial intelligence and achieve new comparative advantages.