The Academy of Information and Communications Technology releases the highest level of credible AI evaluation results: 360 Intelligent Brain has excellent capabilities

Recently, the 2023 Nanjing Artificial Intelligence Industry Development Conference officially opened. At the meeting, the Academy of Information and Communications Technology released the trustworthy AI assessment results. 360 (601360.SH, hereinafter referred to as "360") Group's self-developed cognitive general-purpose large model 360 Intelligent Brain passed two levels of model development and model capability at level 4. Module verification has become one of the few large models in the industry that has passed Level 4 verification, which is also the highest capability level currently achieved by domestic large models.

Previously, the Academy of Information and Communications Technology, together with more than 100 units of the Large Model Working Group, jointly developed a series of standards for the "Large-scale Pre-training Model Technology and Application Verification Methods", including model development, model capabilities, model operation, model application and trustworthiness Five parts are required. At present, two standards for model development and model capabilities have been released, including more than 70 capability items in several major dimensions such as data construction, model training, model deployment, functional richness, performance superiority, and service maturity, which can fully verify all aspects of large models. technical skills.

It is understood that the 360 ​​Intelligent Brain model participated in the verification of compliance with two standards of "model development" and "model capability", and both passed the verification at level 4, with most of the items submitted for inspection reaching 5 points. Full marks.

360 Intelligent Brain is the self-developed cognitive general large model of 360 Group. Relying on the key advantages of large computing power, big data, and engineering accumulated by 360 for more than 360 years, it integrates 360GPT large model, 360CV large model, and 360 multi-modal model. Large model technical capabilities include ten core capabilities and hundreds of subdivided functions such as generative creation, multi-round dialogue, and logical reasoning. Currently, 360 Intelligent Brain has been iterated to version 4.0 and is connected to all 360 products, reshaping a new paradigm of human-machine collaboration. From July to August this year, 360 Intelligent Brain improved its overall performance by 14.55%, its COT capability by 69%, further supporting longer text input of 50,000 words, and increasing the length of multi-round conversations by 18%.

Currently, based on the 360 ​​Group’s accumulation of large model technology and practice, in the large model rankings jointly released by Shanghai Jiao Tong University, Tsinghua University and the University of Edinburgh, 360 Intelligent Brain’s ability outperforms GPT-4, and is ranked first in Chinese. It ranks in the leading position in domestic large model capabilities in multiple third-party evaluations such as the General Large Model Benchmark Evaluation.

Disclaimer: This content is for reference only and does not constitute investment advice

advertise

The above is the detailed content of The Academy of Information and Communications Technology releases the highest level of credible AI evaluation results: 360 Intelligent Brain has excellent capabilities. For more information, please follow other related articles on the PHP Chinese website!