360 Intelligence brain-Visual Big Model released, Zhou Hongyi: multimodal is the only way for the development of large models.

Thanks to CTOnews.com netizens Xiao Zhan cut, rain and snow on the way, the west window of the past, Mr. Aviation's clues delivery! CTOnews.com May 31 news, 360 held a visual model and AI hardware new product conference, launched 360 wisdom brain visual model and a number of AI hardware new products, and announced that 360 Smart Life officially entered the SMB market.

Zhou Hongyi said at the press conference that all software, App, websites and all industries are worth reshaping with big models, while smart hardware is hardware-based App. From the perspective of the development trend of large models, multimode is the only way for the development of large models, and the most important change of GPT-4 is that it has multimodal processing ability. Zhou Hongyi predicts that the combination of multimodal large models and the Internet of things will become the next tuyere.

360 integrates hundreds of billions of parameter "360 brain" big model on the basis of visual perception ability, carries on cleaning training based on billion-level Internet picture and text data, and fine-tunes millions-level industry data of security scene, finally creates a visual and multimodal big model-360 intelligent brain-visual big model. The large visual model is an important component of the 360 brain, which enables it to understand pictures, videos and sounds in the future.

Zhou Hongyi said: "at present, the capability of the big model is mainly reflected in the software layer. When the big model is connected to the intelligent hardware, it will make the ability of the big model move from the digital world to the physical world."