"AI is igniting a creative revolution, and developing applications in the future is as simple as shooting a short video. As long as you can speak, you can become a developer, and in the future, everyone will be a developer and creator."
On the 16th, Robin Lee, the founder, chairman and CEO of Baidu, officially released the tool version of Wenxin Big Model 4.0. He said that since the release of ERNIE Bot more than a year ago, the number of users has exceeded 200 million.
"Compared with a year ago, the algorithm training efficiency of Wenxin Large Model has increased to 5.1 times, the weekly average training efficiency has reached 98.8%, the reasoning performance has improved 105 times, and the reasoning cost has dropped to 1%," Robin Lee said.
Robin Lee further explained, that is to say, the customer used to call 10000 times a day, but now it can call 1 million times at the same cost. Some groups may not be excited by a 99% cost reduction. But whether it's a company or a developer, once used, the most important thing to consider is the effectiveness and cost.
Robin Lee also shared Baidu's specific idea of developing AI native applications based on the big model, pointing out that MoE, small models and agents are three directions worth paying attention to. "This is what we gained by stepping on numerous pits and paying high tuition fees based on our practice in the past year."
Robin Lee stressed that the smaller size model cut out through dimension reduction of Wenxin 4.0 is significantly better than the model directly transferred from open-source. Under the same size, the cost is significantly lower