At the 2024 Baidu World Conference, Baidu founder Robin Li presented a series of引人注目 advancements in the AI field. He noted that the most significant breakthrough in large model technology over the past two years has been the基本解决 of the "hallucination" problem, marking a qualitative leap in the accuracy of AI responses, transforming it from "seriously making things up" to a reliable intelligent assistant.
During the conference, Robin Li highlighted Baidu's latest retrieval-augmented text-to-image technology, iRAG. This innovative technology integrates the vast image resources accumulated by Baidu's search platform and advanced foundational model capabilities to generate ultra-high fidelity images. Robin Li stated that iRAG's overall performance far exceeds traditional text-to-image systems, successfully eliminating the traces of machine-generated images.
The outstanding advantage of iRAG technology lies in its commercial practicality. It not only solves the hallucination problem in AI-generated images but also achieves ultra-realistic effects, while being cost-effective and immediately available, providing new possibilities for commercial applications.
In terms of data, Baidu's Wenxin large model has shown astonishing growth. By early November, its daily call volume had exceeded 1.5 billion times, a 30-fold increase from the 50 million times announced a year ago, highlighting the widespread application and rapid普及 of large model technology across various fields.
These advancements not only demonstrate Baidu's technological prowess in the AI field but also reflect that large model technology is entering a more mature and practical new stage. With the improvement in accuracy and the expansion of application scenarios, AI technology is bringing unprecedented transformation opportunities to industries across the board.