Create 2024 a look at Baidu’s Ernie models and AI tools
Baidu, inc. showcases AI models and toolkits at Create 2024, aimed at making AI technology accessible.
At the Create 2024 Baidu AI developer conference, Baidu, inc a Chinese technology company, introduced AI models and development tools, designed to make AI technology available to all, regardless of technical background.
Robin Li, co-founder, chairman, and CEO of Baidu, discussed the potential of natural language in AI development in his keynote speech. "In the future, natural language will become the new programming language, allowing anyone who can speak to become a developer," Li stated, envisioning a future where creativity and AI intersect.
Baidu's ERNIE, a key component in their AI development, has demonstrated significant usage and utility. ERNIE Bot, a conversational AI based on the ERNIE model, has reached 200 million users since its release on March 16, 2023. The ERNIE API processes 200 million queries each day, reflecting its broad adoption.
Additionally, Baidu Comate, an AI coding assistant, has been involved in generating 27% of new code within Baidu over the past year and serves over 10,000 companies, including Ximalaya, iSoftStone, and Shanghai Mitsubishi Elevator, with a code adoption rate of 46%.
Patrecia Meliana
Baidu's Qianfan platform, part of Baidu AI Cloud, has been used by over 85,000 enterprise clients to develop 190,000 AI applications. The efficiency of ERNIE's training has improved, being 5.1 times more efficient than the previous year.
The model's inference capabilities have also increased, with a 105-fold increase in queries per second and a reduction in inference costs to just 1% of its March 2023 figure. These metrics underscore Baidu's ongoing efforts to enhance AI functionality and reduce operational costs.
Further, Baidu introduced new additions to the ERNIE family at Create 2024. The ERNIE 4.0 Tool, accessible via ERNIE Bot, features a code interpreter that enables users to manage and analyze data through natural language, aiming to streamline data handling and decision-making processes.
The ERNIE model family now includes lighter versions such as ERNIE Speed, ERNIE Lite, and ERNIE Tiny, designed to offer more adaptable and efficient AI solutions by being compressed and distilled from the original model, addressing diverse AI application needs.
Three AI development trends impacting future AI adoption
At the Create 2024 conference, Robin Li, CEO of Baidu, highlighted three significant AI development trends in his keynote.
First, the Mixture of Experts (MoE) model, which facilitates the combination of various AI models to tackle complex problems, signaling a shift towards more versatile AI-native applications.
Second, he emphasized the advantages of smaller AI models, which are designed to be cost-efficient and responsive, making them particularly useful in scenarios where large models are impractical.
Finally, Li discussed the evolution of AI agents that are designed to perform tasks involving understanding, planning, reflection, and self-improvement, pushing the boundaries of AI capabilities.
During the event, Baidu also unveiled the Apollo VTA, an advanced vision foundation model tailored for autonomous driving, trained with over 100 million kilometers of data from urban driving tests. This model integrates several key capabilities such as detection, tracking, understanding, and map building, enhancing the safety and adaptability of autonomous driving technology.
In addition, the launch of the second Baidu ERNIE Cup Innovation Challenge was announced, a competition aimed at fostering the development of groundbreaking AI models and applications, with substantial rewards for the winners.
The Create 2024 event showcased Baidu's commitment to democratizing AI technology. Through a series of over 20 industry forums and more than 30 open AI development classes, along with an expansive AI interactive experience area and music festival, Baidu provided attendees with a comprehensive and engaging overview of the current and future state of AI technology.
ContentGrow is a managed talent network for brands and publishers to work with high-quality freelance writers and journalists worldwide. Sign up to get started or book a discovery call to learn more.
