首页 News 正文

"The big model basically eliminates illusion." On November 12, Robin Lee, Baidu founder, expressed his opinion, and thought that this was the biggest change in the AI industry in the past 24 months.
At the 2024 Baidu World Conference, Robin Lee delivered a speech with the theme of "Application Coming", and released two AI technologies enabling applications: retrieval enhanced literacy map technology (iRAG) and code free tool "Seconds Da".
Robin Lee said that just like websites in the PC era and we media accounts in the mobile era, agents will become new carriers of content, services and information in the AI native era.
Talking about basic models: Large models have basically solved the illusion problem
"In the past 24 months, is the global big model craze a new technological revolution or a new round of foam? As the flag bearer of China's artificial intelligence, I think we are qualified to answer this question." Robin Lee took the curve chart of the daily average adjustment of Wenxin big model as an example - as of the beginning of November, the daily average adjustment of Baidu Wenxin big model exceeded 1.5 billion, an increase of 7.5 times compared with the 200 million disclosed in May, and an increase of about 30 times compared with the 50 million disclosed for the first time a year ago.
"This growth rate is higher than expected", Robin Lee said that this shows that AI is really in demand. He believes that the growth curve of the daily adjustment amount of the Wenxin big model represents the explosive application of big models in China in the past two years.
Robin Lee said that when ERNIE Bot was released last March, Baidu's big model featured enhanced knowledge and retrieval. Over time, enhanced retrieval has gradually become a consensus in the industry. He stated that the significance of enhanced retrieval technology lies in eliminating illusions in large models. To develop applications based on big models, eliminating illusions is necessary. If this model always talks nonsense seriously, no one will believe you, and there will be no applications
Robin Lee believes that the biggest change for the industry in the past 24 months is that the big model has basically eliminated illusion, and the accuracy of the big model's answers to questions has been greatly improved, which has made AI usable and trustworthy from "serious nonsense". We know that a large model is a probabilistic model that generates content with uncertainty. After using RAG (Retrieval Enhanced Generations) technology, the large model will use the retrieved information to guide the generation of text or answers, greatly improving the quality and accuracy of the content
However, Robin Lee also mentioned that RAG at the text level has been well done, but the combination of multimodal content such as images and RAG is not enough, and hallucinations are still common, so the large-scale application of multimodal large models has not yet emerged.
Based on this background, Baidu decided at the beginning of this year to solve the illusion problem of image generation.
At the 2024 World Congress of Baidu, Baidu announced the launch of iRAG (image based RAG), a retrieval enhanced cultural map technology. Robin Lee introduced that the commercial value of iRAG is reflected in: no illusion, super reality, no cost and no wait.
Eliminating the illusion of large models is also the foundation for the explosion of AI applications. In Robin Lee's opinion, today, the basic big model capability is ready, and the star shining moment of AI application is coming.
Talking about Applications: Intelligent agents are the most mainstream form of AI applications
Where do AI applications come from and go? Robin Lee also mentioned two AI application directions in his speech: agent and industrial application.
"Agents are the most mainstream form of AI applications and will soon usher in an explosion point", Robin Lee said that today, all the top technology companies in the world pay close attention to agents, but few like Baidu regard agents as the most important strategic direction.
Robin Lee believes that being an agent is much like being a website in the PC era and an We Media account in the mobile era. The difference is that intelligent agents are more like humans and more intelligent. "Intelligent agents may become a new carrier of content, information, and services in the AI native era.
He cited the example of company type intelligent agents. In the traditional PC website model, companies can only statically display company introductions and product parameters, but lack proactive recommendations, timely responses, and one-on-one service capabilities; And the company's intelligent agent can recommend corresponding products based on customer needs, and in terms of service, it can also respond to needs more directly and quickly, greatly improving the efficiency of interactive marketing. In the future, the company's official intelligent agent is likely to replace the official website and become the most direct interface for consumers.
"Agents are the most mainstream form of AI applications, and will soon usher in its explosion point." In Robin Lee's view, agents have low threshold and high ceiling, which can not only let everyone get started, but also make complex and powerful applications. At present, the Wenxin intelligent agent platform has attracted 150000 enterprises and 800000 developers.
On the spot, Robin Lee released 100 industrial applications based on the big model, covering manufacturing, energy, transportation, government affairs, finance, automobile, education, the Internet and many other industries. He said that Baidu is not going to launch a "super application", but to constantly help more people and enterprises create millions of "super useful" applications.
Based on this, Robin Lee announced the launch of the code free tool "Seconds Da". We do have the conditions to enable people who cannot understand a single line of code to have the ability of programmers, and the ability to quickly and cost effectively turn any idea into reality
"A person can complete the construction of a system through natural language interaction. In addition to the invitation system shown above, he can also do various applications in any scenario, and the complexity of applications will continue to increase with the improvement of our technology." Robin Lee concluded that this means that each person can command multiple agents to complete tasks together. "As long as you have ideas, you can achieve what you want, and we will usher in an unprecedented era in which you can make money only by ideas."
In different historical periods of human information technology transformation, the appearance of applications has also been different: in the PC era, it was individual software and websites; In the mobile era, there are individual apps and followed accounts; In the AI era, Robin Lee believes that the main form of application is agent. He stated that with the exponential leap of big model technology and capabilities, natural language has become the most important programming language of this era, and each of us can create an AI application or intelligent agent that belongs to ourselves and others.
I come from a software engineer background, and there is a saying abroad that 'software devours the world'. But I believe that this world should not be swallowed up, but should be created. In the era of AI, applications create the world Robin Lee finally said.
CandyLake.com 系信息发布平台,仅提供信息存储空间服务。
声明:该文观点仅代表作者本人,本文不代表CandyLake.com立场,且不构成建议,请谨慎对待。
您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

因醉鞭名马幌 注册会员
  • 粉丝

    0

  • 关注

    0

  • 主题

    43