China’s leading search engine provider, Baidu Inc, has claimed that its latest LLM, styled after ChatGPT, can beat the OpenAI model on certain metrics. This statement comes from a report by Reuters which claims that their LLM, Ernie 3.5 can outperform ChatGPT’s GPT-4.
In a statement, the Beijing-based company said, “ChatGPT in comprehensive ability scores” and outperformed “GPT-4 in several Chinese capabilities.” Citing tests that the state newspaper, China Science Daily, ran. The company used datasets including AGIEval and C-Evail. These are two benchmarks used to evaluate the performance of AI models.
Baidu is the first major Chinese tech company to unveil its ChatGPT rival with the AI Ernie Bot in March of this year. Much like GPT 3 to GPT 4, Ernie 3.5 is built on its predecessor Ernie 3.0. Since then, testing was through invite only.
According to Baidu. Ernie 3.5 has better training and inference efficiency with an eye on later iterations becoming both faster and cheaper. Like ChatGPT, it will also have support for external plugins. The goal of which of course will be to provide users with better precision.
This comes after Alibaba Group’s AI model earlier this month and Tencent’s own model as well. China is hoping to usher in an AI revolution as the government is working on creating regularity frameworks for AI. In a meeting with Microsoft co-founder Bill Gates, Chinese President Xi Jinping made clear that his government wants to be welcoming to American AI firms.
Across the world, the AI race is heating up. With firms pushing the envelope of what their models can do on an almost weekly basis. In a report by Goldman Sachs earlier this year. Generative AI is expected to both boost GDP growth and affect hundreds of jobs in the near term.
As generative AI continues to take the world by storm. Now is a great time to get the least information on Gen AI, and go beyond the hype. The Generative AI Summit, coming up July 20th, is a completely-free virtual event devoted to generative AI. Get your free ticket now.