In an exciting breakthrough in the Chinese language AI market, Baidu, the renowned search engine provider, has unveiled its latest model, Ernie 3.5. A large Chinese language model that has been claimed to surpass its counterparts, ChatGPT3 and even GPT4 by Baidu, in several Chinese language capabilities and uses cases. This move by Baidu has put it in the front of the AI race within the country and has concreted its position as a leader in this space in the industry. The claims made by Baidu have been backed by a detailed evaluation conducted by China Science Daily, evaluating the Ernie 3.5 on datasets like AGIeval and C-EVAL, which are standard datasets to test such engines. This noteworthy achievement has inspired other domestic key players such as Alibaba Group, Tencent Holdings, and others to develop AI models that stir competition in this rapidly changing field.
Ernie 3.5 has claimed a remarkable improvement in training and evaluation efficiencies, leading to an optimal inference time than the previous version and less resource intensive. Many different sources have reported almost 50% improvement from the earlier models that Baidu has put forward. This creation is again important because efficiency and cost-effectiveness are highly sought-after characteristics in the AI market. Baidu has a well-laid-out vision for Ernie 3.5. They are planning to launch an external plugin support feature, which will enhance the capability of an existing model to work on specific tasks such as summarization and Question-Answer. It is important to mention that ChatGPT introduced plugin support earlier this year, correctly identifying the growing demand for specialized AI in the market.
Baidu’s claims have been strengthened by an esteemed science journal’s two comprehensive tests. In the first test, Ernie 3.5 outperformed ChatGPT in standard admission and qualification exams, commonly used for college or law school entrance evaluations. Its superior performance in the Chinese language demonstrates its advanced linguistic abilities. The second test evaluated Ernie 3.5 on over 13,000 multiple-choice questions, covering a wide range of subjects. Once again, Ernie 3.5 achieved higher scores than its competitors, cementing its position as the leading Chinese language AI model.
With Huawei all set and locked to unveil its highly anticipated Pangu AI model upgrade on July 7, the competition in the Chinese language AI market is expected to be more fierce and would reach new heights. The intense rivalry among Chinese companies shows their grit and commitment to delivering state-of-the-art AI models. This language processing and understanding advancements potentially revolutionize various sectors, including education, customer service, and content creation.
It will be interesting to see how Chinese companies would trim down their resources and find scalable alternatives amid all the sanctions laid out on them by the US and how they would achieve and maintain domination amid growing competition.