The material in this article comes from the Internet, if it is inconsistent with the actual situation or there is infringement, please contact to delete.
In the latest announcement of the Ali Dharma Academy, they officially launched a ChatGPT-type large-scale language model called "Tongyi Qianwen", which is a highly anticipated move, and as early as the beginning of this month, news about Ali's plan to launch a similar ChatGPT was already broken. However, this official announcement was ahead of schedule, which made everyone interested in Ali's technical prowess.
The main function of Tongyi Qianwen is focused on text generation, and like ChatGPT, it can answer various questions and generate text. In addition to the dialogue function, it also has a built-in "treasure bag" that can generate various types of copy, which increases productivity.
First, let's test the conversational skills of Tongyi Qianqian, including language, context understanding, code, and mathematics. From the perspective of language ability, Tongyi Qianwen has performed well in Chinese, and can even continue to write and create, showing its strong creativity.
Not only that, but Tongyi Qianwen has also demonstrated ability in different language environments, including Japanese and English, and although slightly blunt in terms of English humor, it performs quite well in most cases.
Next, we tested the contextual understanding ability of Tongyi Qianqian, which included the ability to multi-round dialogue. Its ability to accurately extract the above information and use it to answer the next round of questions demonstrates its potential for multiple rounds of dialogue.
In terms of code capabilities, Tongyi Qianwen can perform simple four operations, demonstrating its programming potential. However, it is important to note that it has some problems in converting explanations into annotations, which require further improvement.
Finally, we tested the math ability of Tongyi Qianwen, which performed well in basic calculations and some math problems, but still had room for improvement for advanced math problems.
In addition to dialogue skills, we also tested the scene ability of Tongyi Qianqian, including generating recipes, rainbow farts, and writing love letters for free. In these respects, it demonstrates excellent creative and text-generating capabilities.
Regarding the technical details of Tongyi Qianwen, Ali Dharma Academy did not disclose much information, but according to its answers, the training data came from rich multilingual text data, including Chinese, English, Japanese, French, Spanish and other languages. This shows that Tongyi Qianwen is a multilingual large-language model with networking capabilities, but no access to real-time data.
Overall, despite its excellent performance in conversation and text generation, there is still room for improvement compared to the current industry benchmark ChatGPT. Ali said that they are constantly iterating on the model, so more improvements and enhancements can be expected in the future. It is worth mentioning that Ali has a leading edge in algorithms and computing power, which provides strong support for their competition in the field of large models.
In the era of large models, Chinese enterprises are accelerating their competition, and the demand for state-produced large models is increasing. Although large models bring productivity improvements, they also come with some challenges, such as blocking problems and computing power bottlenecks. Therefore, technology self-research still has important value, and Alibaba's efforts in this regard will play an important role in future competition.
To sum up, Tongyi Qianwen is an important initiative of Ali in the field of large models, although there is still room for improvement, but it shows Ali's technical strength and potential in the field of artificial intelligence. As competition for large models intensifies, we can expect more innovation and breakthroughs.
The above content and materials are derived from the Internet, and the author of this article does not intend to target or allude to any real country, political system, organization, race, individual. The above does not mean that the author of this article endorses the laws, rules, opinions, behaviors in the article and is responsible for the authenticity of the relevant information. The author of this article is not responsible for any problems arising above or related to the above and does not assume any direct or indirect legal liability. Due to the rules of the platform, you will only be recognized as a fan if you have more interaction with me. If you like my article, you can click "Follow", and you can receive the article push as soon as you become an iron fan. This article is only published in today's headlines, please do not move.