Shulou(Shulou.com)11/24 Report--

CTOnews.com, August 2, the University of Neimergen in the Netherlands recently released a research report, pointing out that companies such as Meta and OpenAI tend to mislead users when using the term "open source", and some of the big language models labeled "open source" are actually not open source.

The report specifically mentions Meta's Llama 2 model and OpenAI's GPT / codex model, saying that the code to train these large language models is not open to the public.

The researchers say that the lack of open source large language models in the current AI community is becoming increasingly prominent. The researchers called on companies to release more open source LLM so that researchers and developers can access the code and improve the performance of these models.

OpenAI's ChatGPT model is the most "mysterious" and does not meet the open source standard, while Meta's Llama 2 claims to be "open source", but the actual transparency is only slightly better than the ChatGPT model, and the degree of transparency in the open source ranking is still very low.

According to the study, these companies are mainly for the following reasons:

Want to protect their intellectual property rights

Want to control how the large language model is used

However, researchers and developers have natural weaknesses in the face of open source large language models, and may not be able to study the results of the paper and improve the performance of existing large language models.

Meta's position is ironic because the company proudly says that Llama 2 can be used for research. Even so, the company seems to have locked some parts of the project out of developers and researchers.

CTOnews.com attaches the link address of the paper here, which can be read in depth by interested users.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.