Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

The research shows that the GPT-4 model has the ability of self-correction, which is expected to promote the further commercialization of AI code.

2025-02-25 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

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

CTOnews.com, July 5 (Xinhua)-- researchers at the Massachusetts Institute of Technology (MIT) and Microsoft have found that the GPT-4 model has excellent code self-correction ability, while GPT-3.5 does not have this feature. The paper has been published in ArXiv.

ArXiv, the source of ▲ images, has emerged a number of AI models specifically for code, but at present, most of them only assist developers in writing code, such as Microsoft Copilot assistants familiar to CTOnews.com friends. These AI models can only generate code fragments, so they can not completely replace human developers.

Through the study of GPT-4, the researchers said that at present, it is actually possible to "self-correct the model" to make the model "reflect on its own shortcomings" to improve the length of code fragments and improve the accuracy of the output results.

▲ diagram source ArXiv

Fu Yao, a researcher at the University of Edinburgh at ArXiv, a source of ▲ images, said that only GPT-4 has the ability to improve itself, while the weaker GPT-3.5 does not. This finding suggests that large models may have a new ability to make AI self-correct through a series of user feedback, resulting in satisfactory results for users. This self-correcting ability may only exist in sufficiently mature AI models.

After self-correction, 71% of the output code of the GPT-4 model meets the requirements set by the researchers, while after correcting the code generated by GPT-3.5 using GPT-4, the pass rate of this batch of code also reaches 54%.

The researchers say that GPT-4 's self-correction approach can now be applied to business, and it can still generate some benefits after deducting a series of error correction redundancy costs. This paper can always reflect the future trend of the industry to some extent, so it is expected that a number of GPT-4-based code generators will emerge in the future.

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.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report