Artificial Intelligence

Wolverine, an artificial intelligence system that corrects Python code

In recent weeks, the use or abuse of artificial intelligence in many applications has been discussed, even regulations have been requested to control such use, beyond opinions, these systems have made the work of many professionals more efficient and productive, especially when detecting errors, as in the case of programming.

Developed by a programmer called "BioBootloader", Wolverine is a script that uses OpenAI's GPT-4 model that analyzes Python code with errors and generates possible solutions based on a statistical analysis of previous examples.

BioBootloader demonstrated how Wolverine interacts with GPT-4 to identify and resolve bugs in a Python script on its Twitter profile, going through several stages. The AI tool executes the script, then detects faults, then queries GPT-4 to understand the issues, and then makes the necessary fixes. Finally, after fixing the bugs, Wolverine updates the original Python file with the changes.

While it is true, that the prototype presented is a bit rudimentary, Wolverine showed the potential of AI-driven tools to automatically debug and repair code in the future. BioBootloader believes it could be adapted to other programming languages. Although to use Wolverine, developers need an OpenAI API token for GPT-3.5 or GPT-4, which can generate fees.

The key to Wolverine's functionality lies in meticulously written prompts for GPT-4, a language model programmed in natural language. While AI coding capabilities have limitations, the potential of automatic retrieval code using AI models such as GPT-4 raises the potential and its application in the future of programming.

Every AI system has its risks and limitations and more so when we apply it in programming, one of them is the reliability of AI-generated solutions, not to mention the risk of malicious attacks and poisoning of training datasets.

Ideally, each programmer should have the ability to analyze and solve problems on their own and not develop a dependency on the use of AI-based tools, maintain a critical approach and not neglect their own knowledge and experience in problem solving.

The pros are diverse, including speeding up the code debugging process, facilitating the identification of errors, saving resources and effort in software development, among others. On the other hand, the use of these tools can generate possible problems of reliability and security in the generated solutions, risk of excessive dependence on the AI system by the programmers and of course, limitations in the ability of the AI to understand the context and the developer's intentions, and more.

These tools are not designed to replace human programmers or substitute for the expertise they possess; on the contrary, the responsible use of Wolverine and other similar tools can become a valuable ally in easing their journey.

To learn more, visit:

https://github.com/biobootloader/wolverine

https://hackaday.com/2023/04/09/wolverine-gives-your-python-scripts-the-ability-to-self-heal/

 

02 de Mayo, 2023