This Program Can Fix Itself Using Regenerative AI

Fixing a malfunctioning program can be tedious, which is why developer “BioBootloader” devised a solution using AI. According to Hackaday, the developer created a program, aptly named “Wolverine” that grants Python programs the capability of “regenerative healing,” similar to the Marvel superhero.

BioBootloader showcased a demo video on Twitter and stated that once the Wolverine program is used, it can detect errors in the script and edit them using GPT-4 while providing explanations of what caused the issue. Even if multiple bugs exist, Wolverine repeatedly runs until every problem is resolved.

Finally, the modified code is saved in the original Python file.

The demonstration video features a split-screen display with Python code on the left and Wolverine results on the right in a terminal. He uploads a custom-made calculator script that contains several intentionally inserted flaws before executing it.

Today I used GPT-4 to make "Wolverine" – it gives your python scripts regenerative healing abilities!

Run your scripts with it and when they crash, GPT-4 edits them and explains what went wrong. Even if you have many bugs it'll repeatedly rerun until everything is fixed pic.twitter.com/gN0X7pA2M2

— Scott Swingle (@bio_bootloader) March 18, 2023

OpenAI developed GPT-4, a multimodal AI language model that was launched in March and is accessible to ChatGPT Plus users and beta testers through an API. It leverages its understanding of billions of web documents, books, and websites to complete text-processing tasks such as programming, language translation, and composition.

The source code for Wolverine is available on GitHub, and the developer believes that the same technique could be applied to other programming languages.

However, to use Wolverine, users must have an OpenAI API key for GPT-3.5 or GPT-4, and charges may apply. At present, the GPT-3.5 API is open to all OpenAI account holders, but access to GPT-4 is still limited by a waitlist.

Although it’s currently just a rudimentary model, the capabilities of Wolverine offer a glimpse of a potential future where software programs could potentially rectify their own issues, including unforeseen ones that may surface after release.

However, the wider implications, safety considerations, and feasibility of this approach have yet to be fully investigated.