Code Interpreter is only just available for ChatGPT, but the combination has already made a positive impression.

Professor Ethan Mollick from Wharton University, Pennsylvania tested the alpha version of Code Interpreter for ChatGPT. He was so excited that he wrote an extensive blog about it.

Simply new

The public beta of the new tool has just been announced, but the alpha version has been circulating among select users for some time. Professor Mollick is one of those lucky ones. Take that last word literally, the AI ​​researcher (among others) is very excited about the system.

superlatives

With Code Interpreter, ChatGPT can now also perform more complex mathematical calculations, write code (in Python) and generate charts. Mollick calls the combination the most useful and interesting AI model he has ever used. The blog post is so positive you’d almost look for the disclaimer.sponsored content‘.

The scientist claims that the AI ​​model completed tasks and calculations in seconds that took him weeks during his PhD studies. In addition, according to Mollick, the system makes fewer mistakes than one would expect from a human analyst. He was also impressed with how ChatGPT could reason with Code Interpreter, almost like a human interlocutor.

Mollick had the model perform a variety of different tasks in a matter of months: interpreting superhero powers, showing emotions through code, or even analyzing the Magic The Gathering card game. Furthermore, he also tested the visual talents of this ChatGPT version and unsurprisingly these were more than satisfactory as well.

nuance

The good news show has some caveats. For example, the AI ​​model’s hallucinations may have been reduced, but not entirely gone. They will also have much less impact on the code itself or on the data, according to Mollick.

He also adds that Code Interpreter is not an alternative to human talent. On the contrary, he sees AI models as a tool to relieve people of the more boring tasks so that they can focus on more interesting things. Ironically, unlike Sam Altman, CEO of ChatGPT’s OpenAI, who has previously raised concerns about this.