ChatGPT’s code interpreter democratizes data analytics and could be the most useful AI application ever, said Wharton’s Ethan Morrick. Getty
Today, for $20, anyone can get a personal AI assistant that is proficient at analyzing data.
code interpreter— New OpenAI tools for ChatGPT The tool, which can run code, manipulate uploaded files, analyze data, create charts, edit files, and perform calculations, will be released on Friday for all subscribers to the $20/month ChatGPT Plus service. I was.
For Ethan Morrick, an early adopter of AI and professor of management at the Wharton School of the University of Pennsylvania, this may be the most useful and interesting application of the technology to date. It may also be the strongest case yet for a future in which artificial intelligence permeates society. A valuable partner for advanced knowledge work.
“What took weeks to master in a PhD. AI did it in seconds and generally had fewer errors than human analysts expected. Human oversight is still important, but right now I wouldn’t do a data project without a code interpreter,” Morrick said in a blog post published Friday.
With the ability to write code and access to large memory (users can upload files up to 100MB), it is a new “to the strengths” of large language models, the AI technology behind ChatGPT. Mollick writes that the functionality will be integrated into ChatGPT. , was granted early access to Code Interpreter as a researcher.
“Specifically, it gives AI a general-purpose toolbox for solving problems,” Morrick wrote.
In other words, the code interpreter takes AI assistants beyond just generating text responses. This is of value not only for academics and programmers, but also for anyone trying to do research. As for the future of work, Morrick writes that it is “a harbinger of things to come.”
It’s also less “hallucinating,” he added, because it uses Python, a versatile programming language used for building software and analyzing data. “This fills in some gaps in the language model,” Morrick said. luck, Because the output is not based on text only. The code is being run through Python and will generate error messages if the code is wrong.
For example, if you feed a set of data about superheroes into a code interpreter, the code interpreter turns the tedious task of cleaning and merging the data into a “very sophisticated one” in a “constant” effort to ensure accuracy. in a way” (but it should always be doubled). – confirmed, he added). Users can also switch back and forth between the code interpreters and ask for various changes and improvements when doing data visualizations.
That way, proper analysis can be performed or recommended, Morrick writes. In this case, predictive modeling was done that could predict what powers a hero would have based on other factors.
“While AI can do many other analyzes (after all, AI is just writing Python code), I have often been impressed by AI’s ability to choose sensible analytical approaches.” writes Morrick.
According to Morrick, Code Interpreter’s most salient feature is how it can reason about data “in a very human way.” For example, from our analysis of data on superheroes, we observed that their powers are often visually prominent because they derive from the medium of comics.
From creating cute charts to performing regression analysis, Code Interpreter is the automation people want, freeing them from tedious tasks and giving them time to focus on “deeper, more satisfying” work. It will also democratize access to complex data analytics, he said. luck.
“Code Interpreter represents the clearest and most forward-looking vision yet of what AI means for work. Disruption certainly is, but disruption leads to better, more meaningful work.” I will,” he writes. “I think it’s important for all of us to think about how we can take the same approach to other AI-influenced jobs.”
