The popular ChatGPT code interpreter consumption guide is here

Fengse comes from Ao Fei Temple
Qubit | Official account QbitAI

What is the most powerful and useful function of ChatGPT since its launch ?

Probably none other than Code Interpreter .

This feature will not be officially open to all Plus users until next week, but many netizens who tested it in advance have discovered the same thing:

Its potential is so deep .

Whether it is code writing, data analysis, or chart generation, it is like a "real warrior", able to use sufficiently complex methods to output very high-quality results.

However, many people are put off by its name and think it is a tool exclusively for programmers. In fact, it can work for any novice programmer .

Now, Twitter user @Ethan Mollick has compiled a usage guide specifically for it, and it went viral immediately after it was released.

How could you miss such useful information?

Solve the five major problems of "old ChatGPT"

Before we officially start, we first have a basic understanding of the function of the code interpreter, which is the "why it is important" part explained by Ethan in his blog.

It should be noted that Ethan has obtained testing rights from the alpha version of this function and has been trying it for many months, so he has rich experience.

In his view, this code interpreter is a universal toolbox for AI to help humans solve problems (by writing Python code) .

It can upload files up to 100MB, including compressed form.

Most importantly, it solves five major problems with "old ChatGPT":

First, by writing Python code, it addresses LLM's natural weaknesses in mathematics and language. With it, you can do very complex mathematical problems and process words more accurately (such as counting the exact number of words in a paragraph, as shown below) .

Secondly, reduce the incidence of "hallucinations" and nonsense.

When the AI ​​uses Python code to do tasks directly, the code helps keep it "honest" because if the code is incorrect, the program will error. And because the code operates on the data rather than the LLM itself, the AI ​​won't insert errors in the data.

Third, make the model more general. Because a huge number of problems can be solved with code, and GPT-4 is very good at determining when a code interpreter can be used in new and interesting ways. For example, we can ask it to prove that the earth is round.

Fourth, the entire process does not require you to write a single line of code. We know that basically all LLMs can write code, but you need to run and debug it yourself. This is annoying for people who don't know Python at all (like the author). With the code interpreter, the model automatically corrects its own errors.

Finally, unlock more "AI moments" , which are moments that make us feel as if there really is a soul behind the model, and it has become a spirit.

How to use it correctly?

Taking the data set dealing with "Superheroes and Their Superpowers" (see the reference link [3] for the address) as an example, Ethan showed everyone how to correctly use the code interpreter for data analysis.

First, after turning on this function, click the + sign to upload the data file. If you have a data dictionary, you can also paste it in.

Then, in the dialog box, ask GPT-4 to look through the file and tell you what it found.

It should be noted that when using the code interpreter, just talk to GPT-4 as if it were a real person. There is no need to carefully design prompt words like we usually do. Of course, except for the following two situations:

(1) The model sometimes forgets that it can do certain things, such as making GIFs or 3D pictures, then you can encourage it ("You can make GIFs, please try") ;
(2) When you want the model to further improve the results.

As shown in the figure above, we have loaded the data and can let GPT-4 complete the most troublesome data merging and cleaning work in data analysis. It automatically corrects errors and focuses on how the author asks the question (again, treat the other person as a human being and give orders) .

Then, the analysis officially begins. Code interpreters are very good at various complex analysis methods.

For example, when the author prompted that he was interested in doing some predictive modeling to predict what superpowers a hero might have based on other factors, it went straight to building a random forest classifier, which was pretty cool.

And when you disagree with the way it estimates missing data, you can ask it to try another method instead of giving up directly.

It has so many ideas, you just ask it "Can you perform another very complex and interesting analysis?", and it generates the following results (superhero network diagram) :

The most impressive thing is that it can give some very "humanized" reasoning results, such as "The superpowers of superheroes generally bring strong visual effects to people, because this can be easily used in their context." Presented on the carrier, that is, comic books and movies..."

Finally, various visualization charts are also at your fingertips.

Overall, Ethan said that with the code interpreter, GPT-4 can fully play the role of a data scientist.

And 20 more best examples

Isn’t the above example interesting?

Good news, there is another netizen who searched the entire Internet and summarized the 20 most interesting code interpreter examples. In addition to data analysis, it can also perform more tasks that you would not expect, such as converting images to videos and generating QR codes. , analyze the stock market, etc.

Here are all examples or effects:

1. Convert pictures to videos

2. Extract text from images

3. Correctly use Monte Carlo, ARIMA, PCA, random forest and other data analysis methods

4. Use real physical phenomena to make up science fiction stories

5. Extract color palette from image

6. Generate QR code

7. Convert GIF to video

8. Analyze Apple options data to determine the best deal

9. Perform data analysis on Spotify playlists

10. Real estate data analysis

11. Generate and plot random data

12. Generate matrix rain-like GIF image

13. Clean data, perform data analysis, and generate visualizations

14. Create a gif of each lighthouse location in Beautiful Country

15. Convert data into website

16. Copy STATA code with Python

17. Download and analyze Bitcoin data sets

18. Visualization of crime data trends

19. Generate gene co-expression heat map

20. Exploratory analysis of Titanic data set

For more details and methods, you are welcome to check the original address.

What other powerful functions have you unlocked? Please share~

Original address:
https://www.oneusefulthing.org/p/what-ai-can-do-with-a-toolbox-getting
https://twitter.com/chaseleantj/status/1677679654680035328

Reference links:
[1] https://twitter.com/karpathy/status/1677512911953231874
[2] https://twitter.com/emollick/status/1677313733071085569
[3] https://www.kaggle.com/datasets/ claudiodavi/superhero-set

Click here ???? Follow me and remember to star~