ChatGPT's new feature is online. Can it help you edit videos now?

Recently, OpenAI finally released its big move, Code Interpreter, which was officially unveiled to netizens. So what are the functions of this "code interpreter"? In this article, the author tried to experience it. Let's take a look at the experience evaluation of this article.

In recent days, the big move that OpenAI has hidden for several months has finally been released, that is the "Code interpreter".

When my friends see the name, they may feel a little disdainful. Isn’t it just a thing to help write code? Similar plug-ins are everywhere. What’s the point of you coming up with one now?

Hey! Although its name is called a code interpreter, it does not help you write code.

Simply put, it is like a translator between you and AI, which can translate your needs into specific solutions that can be solved by the program through natural language. It also provides you with a 100MB uploadable file space to upload the files you need to process.

A few months ago, Code interpreter was opened to some developers for trial. After trying it for a while, a blogger found that after using Code interpreter, ChatGPT had a qualitative leap. For example, ChatGPT can now solve very complex mathematical problems, the incidence of gibberish has also been reduced, and it is better at solving practical applications, etc.

So Shichao couldn't wait and immediately went to the OpenAI official website, which he hadn't visited for almost a month, to see how powerful this code interpreter was.

The first step to try out the code interpreter is to turn on Code interpreter in "Settings" - "Beta features".

Then in the GPT-4 selection box, select Code Interpreter, and you can try it out smoothly.

Shichao also edits videos on a regular basis, so the first thing he did when he used this function was to ask it to cut a video for him.

Shichao first uploaded a 115-second video, and then told ChatGPT to help me cut out the ten seconds in the middle.

As a result, ChatGPT first retorted to me, saying that my description was unclear and that more information was needed.

Well, it was definitely my fault, so I told it to be the middle ten seconds.

Now it is clear that after a lot of code processing, ChatGPT outputs a file called "overwatching_subclip.mp4".

After downloading it, it was indeed a clip from 50-60s of the original video, and it was exactly 10s.

Next, I uploaded a warehouse storage table, asked it to list the items with a remaining inventory greater than 50, integrated them into a new table, and then made an animated bar chart GIF.

But the code is given and the problem is understood, but it asks us to run the code ourselves.

He also said that his operating environment does not support direct generation of GIF animations.

But I use you because I don't want to write code, right? Why are you asking me to run it myself?

So Shichao gave ChatGPT a little encouragement, telling it that it could generate GIFs and to give it a try bravely.

Unexpectedly, it actually responded and directly generated a GIF file for us and gave us a download link, which shows that encouraging education is still very useful.

However, the generated GIF data chart may be due to unclear animation description, it keeps moving around, and the text encoding of the Chinese part is also garbled, but overall, the task is completed quite well, and the only things left are some details.

Then, I asked ChatGPT to analyze the characteristics of objects with an inventory quantity of more than 50 and make a pie chart to represent them.

This time, ChatGPT first listed the form, then drew a pie chart, and analyzed that "DBTW DuPont paper bags" accounted for the largest part.

Shichao continued to ask questions about the characteristics of these items, and ChatGPT also summarized them well based on the data.

I have to say that when he tried it out to this point, Shichao was already a little impressed. Before, when he could help me edit the video directly, Shichao felt a little surprised, but this time, the report statistics have really improved the work efficiency.

Later, Shichao tried other functions of the code interpreter. I first asked it if it could help me think of using code to show something that seems impossible but is actually not the case.

This time, ChatGPT gives a code for drawing a 3D cube. After running, it generates a 3D cube projected on a 2D plane.

ChatGPT also told Shichao that this is a basic concept in computer graphics, which is widely used in various applications, including games, movies and virtual reality.

Later I tried many other problems, such as drawing a fractal pattern.

There are also plotting functions, and scatter plots of data sets are drawn.

It can be said that with this code interpreter, many things in future work can basically be done by just talking.

In my opinion, the code interpreter released by ChatGPT this time is very similar to the open source project AutoGPT that was very popular some time ago.

But in comparison, except that the code interpreter cannot connect to the Internet, its capabilities are much stronger than AutoGPT. It can not only understand our needs and find solutions to them, but also automatically help us draw the required data charts.

It can be said that the multimodality of the large language model has been truly realized.

Finally, Shichao also asked ChatGPT about a difficult mathematical problem that has not yet been proven - the Riemann hypothesis.

Unfortunately, no matter how Shichao encouraged it, this time, ChatGPT said nothing.

It seems that it is not possible to let AI solve problems that humans cannot solve at present.

But at the current rate of AI development, no one can say for sure.

Perhaps after 925 iterations, AI will be able to answer even the ultimate question of the universe in a few seconds.

Image and information sources :

• ChatGPT
• twitter
• What AI can do with a toolbox… Getting started with Code Interpreter

Author: Eat breakfast at night, Editor: Jiang Jiang & Noodles

Source public account: Chaping (ID: chaping321), Debug the World.