OpenAI just released a new ChatGPT update called Code Interpreter, and it's available to all plus users and it's one of the most powerful and useful updates to ever come to ChatGPT.
And even though it has the word code in the name, it could actually do so many more things than what developers would use it for.
How to Load Code Interpreter
To switch on code interpreter you'll need a ChatGPT Plus account which costs $20/month and then you'll need to go to settings and beta features and switch on code interpreter. You can then activate it from the GPT4 plugin menu.
And essentially what this does is it enables ChatGPT to use Python script and handle uploads and downloads in a secure environment. The best way to think of code interpreter is it's like having a very junior programmer at you fingertips.
Describe An Image
So the first thing that I tried with the Code Interpreter was actually uploading a photo and then seeing what Code Interpreter could do. So in this case, I took a cute picture of a Labrador puppy and uploaded it into the chat GPT prompt window. It was then able to describe the image.
Now I want to take things a little bit of a step further here so I then got ChatGPT to actually describe a prompt that I could then move over into something like mid journey or dually to generate my own AI version of this image and the results were pretty cool.
And there's lots of other cool things that we can do with images too. We can zoom in and out, we can actually pull off the color palette from a particular photo, we can change the color grading, and we can even upgrade and change the resolution of an image to make it slightly more high fidelity.
We can crop it, we can spin it around, and lots of other things. I'm going to be diving into this a little bit more later on in the video, so be sure to stick around.
One of the coolest things you can do with ChatGPT Code Interpreter is to QR code that then you can use for things like marketing or pointing people towards your product.
For me, I used my website, and asked ChatGPT to create a QR code. And as you can see, it did this really quickly. And if you scan that QR code right now, it will take you to my website, and you can sign up to my newsletter that goes live every Thursday and Sundays, covering tech and AI, and lots more about learning and human performance.
Now, the next thing I tried out with Code Interpreter was actually using it for something that I do every day, which is project management.
At both my companies and in my personal life, I'm pretty organized and I'll often map things into an Excel chart or use productivity software like Asana or Monday or ClickUp to organize everything that I'm doing to ensure success.
But this can be really time consuming, and if anyone's filled out a Gantt chart, you know it can take absolutely ages. In this really practical example, I took a project we're currently working on at Virti, which is all to do with surgical training, and I put down the actual outcomes and the deadlines we needed to complete things by, and then uploaded it into ChatGPT, and asked it to output a Gantt chart into a CSV file.
If I wasn't sure what the project was, I could even use ChatGPT to map out the different steps towards a specific goal.
Now, we can also take this a little bit of a step further, and download that CSV file, and then upload it into Asana to populate our Asana project management, which then we can share with our team.
Now, this next prompt with Code Interpreter is going to be especially useful for anyone who's studying at the moment, either in their professional lives or at school.
I'm a massive advocate of using spacing schedules combined with active recall to learn as quickly and effectively as possible and to negate any of the effects of the forgetting curve.
But actually creating a spacing schedule and study schedule can be really time consuming and a little bit tricky if you're new to getting started.
So I downloaded the outline of a spacing schedule from the internet, and then I fed it back into ChatGPT using Code Interpreter.
Code Interpreter can then break this down. I also used it for more of a zero-shot prompt, where I'm just saying something like, I'm revising for medical finals exam in about 12 months time, and I want you to map out all the topics I need to revise for, and then put it into a spacing study schedule that I can then apply myself.
ChatGPT again does a really good job here of breaking down all of the topics and then using a spacing schedule, in this case I use SuperMemos, which is quite widely used, to then plot in time for revision sessions to ensure that I don't forget things.
Now, one of the things that I've spent ages scouring the internet for is ways to convert files into different file types. Often this might be a PDF into a Word document into a text document, or one type of video file into another.
And it can just take absolutely ages. Luckily, Code Interpreter is here to save you tons and tons of time.
You can upload pretty much any file, from a PDF to a video, and then just ask it, using prompts, to actually change it into a new file type. In this case, I converted a PDF pretty quickly into a text document, and also a video file into an audio file, and it worked really nicely.
Complex Math Equations
Okay, so the next use case for Code Interpreter is really helpful if you're studying a subject like maths or physics, or you're working in an industry like engineering you're doing lots of complex mathematical computations and equations.
Code Interpreter can run some pretty complex math equations and produce graphs and other things.
If you've got a large dataset of numbers or something from a research study that you need to run a deep mathematical equation on to pull out the results of, say, for example, calculating the P value or the impact of a particular intervention in something like medicine, then you can do that through ChatGPT rather than having to use paid formula equation software, which can be very expensive and difficult to use chat GPT is conversational element.
Now, just before we dive into a bunch of prompts that analyze big data sets, which is one of the best use cases for code interpreter. I wanted to throw in just a really practical and straightforward way that we can use code interpreter right now to save us tons and tons of time.
And it's this, you can just upload any document and then analyze that document very quickly, pulling out or summarizing any information that you might need. Previously, you might need to use a ChatGPT plugin to do this or a separate website, but now you can upload any PDF document like in this case, a biology or chemistry GCSE study module and then ask ChatGPT to summarize things or pull out relevant information or even turn things into questions for you from this.
Now this is super helpful if you work in an industry where you're dealing with very large documents and you need to quickly find something. Or you just need something summarized and it's also going to be really helpful for both students and teachers when it comes to analyzing things like projects or homework, which might be sent in a word document or a PDF format.
Where To Get Datasets
Okay, so Code Interpreter saves us loads of time when it comes to analyzing big data sets, but before we actually get to play time, I just want to tell you about one quick website where you can download data sets very quickly. Plug it into ChatGPT and play right now after you've watched this video.
And that website is called Kaggle. It's great for machine learning engineers, and you can search from a variety of different datasets and download them.
Okay, so now we've got our data set and again, this could be absolutely anything from scraped web data to things you downloaded from Google Search Console to this case, where it's a lot of health data, you just drag and drop it into Code Interpreter, and then we can ask ChatGBT to analyze everything contained within that data.
We can ask it to segment the data. Pull out trends and even create graphs as I've done in this case. So rather than having to trace through all of that data, which might take hours and hours of our time, we can actually ask Code Interpreter to pull off the graphs and Code Interpreter can actually format lots of different graph types.
It even produces things like word clouds, which are great for sharing language of your customers or your users with both your product team, your marketing team and your wider audience.
It's not just graphs and word clouds that Code Interpreter can output to. It can also do some really cool things if it's given location data.
For example, I downloaded a data set of UFO sightings globally from Kaggle and then I uploaded it into ChatGPT Code Interpreter, and we can actually get a geolocation map of the entire world showing some hotspots of supposed UFO activity.
We can then in real time Ask ChatGPT to zoom in on specific locations, like for example, the United Kingdom and see exactly what's happening in different locations and areas and map it over different cities or landmarks, which is just super, super cool.
Now, whether you're learning to code or actually building something, Code Interpreter is a complete game changer here. You can actually upload all of your files in a zip document and then get the entire thing analyzed and critiqued by ChatGPT. And then you can work on improving things and optimizing your code.
If you're just starting out, you might want to download some code from GitHub in a zip file and then upload it and get ChatGPT to explain things to you. If you're a little bit more advanced and are running your own projects or coding something for a company and you're stuck, you can upload anything you're working on into ChatGPT.
And a little bit like having a programmer next to you to explain things in simple terms, it can actually point things out that you might have missed.
Create Code From Scratch
And it's not just code analysis that Code Interpreter can handle. You can actually ask it to create code from scratch, just like you can with vanilla ChatGPT. But in this instance, it will actually output a file. So for example, we could get ChatGPT to create a manifest file for a ChatGPT plugin, and then Code Interpreter will actually produce this file, whether it's a JSON or a ZIP file containing all of the lines of code.
We can then look at that ourselves, host it on a local environment and test it, or we could host it on the cloud and then run more code through ChatGPT to then interact with that hosted code and produce something like a ChatGPT plugin, such as a habit tracker.
Code interpreter can shorten video clips and do some basic video editing.
We can also do things like turning the video around, or even changing the video file format, or just pulling out the audio if we've got a video with audio.
You can also get ChatGPT and Code Interpreter to produce things like GIFs, though this is still in the very early stages.
Code Interpreter Limitations
Remember, while Code Interpreter is really great, it's still in beta and it doesn't always do exactly what you want.
When you upload a file, it's also worth mentioning that this is held in a temporary storage area, and if you're working on a big project like I was when I was playing around with lots of data, sometimes that file can be lost and then it irritatingly asks you to re upload things.
Or just forgets other lines of code that it had created in the past. And this can drag that project out for longer than it needs to be.
Now, that being said, Code Interpreter is still an absolutely amazing tool. And if you're working with large datasets, it's going to be completely revolutionary and it's going to save us so much time in the future.