Subscribe. Scale. Succeed.
We’re so confident you’ll love Akkio, we’ve made our service month to month. Ideal for people with commitment issues and ROI desires.
On July 6th, OpenAI announced the general availability of their ChatGPT Code Interpreter, an extension that can run Python code, generate data visualizations, and even perform basic photo editing tasks.
Basically, anything that can be done with Python can now be automated with ChatGPT Plus and Code Interpreter.
Let’s explore the use cases for the new extension, understand its limitations and explore how these types of tools can save you hours every week.
Code Interpreter provides a secure and versatile Python execution environment. It operates in a sandboxed, firewalled setup, ensuring code evaluation without compromising system integrity.
Code Interpreter can only run Python code, so its main use cases include data visualization, data preparation, and data clustering. Many ChatGPT Plus users are also using it to convert files into different formats (e.g. from JPG to PNG, or from GIF to MP4) and cut files into smaller chunks.
Do you recall when OpenAI demonstrated how GPT-4 could take an image and produce a webpage that looked exactly like the image?
Although not quite there yet, Code Interpreter comes very close. ChatGPT can now read photos, process tables, and even unzip directories thanks to the file uploader.
Here are the top time-saving prompts for usage with the new ChatGPT Code Interpreter at work:
Use Code Interpreter to compress, cut, upscale, and process images. A few examples:
After uploading your file, ask ChatGPT to convert your image from format X to format Y. In our example, from PNG to JPG.
ChatGPT will return a downloadable file in the new format. Please download it quickly, as files uploaded to the cloud will not be available forever.
Note that ChatGPT ran into an error during the conversion process. The plugin automatically understood the problem, fixed it, and returned the JPG file.
Code Interpreter is now an effective alternative to many file converters out there, like CloudConvert or TinyPNG.
ChatGPT is now capable of removing the background from any PNG image with great accuracy. To do this, simply upload your PNG file to the platform and let the Code Interpreter run its magic.
PNG files are heavier than JPGs, but it’s the only format that allows for a transparent background. You could then convert the PNG to an SVG if you prefer the file in a vector format for graphic design purposes.
ChatGPT will provide a full explanation of the process. If you don’t need it, just specify it in the prompt (like: “Just do it, no explanation required”).
ChatGPT can’t just convert and transform images, it can process them. This means it can understand metrics such as pixel density, colors, and limits. Anything that a Python library can look for, the Code Interpreter can too.
In this example, let’s have ChatGPT Code Interpreter output the color palette of Stripe’s Banner.
This looks a bit messy. Let’s ask it to limit the output to the “main colors” and leave it free to decide what “main” means.
Great! This looks exactly right. If you need it as a downloadable file, you can follow up by requesting it in PNG or JPG format.
For basic image editing, you can use the ChatGPT Code Interpreter to cut, resize, and compress images in a whiff.
This one is as simple as it sounds. Provide ChatGPT with any URL, and it will be able to generate a QR code for you.
The same capability is available in most browsers, but having it incorporated into the same tool that you may use for other purposes is really convenient. For example, you may access the URL via scraping the web with other plugins or determining the page from a video transcript.
You can also color the QR code if that’s your thing. All steps are clearly detailed by clicking on “Show Work” - in case you’re curious:
You can use ChatGPT Code Interpreter to extract key points from a dataset, get a quick explainer of an infographic, or perform mathematical calculations.
For these examples, we generated a custom dataset with real data from the World Bank. You can build your own here.
If you’d like to follow through with the same dataset, you can find it here. It collects key indicators such as CO2 emissions, GDP, GDP per capita, and government debt for 266 countries.
Getting a quick overview of the dataset in a few sentences can be invaluable when working with large datasets.
ChatGPT Code Interpreter is capable of providing quick overviews of the dataset. You can even ask for a specific length, let’s say, “a single paragraph”.
But it’s not the only tool that can easily do data analysis. Akkio’s Chat Explore feature lets you explore your data, ask questions, and get answer – but even better, it lets you go around ChatGPT’s limits – give Akkio a try!
In a hurry for a meeting? No worries, ask ChatGPT to provide key takeaways, and you’ll have a quick snapshot of everything that’s important at the table.
Of course, these aren’t human takeaways. You might think something is important even if ChatGPT finds it not valuable. Regardless, it can be a huge time saver.
ChatGPT Plus users can create data visualizations with a few prompts in ChatGPT Code Interpreter. This is likely the most popular feature of the plugin, and the one you’ll see thrown around more often on social networks.
ChatGPT can now generate any type of graph, or at least those for which it has libraries. It presently struggles to make word map graphs, but it can readily generate line charts, plot charts, and bar graphs.
If the plugin is unable to build a graph, it will notify you of the cause and allow you to generate another. It will sometimes do it on its own.
Using the same global dataset as a prompt, ask ChatGPT Code Interpreter to build a graph displaying the top ten and bottom ten nations in the world in terms of GDP per capita.
And there you go. In a matter of seconds, you have the data visualized in a nice graph with gradients.
Some users have had luck generating animated GIFs for datasets that include a time column. In our experience, this usually doesn’t work, as ChatGPT will warn you that it doesn’t have access to the necessary libraries to generate GIFs.
It’s very likely that generating dynamic graphs will be part of a future update.
Other tools, like Akkio’s Chat Explore, also lets you build charts and graphs with generative AI, but more importantly, it gives you dashboards and lets you update your data in real time.
One of the unique features of Code Interpreter is how fast it is at generating visualizations for a large set of indicators without requiring multiple prompts.
For example, here I generated 10 graphs showing the key indicators for the USA.
We didn’t provide any real input other than “create 10 visuals to represent this data”. This is perfect if you’re creating a presentation on the go or need quick insights for a team meeting.
Instead of visualizing the charts in GPT-4, which is stunning visually but practically useless in the real world, you can instead generate all graphs as downloadable PNGs.
The output will look something like this:
You can also ask for a specific analysis that includes multiple graphs and output the final result as a PDF. This allows for more complex and in-depth analysis. These are clearly more difficult and will sometimes result in errors for timeouts.
ChatGPT Code Interpreter has access to matlib, a Python library to execute and work on complex math equations.
Many students are already using it to upload pictures of their homework assignments and get Code Interpreter to solve them, but that’s clearly not the main use case.
If you need to process data, similarly to our Chat Data Prep, you can do so in plain English. Chat Data Prep is faster and safeguards your company’s data. You can also keep data up to date when you deploy it to an endpoint.
Standardizing date formats involves ensuring that all dates in a dataset are represented consistently and uniformly. Inconsistent date formats can make it challenging to analyze and compare dates accurately. The standardization process may involve converting dates to a specific format, such as YYYY-MM-DD, or ensuring that all dates are represented using a consistent naming convention.
Removing empty columns involves identifying and eliminating columns in a dataset that do not contain any meaningful data or have missing values across all rows. Empty columns can clutter the dataset and introduce unnecessary noise, making it harder to analyze or process the data effectively. By removing empty columns, the dataset becomes more streamlined and focused on relevant information.
Code Interpreter is not a machine learning platform, but it’s able to create very basic predictions starting from the dataset.
Predictions follow simple mathematical principles, and will not include any sort of deep learning algorithm (something Akkio includes on all plans).
They’re also static, they don’t create a model to predict future outcomes based on new data, nor employ AI models in the prediction process.
ChatGPT Code Interpreter is one of the most powerful plugins released by OpenAI, but it does come with a few limits to keep in mind.
Most notably:
There’s no way to know if Code Interpreter works for you without a financial commitment. Although minimal, ChatGPT Plus costs $20 with no refund guarantee nor free trial options.
The only way to get access to plugins is with a premium subscription. Without it, free users are stuck with GPT-3.5, which is wildly less creative than GPT-4, and won’t see plugins.
If you want to get a taste of AI-powered data analytics for free, you can try out Akkio’s AI platform for free.
OpenAI is under investigation for unsafe data privacy practices. At the moment of writing, this led to the team deactivating the Bing Search plugin in the European Union since it didn’t comply with the data protection laws in the union.
Unless you opt out of conversation history, the corporation is also renowned for utilizing your data to train its models, which greatly limits the value of a product like ChatGPT for data analysis. You can't visualize anything for longer than an active session if you can't see prior chats, which means you have to download all your charts and save them in a folder.
Many organizations are concerned about their staff utilizing ChatGPT when it comes to sensitive corporate data. Many restrict the website on the company's intranet, and many more control what data may be submitted and when. One notable recent example is Samsung's ban of ChatGPT.
Data analysts do not produce visualizations for the purpose of creating them; rather, they require them to demonstrate the effect of the company's efforts to their teams.
To do this, they employ centralized dashboards with several reports and live refreshes, which take in real data from many sources and update the graph.
ChatGPT Code Interpreter is fantastic, however it does not enable analysts to retain or utilize live data for graphs, therefore it cannot be used for continuing work or as a single source of truth when making critical corporate decisions.
ChatGPT is unrestricted while using GPT-3, but has a strict 50 message restriction every 3 hours when using GPT-4. When working on large datasets, analysts must work quickly and cannot feasibly complete jobs in 50 questions.
This is currently the most significant limitation of ChatGPT Code Interpreter, and given how costly it is to maintain a tool like ChatGPT (according to Govtech’s estimate, running ChatGPT costs nearly $700,000 per day), it is unlikely to improve.
Some accounts may have more favorable limits. Enterprise users likely have larger limits as well.
ChatGPT Code Interpreter doesn’t have any real integration with databases right now. You can’t connect it Google Sheet, Airtable, or your CRM. It also doesn’t allow for live data insertion, meaning all graphs developed with it will constantly be out of data.
It’s surely interesting to create graphs from your csvs and xlsx files, but the reality is that most data lives in datacenters and office tools, and having a direct link to these is essential for accurate, up to date information for business intelligence.
ChatGPT, developed by OpenAI, has truly emerged as an expansive platform that caters to the general public. This AI system has been designed to be broadly applicable and accessible, to resonate with the needs of various users, be they individual consumers, researchers, or business professionals.
Yet, despite its robust functionalities, ChatGPT cannot handle every challenge nor does it fit into standard workflows. This is where the symbiotic relationship with developers and other AI technology firms comes to play. OpenAI encourages and depends on third-party developers to leverage its advanced AI model and capabilities, creating niche, industry-specific software solutions. In this context, Akkio is a notable example, expertly taking advantage of the technological prowess offered by OpenAI.
In short, OpenAI’s goal will never be to develop the best platform for data analysts - their scope is much broader.
Chat Explore, Chat Data Prep, and Akkio’s autoML platform are the most effective tools for data analysts. Our features are faster, more precise, and have been battle-tested by users across 30+ industries. Join analysts from Google, PwC, Intel, and more in the best end-to-end ML platform designed for data analysts.
In conclusion, OpenAI's release of the ChatGPT Code Interpreter plugin marks a significant advancement in automating Python code execution and providing time-saving features for professionals in data science, data analytics, and programming.
With the Code Interpreter, ChatGPT can now read and process images, generate data visualizations, remove image backgrounds, extract color palettes, resize and compress images, create QR codes, analyze data, and perform basic mathematical calculations.
These capabilities provide a range of time-saving prompts for users, empowering them to quickly obtain insights, generate visual representations, and perform data processing tasks.
However, while the ChatGPT Code Interpreter offers valuable time-saving features, it may not fully meet the needs of data analysts seeking live data integration or working with sensitive corporate data. Developers like Akkio and features like Chat Explore and Chat Data Prep can provide more tailored solutions for specific industry needs.