01: The Unexpected AI Assistant
OpenAI has released a long-awaited new feature for ChatGPT Plus users called Code Interpreter, and code is only one small part of its capabilities.
This feature is like having a junior programmer, graphics assistant, and data analyst at your fingertips, ready to assist you with various tasks. It can learn, write & use its own Python code, access files you’ve uploaded, and you can even download the results of your work.
Here are just a few examples of what you can do with it:
- Image manipulation: You can upload an image and ask the AI to manipulate its size and color.
- Data analysis: You can upload a CSV file and analyze it with simple questions. It can create charts and graph visualizations to help you better understand your data.
- Converting files between formats: AI can handle the conversion for you if you have a file in one form and need it in another.
One real-life example of how this can be used is when Wharton professor Ethan Mollick uploaded a CSV file of every lighthouse located in the US. He asked the AI to “Create a gif of a map of the lighthouse locations, where the map is very dark but each lighthouse twinkles.” The AI completed this task, showcasing its ability to handle complex requests with nothing more than natural language.
I’ve also been experimenting with it to test its limits, and I uploaded a PDF with data rows to extract. When the AI couldn’t complete the task flat out, it recognized the error and problem solved proactively, asking if I wanted to use OCR (Optical Character Recognition). It then rewrote its code and process with the new method. It showed a sample of success but was self-aware enough to realize it couldn’t process the full PDF mixed with copy and the rows I needed, suggesting the best method to identify which page[s] it should prioritize so it could run batches. I identified the key pages for Code Interpreter so it could finish its own code rewrite, and it gave me the data I was looking for.
This whole process took less than 90 seconds.
This feature is incredibly convenient for those tasks that require complex tracking and organizing or just getting started fast. It’s like having an AI assistant who thinks of things you might not have considered and pre-formats them all with the power of doing them for you.
🧠 This changes a lot for your business. Code Interpreter can handle coding tasks requiring a dedicated programmer, analysis tasks needed for your data, and even graphics tasks for your website, marketing, and more.
The Code Interpreter feature is being rolled out to all ChatGPT Plus users. To enable it, click on your name, select Beta Features in settings, and toggle it on when available.
02: A Data Scraping Dilemma
Data is the lifeblood of AI.
It’s the raw material that fuels machine learning, powers predictive models, and enables AI systems to understand and interact with the world. But as AI becomes more pervasive and powerful, how data is collected and used is under increasing scrutiny.
Google recently updated its privacy policy to collect public data for AI training. This data will be used to train Google’s AI models and build products/features such as Google Translate, Bard, and Cloud AI capabilities. The key shift here is that Google can now use any information people post publicly online, not just data provided directly. This broadens Google’s focus from improving “language” models to enhancing all its “AI” models. However, analyzing people’s online posts using an AI system does pose privacy concerns.
On another front, Elon Musk and Twitter have instituted viewing rate limitations, forcing non-paying users to view fewer posts per day due to data scraping by AI companies. Elon is doing everything he can to stop AI companies from stealing his precious data.
While Twitter is in severe pain and frustrating loyal users, the competition is heating up.
Meta has released Threads, a new social network that has been wildly successful, racing over 100 million signups in only five days. If you’re on Threads connect with me here 🔗 for some quality dad jokes and AI fun.
Then to add even more public awareness to the problem, OpenAI is being sued 🔗 over data privacy. The lawsuit could decide whether the company’s use of training data scraped from the public internet may continue or require a complete shift.
*exhale*
While I believe in free and open information, there are limits to how I want that data used, and I’d like to think more agree with me than not.
We need a way to instruct AI companies on what data we want to be indexed and trained on, similar to how we use the “no-index” flag with SEO and search. The industry is evolving rapidly; there must be a healthy mix between innovation and contribution.
If you’re with me on this stance, here’s how to be more intentional about data protection when you need it for your privacy. Use alternative services prioritizing user privacy (DuckDuckGo for search, ProtonMail for email, Vimeo for video sharing, and Brave for web browsing). And of course, use incognito or private browsing mode when you’re on sites you don’t want to be tracked against, and be mindful of the information you post publicly in those sessions.
As AI continues to evolve, it’s essential to understand the implications and potential benefits it can bring to your business. But it’s equally important to be aware of the potential risks and take steps to protect your data.
03: A New Era of AI Tools Is Here
The long-awaited GPT-4 API is now generally available for all paying API customers & this is a section of the newsletter you should forward to the developer on your team.
GPT-4 is OpenAI’s most capable model, and since its introduction in March, millions of developers, including myself, have requested access to the GPT-4 API. The range of innovative products leveraging GPT-4 is growing daily, and the results are already showing significant improvements.
This is a significant development for developers and those using AI tools where GPT-4 is an option. The GPT-4 API is a powerful tool that can be used to create more advanced AI applications. It’s a significant upgrade from the previous GPT-3 model, offering improved performance and capabilities with huge gains.
It’s important to note that we’re highlighting GPT-4 in the API only, not as part of ChatGPT.
While GPT-4 is the underlying model that powers the API, ChatGPT is a separate application that uses the GPT-4 model to generate human-like text. The API gives developers direct access to the GPT-4 model, allowing them to integrate it into their applications and services.
🧠 This is a significant development for businesses and developers.
The GPT-4 API is more precise, more accurate, and better at editing, with an excellent ability to spot errors in the text it works with. It’s also better at making jokes, adapting to challenges, and acing professional tests and exams. For example, GPT-4 scored in the 90th percentile on the Uniform Bar Exam compared to GPT-3.5, which scored in the 10th percentile.
GPT-4 is also 82% better than GPT-3.5 at detecting inappropriate requests and has better guardrails. This means using it in a professional setting is safer and more reliable.
If you’ve been receiving Top 3 Tuesday for a while, you’ll remember my chatbot write-up and the challenge of having an AI model act as you. This API directly adds toward improving this, amongst many other improvements.
The GPT-4 API also opens up possibilities for creating more advanced AI applications, and it’s something that every developer should be aware of. It’s not just about creating more advanced AI applications; it’s about unlocking the full potential of AI and using it to drive innovation and growth.
In short, if you’re a developer or a business owner, now is the time to start exploring the possibilities of the GPT-4 API.
