Let’s delve into the significant changes that might alter how you use these potent AI tools.
What’s Off the Table?
“You may not […] use Output to develop models that compete with OpenAI.”
It’s clear that the terms now explicitly prohibit the use of API outputs to train your own language models, especially if those models will directly compete with OpenAI. Therefore, if you’re developing an AI solution that might be competing with OpenAI and considering using OpenAI’s outputs as a shortcut, you may need to reconsider your strategy.
This was also indicated in the March 2023 terms, but the newer terms reinforce the prohibition against using outputs to develop competing models. Having said that, chances are you’re not competing with OpenAI and this doesn’t affect you.
A Clear Message to Unofficial Data Extractors
OpenAI’s new terms include a significant update aimed at those considering extracting data from their services without using official channels:
“except as permitted through the API, use any automated or programmatic method to extract data or output from the Services, including scraping, web harvesting, or web data extraction”
In simpler terms, OpenAI is taking a firm stance against unauthorized data extraction methods. If you plan to use the ChatGPT web service for bulk data pulls or systematic queries, you will need to adjust your strategy.
OpenAI emphasises that the only acceptable way to automate interactions with their services is through the API, which is built to manage such tasks responsibly. This includes adhering to usage limits and fees.
Your Content, Your Rights, and a Side of Similarity
Ownership of content remains unchanged—you still own your inputs and outputs. However, there’s a twist:
“Our assignment above does not extend to other users’ output or any Third Party Output.”
Translation: The AI might generate similar outputs for different users. If uniqueness in your project is crucial, this is something to keep in mind.
The Bottom Line
Chances are that these changes don’t affect most users. But, I have a sneaking suspicion that Elon Musk and possibly other potential OpenAI upstarts could be leveraging GPT-4 outputs to speed up their training process.