OpenAI’s Mira Murati Quizzed About Data Used To Train Sora AI Model: Here’s What She Said

Last Updated: March 15, 2024, 11:33 IST

United States of America (USA)

OpenAI is an AI juggernaut which has quickly moved beyond ChatGPT and now offers its AI model in other forms. We have seen AI being used to create images and videos that can be generated with a simple prompt or text.

But the company has been facing questions about its policies and how it manages to get data that is used to train the AI model. OpenAI has already gone through these allegations via the New York Times, which has sued the AI giant, claiming its AI model was trained using its articles without their permission.

So, when OpenAI CTO Mira Murati was interviewed recently, and asked about the data used to train its new Sora model, the executive clearly didn’t prepare for the grilling that was coming her way.

Murati was put through a set of questions by the Wall Street Journal, but all of us were eager to see if the company finally reveals its secret sauce to train the AI models. The journalist specifically asked Murati, what data is used by OpenAI to train its new text-to-video AI model, Sora.

Murati was quoted saying that the open has used publicly available data and licensed data to train Sora. She was further quizzed about the source of the publicly available data, with a specific reference to YouTube and Facebook. The OpenAI CTO had a befuddled reaction to that question, and decided against talking about the details of the data OpenAI uses to train its AI model.

And that’s where it has become a problem for OpenAI or any other company in the AI vertical. It is ethically important for these companies to disclose their data source, because the lack of transparency raises questions about the legitimacy and credibility of the responses generated by the AI chatbots. Recent events with Google’s own image generator have raised the concerns further.

OpenAI has Microsoft in its wing, but episodes like these don’t fill the public with confidence about the future of AI, especially when you have AI models that are slowly but surely being prepared to replace manpower. It also is a big concern when OpenAI aspires to have an AI-powered search engine out in the market in the near future.