OpenAI releases ‘Deepfake’ detector to disinformation researchers

Sanjana Dhar
Sanjana Dhar May 8, 2024
Updated 2024/05/08 at 10:02 AM

OpenAI said that it would share a new deepfake detector consisting of a small group of disinformation researchers. The popular AI startup is on its way to join a global effort to spot content made with Artificial Intelligence.

Get to know more about OpenAI’s deepfake detector

OpenAI is all set to release a tool which is designed to detect content made with artificial intelligence specially which are created by its own popular image generator – DALL-E. However it is acknowledged by the AI startup that this tool will only contribute in being a small part of what will actually be needed to fight the so-called deepfakes in the coming months and years.

The new deepfake detector comes with a small group of disinformation researchers who would help in testing the tool in real world situations. They would also suggest ways how it could be improved.

Sandhini Agarwal, an OpenAI researcher who focuses on safety and policy stated,“This is to kick-start new research,That is really needed.”

Features of the new deepfake detector

According to OpenAI, the new deep fake detector can be used for correctly identifying 98.8% of images created by the latest version of OpenAI’s image generator, DALL-E 3. However as of now the deepfake detector tool is not yet designed to detect images produced by Midjourney, Stability and other popular image generators.

These kinds of deepfake detectors cannot be perfect because they are driven by probabilities, trials and errors. Therefore the AI startup is working to fight the problem in other alternative ways just like many other companies and academic labs.

Open AI is also developing or working on ways of “watermarking” AI generated sounds so that it could be easily recognised that the sound is generated by AI. The company aims to make these watermarks difficult to remove.


For more information please keep reading techinnews





Share this Article