Artificial Intelligence (AI) is transforming the internet. Tools like Grok Imagine and Sora have made it easy for anyone to create realistic AI-generated videos. This has raised alarms for many who believe that this could allow anyone to create deepfakes online. Now, Elon Musk has revealed that his AI firm xAI is working on getting Grok to soon detect AI videos and find their source of origin.
On X, a user expressed concerns about the ever-improving AI capabilities in video generation. The user warned that soon, anyone may make deepfakes of people without their consent. The videos may be of such high quality that it will become hard to identify whether it was actually made with AI.
The user wrote, “Literally nothing is being done to prevent this from happening. We can all see it coming and our leaders are doing precisely nothing at all to stop it.”
Elon Musk responded to this post. The tech billionaire revealed that his team is working on a new way to detect AI videos with Grok.
Musk said, “Grok will be able to analyze the video for AI signatures in the bitstream and then further research the Internet to assess origin.”
Grok explains Elon Musk’s vision for AI detection
The Grok AI bot replied to Musk’s comment. The AI explained that it will soon be able to better detect “subtle AI artifacts in video bitstreams, like inconsistencies in compression or generation patterns that humans can’t spot.” The AI will then go through its database as well as look for “web footprints for verification.”
The purpose of such advancements in AI detection is to ensure that users do not get flooded with AI content, and this would seemingly restore “trust in visuals” at a time when more and more videos are being generated by AI. Many are calling the AI-generated videos ‘AI slop.’
The concerns about deepfakes are turning out to be true. After OpenAI released the Sora app on the Apple App Store, there was a rise in the creation of deepfake content of other people. After backlash, OpenAI brought new restrictions to the cameo mode, allowing users to select situations they don’t want to be seen in, in the AI videos made by others.
