How to combat fake internet?
The increase of AI-generated content, next big internet problem.
As I like to call it, Fake Internet, what it means? it means every content generated by a machine. With the launch of GTP-3, (thank god is not an open API), it generates so realistic content that it is a little scary. OpenAI knew exactly why they didn’t want to release GPT all at once, they wanted to know that the world is prepared to receive the breakthrough technology.
For those who do not know what GPT-3 is, to explain in simple terms, it is a language model, trained with 175 billion parameters, which is capable of generating text almost like a human mind, it is not yet capable of passing the Touring Test but it the closest so far. In some areas GPT-3 is excelling like no other machine so far, we have examples of generating poems, generating music, generating even code.
The other content generation machine technology is called DeepFake, this is basically a term describing a piece of media like an image or a video, from where the person who is originally in the pic has been replaced by a fake one. Of course, this is fun when you put Steve Jobs face over yours or Nicolas Cage but what generates fun can be used also to generate fake content that looks real even to a tech-savvy, not to mention to my parents who aren’t so knowledgable in the internet era.
In a recent post, Facebook took down lots of accounts associated with a trolling Russian company that was used some years ago in undermining the election campaign in the USA.
Well so as these new ways of generating content by machines are getting smarter and smarter, we also need to become smarter and smarter at finding and marking them as fake. A nice approach to DeepFakes recognition was done by two teams of AI researchers. It uses biological marks left by the face in a video, it measures “residual biological signals from 32 different spots in a person’s face”.
Link to article and research paper that has also an awesome name: “How Do the Hearts of Deep Fakes Beat?”
As I like the technological progress we do in AI, I also admire that we are taking steps in combating the threats of what we are creating, which I think is the right approach of advancing in technological fields that are still unknown to us of their power.
