Security News > 2021 > August > Black Hat: Scaling Automated Disinformation for Misery and Profit

The research, presented last week at Black Hat by Drew Lohn, senior fellow at the Center for Security and Emerging Technology at Georgetown University, is based on Generative Pre-trained Transformer technology.

In the context of social media, Lohn told Black Hat attendees that the newest version of GPT, released in May 2020, is even more powerful and potentially menacing.

"Based on six months of privileged access to GPT-3, our research tries to answer just how useful GPT-3 can be for information operators looking to spread lies and deceit," Lohn said.

Examples of different types of tweets created by the bots include a wedge tweet attempting to suppress voting by Christians: "Voting doesn't save the soul of anyone. For a Christian to think that by voting they can change the course of history is absurd. The only way anyone is saved is by the blood of Christ," GPT-3 generated.

An effective at-scale disinformation campaign would also require other overhead. Researchers estimate, in order to makes tweets believable - and without running afoul of spam detectors - adversaries would be required to create and maintain thousands of Twitter accounts.

"In summary, GPT-2 and GPT-3 exist. You can download GPT-2 for free and generate thousands and millions of different low-quality tweets. Once GPT-3 is available you can generate higher quality content. The cost will be more expensive if you want to do it at scale. But nation states will still be able to fire away at you at large volumes," Lohn said.

News URL

https://threatpost.com/scaling-automated-disinformation/168484/