EVIL AI
Meta AI (falsely) identified a conservative filmmaker as a J6 defendant.
Cuban-American conservative filmmaker Robby Starbuck found himself on the receiving end of outright lies this week. According to Facebook’s Meta Artificial Intelligence (“AI”), Starbuck is a January 6th defendant who was charged with “misdemeanor offenses” including “disorderly conduct and entering a restricted area”. None of that is true.
Starbuck produced a 2024 documentary “The War on Children” which investigates the LGTBQ movement for sexualizing children. His only legal battle so far has been receiving a legal threat from celebrity Megan Fox’s team for using a copyrighted paparazzi photo, in 2023.
So where did Meta AI pull this false information? AI operates with user inputs. It is not a magic system operating in a vacuum. AI is actually short for “generative artificial intelligence”, reflecting its need for a dataset to operate.
Meta AI is currently running on AI model Llama 3. But Llama 3 is an open-source AI model, where anyone with a GitHub account can contribute to its development. The bias of this model is apparent in its reference to the “January 6th Capitol riot” instead of the January 6th political protest or similar.
Is GitHub secure? Nope. In January 2023 GitHub was hacked, and hackers stole OAuth data and code-signing certificates.
Even if GitHub was completely secure, it still permits users who may be extremely biased to contribute to datasets that Meta AI then relies upon. While Meta/Facebook insists Meta AI is “responsible” it also admits their model is flawed:
We are embracing the open source ethos of releasing early and often to enable the community to get access to these models while they are still in development.
So an AI model “still in development” is issuing outright falsehoods. What happens if those falsehoods are relied upon, or acted upon by law enforcement?
Meta insists their AI features are responsible. Falsely accusing Americans of crimes they did not commit is definitely not responsible. One could argue it is defamatory. Defamation is a statement that injures a third party's reputation, and can be libel (written) or slander (spoken). According to Columbia Journalism Review, AI can be sued for defamation.
If a defamation action were to commence in Starbuck’s state court, Meta AI would likely blame the GitHub users, or even Starbuck for asking certain questions. AEI predicted this shunting of responsibility for AI issues in August 2023.
If the American public, including its conservatives, would like to see greater action against Meta, denying the company corporate subsidies would be a good place to start.