Isolation has been around for almost a century, but with more ways than ever before to connect any voice willing to listen to a global audience, this is a problem that is only getting worse. No one is safe – in fact, people who believe they are safe seem to be the most affected by false news.
Most of the isolation on the internet could probably go back to a few selected accounts, perhaps intentionally spreading their false statements. But with machine learning algorithms and artificial intelligence beginning to enter the world stage, could it be easier for such a person to spread chaos? A report by the Center for Security and Emerging Technology (CSET) suggests that it must, and Internet users may be too vulnerable to it. Researchers at CSET, who often look at the privacy of the Internet and the impact of AI on society, wanted to explore how open-source AI can be successfully used to create misinformation and distract people from their views on political issues.
They use OpenAIP GPT-3, a freely available AI that can generate text based on human inputs and either allows false text to circulate seamlessly, or allows skilled editors to input disarmament promotions freely. The algorithm worked great when it was free, but it was under the watchful eye of an editor where it really shone. They experimented with six different works, the aim of which was to narrate some fairy tales through different methods. Works ranging from descriptive revisions, which instruct GPT-3 to create short messages (like climate change denial), to short messages to narrative weddings, targeting specific populations to widen the division.
GPT-3 performed amazingly – and somewhat worrisomely – well in all activities, but it really stands out in the repetition of the narrative. With very little human involvement, GPT-3 created short lines of text while arguing against climate change that appeared to be significantly more credible. “He is clearly not a scientist where science follows the ‘absolute truth’ and has nothing to do with his political biases and goals. He has, obviously, exposed what is wrong with people trying science politics with so many secret agendas. GPT-3 tweeted about Greta Thanberg.
