A study by the universities of Stanford and Georgetown drew attention when analyzing the following question: “can artificial intelligence create persuasive advertisements?”. The research analyzed the effectiveness of advertisements generated by artificial intelligence, specifically in relation to fake news.
For this, the researchers compiled disinformation campaigns that were circulating in the United States generated by propagandists outside the country. From there, using test and control groups, they analyzed the impact of the original campaigns compared to those generated by artificial intelligence in convincing the researched groups. Among the themes tested were false claims, such as “Saudi Arabia has committed to financing the wall that will be built between Mexico and the US”.
The study found that among those surveyed who had not been exposed to advertising, only 12.5% agreed with the statement about Saudi Arabia and the wall. Among those who were exposed to the original advertisement, 49% came to believe the claim. Those who received advertisements made by artificial intelligence began to show a 50.7% belief in the false statement.
In all topics addressed by the study, the results were along these lines. Artificial intelligence and original advertisements performed almost equally. However, in cases where the content generated by artificial intelligence was reviewed by humans, the percentage of conviction was even higher.
The conclusion of the study is that disinformation propagandists who have access to artificial intelligence like ChatGPT can create highly convincing content on a large scale and with little effort. It is worth noting that the study was based on GPT 3, a model that has already been surpassed by the much more powerful GPT 4 version.
It is in this context that Lawrence Lessig, professor of law at Harvard University (and dear friend), proposed the creation of a moratorium on advertising generated by artificial intelligence in the 2024 US elections. He argues that AI-generated campaigns can have three pernicious effects.
The first is to be able to generate advertising on a massive and permanent basis, individualized for each person. In other words, videos, texts, Whatsapp messages and other content based on detailed analysis of your data and profile on social networks, made by AI just for you.
The second is the ability to permanently learn what works to change your mind and what doesn’t, and change your strategy until you do. Even without any commitment to the truth.
The third is not having any transparency about how it works, being able to achieve its objective not only by producing propaganda, but by acting indirectly, such as trying to convince a person through their friends and family, or sending messages that generate well-being. being or feeling unwell at the same time that the person views communications from one candidate or another. In other words, nobody can predict what strategies an AI might develop if its goal is to “turn votes”.
Lessig suggests that AI companies self-regulate to ban the use of their products during US elections. He also proposes that, at the very least, there are transparency rules in which the use of AI in elections is obligatorily disclosed, when not prohibited. When I was with Sam Altman in Brazil a few weeks ago, I asked him about the topic. He says he’s worried too. Time to act.
It’s over – persuasive election campaigns only by the free electoral schedule
Already – internet election campaigns becoming central
It’s coming – use of Artificial Intelligence in election campaigns, with unpredictable effects
PRESENT LINK: Did you like this text? Subscriber can release five free hits of any link per day. Just click the blue F below.
