NEWSROOM

A verification platform that aims to foster a critical perspective on artificial intelligence technology through its guide-like content conducted a deepfake detection test. Two out of three participants were able to distinguish AI-generated images in the test, which saw 2,579 participants. The platform called for increased vigilance against content generated by artificial intelligence during election periods.

TURKEY - Concerns have risen regarding the rapid spread of artificial intelligence technology affecting the misinformation issue and blurring the truth. With 2024 being an election year and the possibility of deepfake images and sounds created with artificial intelligence technology influencing voter behavior, these concerns have been underscored. teyit.org, which has been examining dubious information on the internet since 2016 and conducting educational activities to improve digital literacy and critical thinking skills, prepared a deepfake test to raise awareness on this issue.

Zeynep Şahin, responsible for Teyit's Educational Content, shared her evaluations, saying, "Artificial intelligence technology has developed so rapidly that visually close-to-real outputs can now appear almost every day. The clues for detecting fake constructs generated by artificial intelligence also need to be updated at the same speed. Not only in Turkey but nearly a quarter of the world's population is going to the polls this year, and we need to be more cautious against content produced by artificial intelligence."

The correct answer rate of participants in Teyit's test was 67.84%

In Teyit's test, participants were asked whether the human visuals generated by the StyleGAN2 artificial intelligence algorithm from ThisPersonDoesNotExist.com and real human photos were AI-generated or real. While 2,579 people participated in the test, the rate of correct answers was recorded as 67.84%.

Expressing concern that artificial intelligence could influence voter behavior in the election year, Zeynep Şahin said, "The use of deepfake images and sounds of candidates created with artificial intelligence reminds us that we need to be cautious during the election period. The increase in misinformation, the production of false information with unusual methods, and the distortion of the perception of reality by sounds and images generated by artificial intelligence are among the expectations for this global election year. The test prepared by teyit.org offers an effective tool to learn what to look for to catch clues and understand whether a visual is generated by artificial intelligence in a practical way."

"Artificial intelligence can also be utilized for detection"

After the introduction of Sora, which can produce videos from text commands announced by OpenAI, the close-to-real production capacity of artificial intelligence became even more apparent, stated Zeynep Şahin,

"Although images and sounds produced by artificial intelligence are now very close to reality, it is still possible to verify their authenticity. Our main task as users is to question the source of the visuals, videos, and sounds we come across and to slow down to look at our social media feeds with a more critical eye. Even if the distortions in the suspicious visual are not visible to the eye, artificial intelligence tools can be used to understand whether a visual is produced with AI. Artificial intelligence offers a solution to the problem it is thought to cause. As teyit.org, which has been fighting misinformation for eight years, we are verifying misleading information before elections and compiling false and misleading information frequently encountered in past election periods under specific headings to share."

"It may be too early to ring the alarm bells"

Zeynep Şahin, stating that opinions on artificial intelligence are positioned at two extremes and cause tension, concluded her evaluations with these words: "On one hand, we discuss its benefits, and on the other, we fear that our jobs will be taken away or that the most dangerous crimes of the future will be produced with artificial intelligence. The risks of content with fake constructs generated by artificial intelligence, such as convincing voters not to go to the polls or blurring the perception of reality, become more prominent during election periods. For example, in January 2024, a fake audio recording in the United States, which imitated Joe Biden's voice using artificial intelligence, was asking voters not to go to the polls. It is important to be aware that being aware of the risks is more important than ever.

However, on the other hand, some experts share the view that the alarm bells for this technology are being rung too early. Research penned by scholars from Harvard University indicates that new technologies always cause panic among people. As Teyit, contributing to the development of sustainable solutions to the misinformation problem, we focus on the subject of artificial intelligence in our educational content and adapt verification to the age of artificial intelligence. Anyone wanting to gain a critical perspective can become knowledgeable about the good potential of artificial intelligence by benefiting from Teyit's content."