🤺AI Combats Conspiracies

AI chatbot successfully changes minds of conspiracy theorists

Key Points 

• Researchers created a custom chatbot to counter conspiracy theories and promote self-reflection.

• Interaction with the chatbot led to a 21% reduction in participants' belief in conspiracy theories.

• However, the findings might not accurately reflect the views of hardcore conspiracy theorists, as the participants were compensated for their involvement.

☕News - Researchers have found that AI can help in fighting conspiracy theories. They've developed a chatbot that can debunk false information and encourage people to question their own beliefs.

🤔How so? The researchers created a custom chatbot using GPT-4 Turbo, which was trained to argue effectively against conspiracy theories. They recruited over 1,000 participants, ensuring a diverse range of backgrounds matching the US census data on gender, ethnicity, and other factors. By including people with varied life experiences, the team aimed to evaluate how well the chatbot could debunk different conspiracies.

Each participant was asked to describe a conspiracy theory, explain why they believed it, and rate their conviction as a percentage. This information was then shared with the chatbot, which engaged in detailed conversations with each participant. During these exchanges, the chatbot provided evidence and information countering the conspiracy theories and answered participants' questions.

The chatbot's responses were comprehensive, often spanning hundreds of words, and the average conversation lasted about 8 minutes. The results were promising: participants' confidence in their chosen conspiracy theory decreased by an average of 21% after interacting with the chatbot. Additionally, 25% of participants who were initially confident (with a score of over 50%) became uncertain. Follow-up surveys conducted two months later indicated that many participants maintained this shift in perspective over time.

😏What's the catch? While the study's findings are encouraging, the researchers caution that the participants were paid survey takers and may not accurately reflect those deeply committed to conspiracy theories.

China proposes stricter AI content labeling rules

Key Points 

• China's new rule requires digital platforms to mark AI-generated content with visible logos and metadata.

• Users posting AI content must also label it and disclose their identity if using platform-provided tools.

• Non-compliance will result in penalties for both platform operators and individuals.

💻News - China's internet regulator has proposed a new rule that would mandate digital platforms to label content generated by artificial intelligence. 

The draft plan from the Cyberspace Administration of China requires that AI-created material be marked with a visible logo and include embedded metadata. It also specifies that logos should be placed in various locations within text, images, and videos.

For audio files, a voice prompt must indicate AI origin at the start, end, and possibly during playback, and software should alert users when they access AI content. Video players must show notices at the beginning, end, and key moments of the video.

📖What's more? Users who post AI-generated content must label it as such and disclose their identity if using platform-provided AI tools - a record of their activities retained for six months. 

Labels may be applied based on embedded metadata, and if metadata is missing, service providers must assess and label the content if AI involvement is suspected. 

The draft code mandates that China's internet platforms share their techniques for detecting AI content to help improve detection efforts across the board. Non-compliance will lead to penalties for both platform operators and individuals.

🍵See also - While the draft is open for comments until October, it's unlikely that major changes will be made before it becomes official, as China's regulators generally stick closely to their draft proposals. This proposal also fits with China's usual approach of having tech giants monitor online content and link user posts to identifiable individuals, which helps the country maintain control over the local internet.