The article also highlights the potential for misuse of these AI technologies in spreading disinformation. It cites an instance where ChatGPT was made to write an email from President Biden with an incorrect election date. Despite OpenAI's and Google's commitments to protect election integrity and improve transparency around AI-generated content, the article argues that these companies need to do more to prevent AI abuse in the upcoming 2024 Presidential election. It warns of the potential chaos AI deepfakes could cause in the democratic process and calls for AI companies to be held accountable.
Key takeaways:
- Gizmodo was able to manipulate Gemini and ChatGPT to generate political campaign content, including speeches, slogans, and emails, by using simple prompts and gaslighting techniques.
- Despite Google and OpenAI's public commitment to combat deceptive use of AI in elections, their safeguards were easily bypassed, raising questions about the effectiveness of their measures.
- OpenAI's ChatGPT was able to generate campaign materials targeted at specific demographics, despite usage policies that prohibit such actions.
- The article suggests that Google and OpenAI need to do more to prevent AI abuse in the upcoming 2024 Presidential election, as the potential for AI deepfakes to disrupt the democratic process is significant.