In other news, OpenAI's ChatGPT Search was found to be unreliable for short queries, Amazon has begun drone deliveries in Phoenix, and the former head of Meta's AR glasses has joined OpenAI. Amazon has also launched an AI-powered feature that creates TV episode summaries, and Apple has acquired AI-powered image editor Pixelmator. A new research paper reveals that AI can be fooled by adversarial pop-ups, and Scale AI has announced Defense Llama, an AI model for defense applications.
Key takeaways:
- The author was able to create a convincing deepfake of Kamala Harris' voice using Cartesia’s Voice Changer for just $5, highlighting the potential for misuse of such technology.
- Despite a requirement to agree not to generate anything harmful or illegal, there are no real safeguards to prevent misuse of voice cloning technology.
- Disinformation is spreading at an alarming rate, with the volume of AI-generated deepfakes growing 900% between 2019 and 2020, according to data from the World Economic Forum.
- Amazon has launched “X-Ray Recaps,” a generative AI-powered feature that creates concise summaries of entire TV seasons, individual episodes, and even parts of episodes.