Hello and welcome to Eye on AI. In this week’s edition: The difficulty of labeling AI-generated content; a bunch of new reasoning models are nipping at OpenAI’s heels; Google DeepMind uses AI to correct quantum computing errors; the sun sets on human translators.
With the U.S. presidential election behind us, it seems like we may have dodged a bullet on AI-generated misinformation. While there were plenty of AI-generated memes bouncing around the internet, and evidence that AI was used to create some misleading social media posts—including by foreign governments attempting to influence voters—there is so far little indication AI-generated content played a significant role in the election’s outcome.
That is mostly good news. It means we have a bit more time to try to put in place measures that would make it easier for fact-checkers, the news media, and average media consumers to determine if a piece of content is AI-generated. The bad news, however, is that we may get complacent. AI’s apparent lack of impact on the election may remove any sense of urgency to putting the right content authenticity standards in place.
