Right now, OpenAI's GPT detector is 26% accurate. Even if you got it to a level that got it right 70% of the time, it is still not what we can rely on. And I think people are worried about the wrong thing. Cheating already happens. It's ubiquitous. In every study we have, students cheat. There's 20,000 people in Kenya writing essays for students. Like this is happening everywhere all the time, right? The value of homework is dropped as people cheat on it. So sure, it might be helpful, but at scale, no.
ChatGPT, the artificial intelligence chatbot capable of generating human-like text, seems to be everywhere. But how trustworthy are these tools -- and what do they mean for the future of writing and work? Adam brings AI entrepreneur Allie Miller and innovation and entrepreneurship professor Ethan Mollick to discuss the capabilities of ChatGPT, debate its merits and downfalls, and ponder what we should -- and shouldn’t -- leave to AI. Transcripts for ReThinking are available at go.ted.com/RWAGscripts