Detecting AI-Generated Text. Did ChatGPT write this? We deserve to…

OpenAI is aware of the issue and plans to address it by watermarking all ChatGPT-generated text. OpenAI guest researcher Scott Aaronson said it would work using a "cryptographic pseudorandom function," where only OpenAI could access the key. So for any text that GPT generates, there would be a hidden signal in its choices of words, which only OpenAI can prove later that it came from GPT. So when this is released, ChatGPT will continue to generate text that looks random to the public, but OpenAI will be able to uncover a watermark with the secret key.