The tool to outsmart ChatGPT
The Future. It’s no secret that AI-generated text is blowing up. Everyone from students to job applicants to employees has been wowed by the power of ChatGPT. But now, yet another tool has emerged — engineers are developing ways to detect text written by AI. Looks like you may have to write that history essay after all.
Engineers are creating tech to screen for ChatGPT-generated text.
- OpenAI has its own GPT-2 output detector model, which reportedly has a “relatively high detection rate.”
- AI scientists at the MIT-IBM Watson AI Lab and the Harvard Natural Language Processing group created GLTR (Giant Language Model Test Room), which can detect if the text was written by a bot.
- GPTZero, built by 22-year-old Princeton senior Edward Tian, can measure randomness in sentences and calculate the probability that the text was written by ChatGPT.
- And even anti-cheating software Turnitin says it’s focused on detecting whether students are using ChatGPT for assignments.
Meanwhile, OpenAI is looking into ways of watermarking ChatGPT text with “an otherwise unnoticeable secret signal in its choices of words.”
What does this mean for the AI-generated text world? Well, in the education realm, teachers may be heaving a sigh. After concerns over students accessing ChatGPT during school, NYC banned access to the program, citing “negative impacts on student learning.”
Still, things might not all be smooth sailing from here. At the rate at which AI technologies are progressing, who knows what will come next. Buckle up!