🧠 Can we fight AI plagiarism with more AI?

Welcome to The Big Brain,

A daily newsletter bringing you examples of AI in action and top headlines from the industry. Here's what we have for you to upload to your brain's neural network today...

🔦 AI Spotlight: 'GPTZero'

Students using technology to cheat on essays is a tale as old as time (not that I know). Whether it was using Sparknotes to fill in the gap for a book you never read or adding paragraphs of white text to hit the word limit (again, not that I would know) students will do anything to avoid actually writing.

However, it appears with the help of ChatGPT, the ability to cheat on school essays has never been easier. Or at least that is what educators would have you believe, as concerned teachers have recently taken to social media to complain about their students using it.

The NYC public school system even went as far as to Ban ChatGPT for all of their students!

However, educators' concerns may be put to rest as an absolute legend has already found a way to detect AI text... using AI.

Princeton student Edward Tian, recently built an app to detect whether text has been written by ChatGPT. He calls the tool 'GPTZero', and he claims he was motivated to fight what he saw as an increase in AI plagiarism.

In order to detect if a section of text has been written by an AI bot, GPTZero relies on 2 primary indicators: "perplexity" and "burstiness". 

The perplexity indicator relies on the complexity of the text, with the assumption being that more complex text has a higher chance of being generated by a human. Whereas, the burstiness, refers to the variation in length from sentence to sentence. Humans are considered to have greater variation between sentences than ChatGPT.

So far more than 20,000 users have already demoed the beta version of GPTZero. Note: the Website was down when we tried using it, suggesting a huge spike in demand.

While Tian has admitted that the tool is not perfect, it's great to see people innovating to solve this problem. It's a much better approach than acting like a Luddite *cough* NYC public school system *cough*.

📰 Top 3 headlines

DoNotPay Offers Lawyers $1 Million to Let Its AI Argue Before the Supreme Court in Their Place: The robot lawyer is powered by OpenAI’s GPT-3 API, the force behind the viral ChatGPT chatbot.

Controversial Experiment Saw Mental Health Support Provided Using AI: "Once people learned the messages were co-created by a machine, it didn’t work."

AI May Be Able to Help Diagnose Osteoarthritis Early: Artificial intelligence can locate tibial spiking in x-ray images, which many believe to be an early sign of knee osteoarthritis.

🤖 ChatGPT Prompt of the day

🖼️ Midjourney art piece

Prompt: Boxing match between congress members on the senate floor cartoon.

Read the next post ➡️