正在加载视频...
视频加载失败
here's a quick demo with john mcphee's "frame of reference"
25 条评论

I spent New Years building GPTZero — an app that can quickly and efficiently detect whether an essay is ChatGPT or human written

here's a demo with @nandoodles's Linkedin post that used ChatGPT to successfully respond to Danish programmer David Hansson's opinions

the motivation here is increasing AI plagiarism. think are high school teachers going to want students using ChatGPT to write their history essays? likely not.

the analysis is based on some ongoing research with and @sreejan_kumar and @princeton_nlp. hopefully we'll publish something empirical soon. but in the mean time this was a fun app to make :)

@sreejan_kumar @princeton_nlp in short, there's so much chatgpt hype going around. is this and that written by AI? we as humans deserve to know!

@sreejan_kumar @princeton_nlp feel free to try the beta yourself at or it's still barebones right now, but will be spending the next few weeks improving the model and analysis.

@sreejan_kumar @princeton_nlp wow! woke up this morning, and did not expect so many ppl here! if the apps' not working rn it's likely because my free @streamlit hosting was not meant to handle so many ppl. going to reboot and try again :)

thanks @DataChaz and rest of the @streamlit team for generously upping gptzero’s hosting and memory :))

You bored bored guy, I hope your code doesn’t work when you deploy it

Very interesting... But I'm worried about false positives on writers less verbally virtuosic than Pulitzer-winning John McPhee. How would it fare with an average high school essay, say?

For those who want to finish reading the McPhee article:

Hi Ed! Any chance you're free at 1750 GMT to come and talk over the phone about your app on the BBC World Service for OS Radio?

how does it fare with normal human written text? I have used a similar tool and I ran it on an AI generated article that I had gone in and tweaked, and it gave a 80% human written score. I wrote a reddit comment completely myself and it gave 78% score.

API? I'll integrate it into @Workello

Nice tool but more or less useless - we tried 4 GPT3 generated essays, paraphrased afterwards in seconds and all were classified as "human generated".

Is it accurate in Spanish? I've just made a try with an OPEN AI text about "Natural Law". The result is 165: "Your text is likely human generated!", which is wrong.

Consider using to make this demo more readable on mobile

It should display the pvalues: i.e. With a prob of 1-alpha, your text is human generated. That way, you can't be wrong.

This looks very interesting and I really like the 'reasoned' approach through which you can see how the text is analyzed. Congrats! Do you have an idea how this compares to for example @huggingface's older GPT-detector?

Help... lovin' the concept but unable to enlarge the demo enough to read it

LMAO

I love it that you used McPhee for the demo. (A very Princeton choice!)

awesome... thx @edward_the6

Nice work and great demo choice!

What, is this another instant AI?
