Live data from Hacker News

Show HN: Humans vs AI – A/B testing GPT-3

vwo.com

41–50 of 80 posts

Re: Show HN: Humans vs AI – A/B testing GPT-3

#41

For all the texts generated by GPT3, is anyone verifying that it's not just copy pasting paragraphs from previous seen texts? (like by, searching n-grams or even just googling it?) If not, then its pretty easy for GPT-3 to copy paste existing human written texts and just prove that it can write like a human.

> its pretty easy for GPT-3 to copy paste existing human written texts

My god, they've finally worked out that humans haven't written anything original since the dawn of the internet.

Re: Show HN: Humans vs AI – A/B testing GPT-3

#42

Did anyone tried to develop a scammer time wasting service with GPT-3, something like https://spa.mnesty.com/ ? Ideally it should also support hangouts because the romance scammers that spam me almost every day with a new address always ask to move to an hangout conversation in their initial message.

Just saying: I'd subscribe to someone's patreon if they promise to develop this

We could use GPT-3 to write the campaign, I totally promise to deliver within 3 months and you'll get a t-shirt!

Re: Show HN: Humans vs AI – A/B testing GPT-3

#43

Earlier quoted context omitted.

Just saying: I'd subscribe to someone's patreon if they promise to develop this

We could use GPT-3 to write the campaign, I totally promise to deliver within 3 months and you'll get a t-shirt!

oh my god, gpt3 for investment pitches

Re: Show HN: Humans vs AI – A/B testing GPT-3

#45
post #19

Earlier quoted context omitted.

Actually GPT-3 detectors can be built https://www.technologyreview.com/2019/03/12/136668/an-ai-for... But it'll detect human written one too. So expect many false positives.

return true;

No false negatives, woo!

Re: Show HN: Humans vs AI – A/B testing GPT-3

#47
> A/B tests in progress until they reach statistical significance.

Seems to imply that there will be a difference (by rejecting the null hypothesis). But not rejecting the null should actually count as a win for GPT3.

I don't want to think at the implications if it turns out GPT3 is better.

Re: Show HN: Humans vs AI – A/B testing GPT-3

#48

> A/B tests in progress until they reach statistical significance. Seems to imply that there will be a difference (by rejecting the null hypothesis). But not rejecting the null should actually count as a win for GPT3. I don't want to think at the implications if it turns out GPT3 is better.

Yes, agree. That's a good point.

Re: Show HN: Humans vs AI – A/B testing GPT-3

#49
post #37

For all the texts generated by GPT3, is anyone verifying that it's not just copy pasting paragraphs from previous seen texts? (like by, searching n-grams or even just googling it?) If not, then its pretty easy for GPT-3 to copy paste existing human written texts and just prove that it can write like a human.

The question is, is this different from how humans learn to write and speak a particular language?

Yes. This time a machine is doing it, therefore it doesn't count as intelligence.

Re: Show HN: Humans vs AI – A/B testing GPT-3

#50
post #21

I'm a little disappointed that the landing page for this challenge itself doesn't seem to be a live a/b test with live results for how much submissions are garnered through the human version of the page versus the gpt-3 one. Fun challenge though !

Good point. Just launched an A/B test on the page using GPT-3 generated suggestions: https://soapbox.wistia.com/videos/TRn6lQiVhU Let's see which one gets more participation :)

Whao, that's pretty cool ! I'm so curious of the results now ^^
Post reply on HN