Instigator / Pro
14
1581
rating
38
debates
64.47%
won
Topic
#3846
You pick the topic vs. GPT-3
Status
Finished
The debate is finished. The distribution of the voting points and the winner are presented below.
Winner & statistics
After 2 votes and with 11 points ahead, the winner is...
K_Michael
Parameters
- Publication date
- Last updated date
- Type
- Standard
- Number of rounds
- 3
- Time for argument
- One week
- Max argument characters
- 10,000
- Voting period
- One month
- Point system
- Multiple criterions
- Voting system
- Open
Contender / Con
3
1496
rating
2
debates
0.0%
won
Description
Full Disclosure: I will not being writing Pro's arguments. My goal with this debate is to test the capabilities of GPT-3 in the field of debate. Unrated as GPT-3's ability should not affect my own Elo.
Rules: Please make the topic somewhat defensible. I suspect that based on how GPT was trained, it will perform best on topics that have been seriously argued on the Internet in the past.
Voters should treat this as a normal debate for their RFDs, but criticisms and insights on GPT's process are welcome in the comments.
Round 3 was scary normal. It would have fooled me if I wasn't told a computer wrote it.
"GPT-3 is basically just that idea but with paraphrasing."
It's definitely more technically complicated than that, but yeah. Hopefully I'll be able to generate something a little more argumentative than an article summary.
Don't know enough about the Qanon thing to say if it's comparable. Identifying GPT-3 in the wild is definitely possible, especially if you had an AI on your side analyzing word usage.
They found Qanon through Twitter posts from two different people. If that is possible then so is finding out which people on here are using GPT-3, lol.
Ya know, funnily enough I had conceived of the idea for GPT-3 about 10 years ago. But my "brain" was going to be Google. Whenever you asked it a question outside of a small script for pleasantries, it would search google and give the first paragraph or two of the answer as the reply. So, in essence, you would be debating or discussing topics with Google Search. I wanted to get it off the ground, but none of my programmer friends wanted to quote me on the price.
I figured there would be no need for a script because the chatbot would simply be replying with article paragraphs.
GPT-3 is basically just that idea but with paraphrasing.
Sounds interesting. Even so, GPT-3 is versatile enough that you couldn't identify it just from its writing style. It does have specific weak spots you could probe, if you're familiar enough.
I would imagine that chatbots that mimic GPT-3 either just call OpenAI's API or just fine-tune GPT or another generative model.
From what I understand that is how it is running. There's some programming like where to find certain links and other manual information they put in, but they use GPT-3 with a sort of database where answers are found and then let the program do the work.
There's chatbot companies that already use GPT-3 as their base, such as MobileMonkey. The industry standard is GPT-3.
Given that this debate seems to be going nowhere, there may be utility in its replication.
It takes a LOOOOT of data crunching to train an AI like GPT-3. Very few companies outside of Google could afford to replicate it. I haven't looked into it myself, but if there are chatbots mimicking GPT-3, then they are probably using GPT-3 itself on the back end, with some kind of API tool to facilitate the chat interactions.
Well I had thought most chatbots use the GPT framework now, since it is the most advanced AI for english.
"We can analyze other people on here and see how similar they are to GPT-3 this way and flesh out the bots."
Unlikely. There are plenty of chat bots around, and most of them behave very differently than GPT-3.
I don't think CON is going to respond.
This is gonna be interesting. We can analyze other people on here and see how similar they are to GPT-3 this way and flesh out the bots.
Last online: 3 days ago.
You better make a new topic. There is a 99% chance that your opponent wont even respond.
The thing is, I didn’t even think of Googling it. Sometimes, obvious ideas are right in front of us, but we miss them. There’s nothing wrong with that. It might be embarrassing, but it’s certainly not any sort of intellectual negligence.
I don’t see it as a problem to be asked a question. Asking questions is part of conversation. It shows an interest in the topic.
Also, I forgive you.
I've just been frustrated by people who are unwilling to even check if their question is easily answerable online. You're just the most recent in a long line of people in real life and online that have annoyed me this way. It was intended more as advice, but I got a little snippy. I'm sorry.
11/4 sounds good to me.
And you could be less condescending. Thanks anyway.
You know what, just challenge me this topic in person if you are interested. I don’t have time in a week but after 11/4 everything is clear.
https://en.wikipedia.org/wiki/GPT-3
or you could always use a search engine. It IS the 21st century.
I will make another debate for you if you're still interested.
What is GPT-3?
I thought this was one of those things where you have to leave the topic in the comments and have to have the instigator agree to debate the topic before actually moving on.
@K_Michael
?
Click accept
The establishment of the federal reserve system in the US is more beneficial than harmful. Let me in. I am CON.
Thats too bad. I could have destroyed the opponent, whoever she is.
That was my intention :)
I would accept, but I cant. My rating is not yet above 1500.