Today I applied for beta Access to the GPT3 general purpose text processing API from open AI.
It produces shockingly good results.
https://boingboing.net/2020/07/17/the-new-gpt-3-powered-version.html
I believe that GPT3 is a threat to BLURT.
Here is how I propose to respond to this threat:
- Kill the current reputation system
- Implement social proofs
- Think of identity as a gradient ranging from non-confirmed to fully-confirmed real-name ID
- Never require users to confirm their identity, ever
- Provide tools for strong, robust verification of identity when that is what users desire
rid of any reputation system is a gd idea cause its very subjective hence unfair. would not bother about id verification since even those may be spoofed/hacked and sounds like presenting a big bro, virtue signalling etc anyways, not blurt-like imho. cetrainly agree no id requirements ever! what do we think?
It's a threat to all blogging systems really. But even if it riffs really nice it's not going be able to write stories that go with picture sets and videos and other things indicating a human is behind it.
Do you think this GPT-3 can pass the turing test?
No, apparently what it does really well is mimic writing styles.
just another thing for SEO spammers I guess.. I really can't see spamming happening anymore here since the fees.
another comment if i might, does it actually matter if behind a post is a machine or a human? if the post is good then enjoy it or learn, if one thinks it is bad then so be it. both machines and human can produce gd and bad, real and fake etc. always trust but verify makes sense? i.e. gpt3 is no threat imho
Phew, that is complex.
You're right, but I don't like that you're right :).
dont like the robots also but we prob cant stop them, prob just allow us make our own human evaluation of robots output on case by case basis.