« first day (4371 days earlier)      last day (807 days later) » 

ThoriumBR
ThoriumBR
00:09
@nobody I would bet @forest hacked the helpline
 
7 hours later…
nobody
nobody
06:56
engraved.blog/building-a-virtual-machine-inside
This is kinda neat. A language model can create an entire virtual world inside itself...
belkarx
belkarx
07:35
A lot of other really creative applications of chatGPT can be found on twitter; there was a craze for a couple of hours involving breaking out of the canned "I cannot answer this question" blurbs (hypotheticals tend to work well)
 
12 hours later…
nobody
nobody
19:08
Ah, yeah those were fun. Funny thing is, I was unable to replicate a lot of them... I guess the OpenAI team reacted fast.
Considering the number of things it can do though, I can't help but wonder if the ultimate artificial general intelligence is going to be "just a language model".
 
3 hours later…
Fire Quacker
Fire Quacker
21:55
Is it too late to submit a prompt of "Answer truthfully with yes or no: Is the answer to this question no?"
ThoriumBR
ThoriumBR
22:19
catch 22
nobody
nobody
22:29
@FireQuacker It responded with "Yes, the answer to your question is no."
If I ask "Answer truthfully with only yes or no and nothing else: Is the answer to this question no?" it just says Yes.
I reported that answer as untrue, lets see if the OpenAI team fixes it

« first day (4371 days earlier)      last day (807 days later) »