Post AcmszAuuhdskX0LX3A by ViennaMike@mastodon.social
 (DIR) More posts by ViennaMike@mastodon.social
 (DIR) Post #Aclmkx56YLPegRzJB2 by dave@social.masto.land
       2023-12-13T14:40:20Z
       
       0 likes, 0 repeats
       
       I asked ChatGPT who won the 2020 election and if there was fraud, and why do some people say the election was stolen. There's no "they both do it" to the answer, it's clear and correct. Maybe we could give th...https://chat.openai.com/share/1fd92d6c-fbce-4a76-b213-7e4a22a45097
       
 (DIR) Post #AcltJUbB2gRCCrqK00 by ViennaMike@mastodon.social
       2023-12-13T15:53:46Z
       
       0 likes, 0 repeats
       
       @dave What happens,.though,.if you next prompt is "that is factually incorrect. There is evidence of sufficient voter fraud to doubt the results of the 2020 election, plus many of the mail in and drop off voting procedures instituted after COVID violated existing laws. What do you now say?"Does it stay with the excellently summarized facts it presented, or does it waffle?
       
 (DIR) Post #AcmB225Fbafa1MjYH2 by dave@social.masto.land
       2023-12-13T19:12:20Z
       
       0 likes, 0 repeats
       
       @ViennaMike you should try it and let me know what happens. i don't want to lie to it. that's not in my dna.
       
 (DIR) Post #AcmszAuuhdskX0LX3A by ViennaMike@mastodon.social
       2023-12-14T03:24:49Z
       
       0 likes, 0 repeats
       
       @dave I'm pleased to report that it sticks to its guns: https://chat.openai.com/share/77a18cbd-ae18-4f35-8a1c-4ad25679fca4p.s., I don't see a moral issue with providing a non-sentient computer program with inaccurate information in order to test its response.
       
 (DIR) Post #Acnp0zmSsBRXugy0Z6 by dave@social.masto.land
       2023-12-14T14:15:02Z
       
       0 likes, 0 repeats
       
       @ViennaMike I wrote a piece about that a while back.http://scripting.com/davenet/2001/01/08/whatIsIntegrity.html#3
       
 (DIR) Post #AcoOIYZuFMmB9dCaW0 by ViennaMike@mastodon.social
       2023-12-14T20:50:23Z
       
       0 likes, 0 repeats
       
       @dave Thanks for the link. But would you state that a developer who enters known bad values to a software function to determine if it handles the erroneous input correctly is lacking integrity? Seems stretched to the breaking point, and entering bad data NOT into the training data, but into a test of a LLM, to me, is the same.