Startup Uses GPT-3 to Provide Counseling and Then Realizes It 'Feels Weird'
vice.com
vice.com
> “Koko Bot connects you with > real people who truly get you”
> We were not pairing people up to chat with GPT-3, without their knowledge.
> Once people learned the messages were co-created by a machine, it didn’t work. Simulated empathy feels weird, empty.
We have a vending machine in the lunchroom that says "Thankyou!" in breathy broken English. I am not impressed and have stopped using it. Consumming less junk food is probably better for my physical health as well as my emotional health.
Other times, it helped me to explore relationships in my life. In that way it worked a little bit like free-writing, but an interactive / guided exercise. It turns out that a little objectivity and reflection was all I needed.
Is it a substitute for a human? No. Then again, are you volunteering to listen to me complain? I don't want to abuse my limited friendships and I'm not going to pay someone.
My frustration was with the limitations of the AI, not the fact that it was a bot. If it remembered my past conversations, built a profile of my personality and preferences, and wasn't such a stick in the mud regarding policy then I suspect the companionship would be socially rewarding, at least to me.
> Is it a substitute for a human? No. Then again, are you volunteering to listen to me complain?
People at 7cups.com are volunteering to listen to you complain. As much as you like, on any topic (with obvious caveats around abuse, etc.).