20 Apr 2023

More AI crappery - NOT amateur radio

As an experiment just now I asked Chat GPT4 about my son, who is a professional jazz pianist. I tried about 7 times. The answers were all correct in parts, but every one was totally wrong on when and where he was born and many other details. 

My worry is people will believe that AI always is factually correct, when it isn't. 

Yes, I am sure a great deal of good can come from AI, but also it could result in a lot of incorrect data being assumed as correct.

At the moment, Chat GPT4 gets about 3/10 from me and "must try harder".

Artificial intelligence (AI) is coming, but it still has some way to go. I wonder what AI will be capable of in 10 or 20 years' time.

1 comment:

  1. Hello Roger, it is possible to learn chatGPT the correct answers. Actually it is a little child that thinks it knows everything and has a lot of fantasy, but it is learning fast. In 10 years chatGPT will be grown up and knows more as a human being ever can. Could be a dangerous situation, time will learn. 73, Bas

    ReplyDelete