Janelle Shane’s AI Weirdness has posted an Advent Calendar for 2023. She did not include the aptly named icon (left) in the calendar. It was generated while she experimented with DALL-E and ChatGPT. In her comments about the AI-generated results, she noted:
The more you look at an AI image, the more you see wrong with it.
Turning this around:
AI improvements make the mistakes harder to see.
The traditional Turing Test argues that if the AI can convince a human that the other end of a digital conversation is also a human, than the AI is legitimately “intelligent.” In other words, if you fool enough people, than it is indeed intelligent, by Turing’s standard, at least.
I wonder what Turing would think of all this.
Smatters 
Leave a comment