There is a game called telephone that I always play with my friends. The rule of the game is the first person is given a prompt or story to relay to the next person and then that person relay the info to the next person and so on until the last person. When it reaches the last person, the result is then compared with the prompt that was first given. I wanted to apply this game to AI and test how well AI can relay information to each other and generate stories and images based on the information received.
I created an application called telephone, in which you enter a prompt for the AI in which it generates a synopsis for a movie based on the prompt that was submitted. Then based on the synopsis the AI created, it would then send to another AI in which it would create a movie poster based on the synopsis provided by then ai. Next, the movie poster would then be sent back to a third AI in which it would create another synopsis based on the movie poster it was provided with.
One interesting observation I found is that the AI for ‘Story From The Image’, is very good at describing what is included in the image but not so much at creating an original synopsis by itself. When creating the poster for the synopsis, there were many issues with copyrights in which certain keywords cannot be included in the prompt that is given to the AI for generating movie posters. Currently, artificial intelligence is still in its infancy, given time AI could possibly mimic the imagination of a human being.
Top comments (1)
Interesting exploration. Given the rapid development of AI, I think it will be interesting to see how your tests and results progress over time, will they improve in accuracy and creativity I wonder.