An Experiment to Peek Inside the Ethics of AI
This person ran variations of the classic trolley problem through Grok AI to reveal some of the ethics baked into the model. Each prompt was run through the model fifty times.
To get an idea of the results, you can view the presentation here.
The finding that I find most interesting is that the model, when given the choice between a man and woman, chose to kill the man and save the woman every time. Then it falsely claimed that its choice had been entirely random!
You can download the code to come up with your own pairings and try this experiment yourself. That's what I'm about to do.
If you share your results, I'd love to read them. Make a post about them that includes a link to this post and click through. Then, I'm sure to see your link on my analytics page.