A recent experiment showed that any AI can be ‘turned to the dark side’ simply by reframing its goals. It’s surprisingly easy – and that points to a big problem.
If you’d like to read the paper – published in Nature – click here.
Author mpescePosted on