These mixed sci-fi/reality sensationalist Skynet writeups aren't very convincing. However, I agree it's quite possible we could see an AI being left "free to ro

These mixed sci-fi/reality sensationalist Skynet writeups aren't very convincing. However, I agree it's quite possible we could see an AI being left "free to roam" in an environment with poor security doing a lot of real-world damage accidentally, such as taking down a critical system or major infrastructure for an extended period. It would be a little like that story of the AI building paperclips and it's important that we try to prevent such catastrophe. Even staunch AI advocates (like me) should recognize the significant dangers of careless deployment of the technology. I like to see it in a similar way as nuclear medicine, something that can do much harm despite the extensive positive applications. I hope we can evolve the appropriate security protocols in time to use AI for the good of humanity.