Someone could’ve made the same arguments about automating horse-drawn carriages. Even the safety arguments would fit: going faster would be much more dangerous.

Someone could’ve made the same arguments about automating horse-drawn carriages. Even the safety arguments would fit: going faster would be much more dangerous. I think it’s a good idea to keep an open mind and see what happens. You might be right, but if you’re wrong, you’d be better served to be ready for it. Which seems more likely: that it’s impossible to automate testing with AI, or that it might be possible?