Comment by lucidbee

I worked as a principal engineer in an AI company until a year ago and I was impressed at how hard it is to get models robustly trained. They are fragile in real world contexts (in the field) and training is full of pitfalls. I have heard so much marketing enthusiasm but the real world situation is different. Some fundamental advances are not even in sight yet. We don't know what we are missing. My view is we don't know yet whether the singularity is possible and have no idea when it could arrive.