Not to mention recent advancements in deep learning / machine learning with neural nets. It seems like that field has really proliferated over the past 5ish years. (I'm not sure what breakthrough specifically has led to all this, but we have TensorFlow, Torch, etc.)
If you've checked out recent papers in machine translation with deep learning, they're about as good as current products with supposedly just a week of unsupervised training. So it's really going places.
Of course, the demos (http://104.131.78.120) fail all the time because they haven't done the work to feed in the whole internet and handle proper names and stuff.
I feel that he is right being paranoid, because it is not impossible to build an AI, what his goal is to build an open AI so an AI isn't in the hands of a private organization, who knows what they can do. That being said, it is a long time before we have an actual ASI, but its great that we have the open AI initiative, so the AI itself is built in the open, or squarely in the open.
1. Are the any laws of physics that stop us from creating AI ? Answer is No. Nature already built it so its possible.
2. Is it possible to improve AI significantly ? Yes, because if we can build AI, AI can build better AI.
3. How soon can it happen? Since the trend is exponential. It can happen within 100 years.
4. Is it good idea to start worrying about it now? Sure why not.