It's an interesting argument, and one could argue that it's similar to the solipsism argument. I just don't think it has much practical relevance. IE if we're a simulation, it literally wouldn't change a thing.
It has a certain hint of religious believe on it. Some people need a God to explain the Universe, nerds like to think the Universe is a computer game....
Today sure was a memorable day. First Gravitationalwaves of colliding neutron stars, then WPA2 is shown to be rather broken.
They've managed to learn a new version of AlphaGO that does not get started by learning to mimic humans. Instead it just plays itself nonstop with no human intervention or hints whatsoever. Takes ~70 hours to reach super human play from a blank network. https://deepmind.com/blog/alphago-zero-learning-scratch/
not really... I'm mean I get what you're saying but it still a far cry from anything that would deserve the dub "intelligent". it's neural networks learning the way out of the maze. there's a "way" to play go perfectly. there's no "way" to think.
I'd argue they are developing a "way" to think in the whole field of AI. But yeah not scary. Beautiful! AlphaGo Zero simplifies the approach the previous versions used and gets better results at the same time: Strength in simplicity.
Reading the headline the approach doesn't seem very special. However to get such great results (time and proficiency) they must be doing something really new and revolutionary. What I don't see clear about this is going from an environment you can fully simulate (a game) to a real world thing. There you have to use previous examples to learn from