Theoretically, yes. But anyone who has tried to implement scientific papers will tell you it's far harder. The papers often lack critical details and implementation hacks: all the little rough edges that go into making a production system work. They also lack context in many cases, so you spend more time reverse engineering the paper than figuring out how to make it work.
Last time this was brought up, someone implied it's closed source so as not to boost the Chinese competitor.