I was talking to a friend who moved from the Google Brain team to the DeepMind team and he said point blank that no one was going to reproduce their work and that they were not going to reproduce other groups' work. Everyone has their own research agenda and the cost of the hardware and compute time needed to process the data is available to a tiny number of research groups around the world.
> the cost of the hardware and compute time needed to process the data is available to a tiny number of research groups
You'd think this would make them more interested in replication, since otherwise the likelihood that they are actually doing something of value is uncertain.
Science and research is full of all kinds of false starts, you don't increase your speed by blinding yourself.
