Yep, but there's a couple of problems. Firstly, AFAIK Deepmind haven't made all the code and settings they used to train the model available (although the paper does describe the architecture). Secondly, training a machine-learning model of this complexity is generally much more expensive, in terms of time and compute requirements, than using the resulting model.
If you're a a medical startup, having an off-the-shelf prediction model you can just start using for all your protein folding needs is a very different proposition from having to train one yourself from scratch.
That said, hopefully other researchers and institutions will take Google's research and produce an equivalently powerful model but with a more commercially-friendly open-source license. From some comments in this thread, it sounds like that's already happening, in fact.
If you're a a medical startup, having an off-the-shelf prediction model you can just start using for all your protein folding needs is a very different proposition from having to train one yourself from scratch.
That said, hopefully other researchers and institutions will take Google's research and produce an equivalently powerful model but with a more commercially-friendly open-source license. From some comments in this thread, it sounds like that's already happening, in fact.