There’s an interesting approach used by DARPA:

1. Prove its not impossible

2. Prove it’s possible

3. Do it