Ask HN: Is a PhD in theoretical neuroscience or ml worth it?

I'm a master's student at Harvard, where I've been studying computational biology and artificial intelligence. During my studies, I've had a lot of intuitions about new neuroscience inspired architectures, which I write about here https://medium.com/technomancy. With the plethora of job openings and startup resources, do you think doing a 4-6 year phd is worth it if I want to advance ML and AI by building new theoretical architectures? Is sticking to openAI/deepmind/google brain/FAIR residencies or fellowships better? Or should I just say screw it an apply as a deep-tech/moonshot startup at Y Combinator