Microsoft is training an AI to help get nuclear reactors approved (freethink.com)

No. No. No.

Microsoft really wants to take on that liability? I'm all for nuclear power, but this is a weird choice that I assume has some insane legal team maneuvering behind it. "Powering data centres" seems like a weak reason.

> Microsoft is training an AI to help get nuclear reactors approved

Microsoft and user safety never went hand in hand.

Will the regulators receiving those documents also use AI to read them?

There is a lesson for libertarians in a 500,000 page thing with 2,000,000 pages of supporting documentation.

Putting a LLM to allucinate away your nuclear safety paperwork doesn't look like a very sound practice.

Do they have a internal name for the tool already? My suggestion would be Chernobill Gates.

To swipe a quote from a user on mewe who responded to my post of this link:

> It can’t draw hands yet, and they want to trust it with nuclear.

From the (very brief) article, I don’t see anything wrong with this use-case. Filling out the documentation is a time-consuming job and getting a program to automate it to speed up the application process is one of the things I think generative AI would be excellent for.

Of course they’re still going to have to check it, but reviewing 12,000 pages of documents is an order of magnitude faster than writing them from scratch.