By definition a simulated brain is intended to be exactly as smart as the meat brain being simulated. Once you get the simulation working you might be able to run it faster though.

If a simulated brain would be even remotely comparable to a human brain in terms of computation it would be immediately the smartest thing on the planet simply by virtue of having immediate or near immediate access to all recorded knowledge.

But we might not need to simulate anything..

In the end it’s quite possible that AGI would end up being brains in vats with some machine brain interface for IO.

I literally might be more cost effective to do it this way.

And this is what people that dismiss AGI often miss. They assume the advancement would only come from one place and that programming would be the only way to solve the problem.

Analogue computers are making a comeback in many areas, and an analogue computer for an AGI might just be a bunch of brain tissue rather than some neuromorphic chips or some other nonsense.

If we get to a point in which we can clone meat for food and organs for transplant making a brain isn’t out of the realm of possibility if we can get BMIs working then a lot of the problems might be solved by just leveraging existing solutions already.