User blog comment:DWither/Doubt about busy beaver/@comment-5529393-20180904212819/@comment-5529393-20180907010528

I agree, the powers of a super intelligent AI combined with the speed, precision, and lack of errors of a computer would be an amazing sight to see. However, it still couldn't compute the Busy Beaver function. Algorithms using artificial intelligence are still algorithms, and hence the proofs that the Busy Beaver function cannot be computed by an algorithm are still valid.

There may be hope if you don't believe in the physical Church-Turing thesis, which states that the functions which are computable by any physical system are exactly the functions computable by a Turing machine. So if this thesis is true, then since the Busy Beaver function cannot be computed by a Turing machine, it cannot be computed by any physical system either. But, if the thesis is false, there could be some exotic type of physical system which is "hypercomputational", and can compute functions like the Halting function and the busy beaver function.

Anyway, the point is that this is not really about artificial intelligence. The artificial intelligence that is currently studied is run on conventional computers, which have the same computing power as Turing machines. To compute the Busy Beaver function would require a change in the basic computational model, and I imagine that a computational model that could compute the Halting function (and therefore the Busy Beaver function) could do so in a straightforward fashion, without the need for any sophisticated artificial intelligence techniques.