Changing Algorithms compatible with AI?

[FallenApple]

Many have argued using a Godel diagonalization argument that there is no program that can tell ahead of time that the Turing Machine would halt. But would the way to get around this is to have a continuously, changing algorithm in response to learned input? I probably should say rapidly changing, since continuously is impossible.

 

[mfb]

If the algorithm changes according to deterministic rules, that just means these rules are the actual algorithm and what you call the algorithm is part of the memory.
No.

 

[QuantumQuest]

Additionally to what mfb said, you can go the classic way of the halting problem and see what you can feed a program P that allegedly can tell ahead of time that a TM will halt and render it useless.