With the speed they work, and of course the proper software, they will be able to peruse, weigh, and evaluate every possible response and make one that simulates intelligence.
As I said, we might not consider it “real” intelligence, but it will be so convincing, the difference may be irrelevant.
> evaluate every possible response
I disagree because in my experience they aren’t creative - they can only do what they were previously programmed to do. Even self modifying software is extremely limited in what it c an accomplish. Neural Nets are limited to whatever the training data contains - it can only come up with variations on what it is fed - nothing fundamentally new. Furthermore, they cannot ACT on what they process without getting a human involved.