What are the implications if P=NP in Artificial Intelligence?

[Gjmdp]

If it turns out that there is, indeed, an algorithm that can solve any NP problem (now also P) in polynomial type, so that P=NP, creativity would prove worthless with such algorithm. Thereby, AI will not be needed as any intellectual inquiry we would ever had could be easily solved with this algorithm. Hence, we should try to solve the P=NP problem rather than research in Artificial Intelligence. Right?

 

[jedishrfu]

Here's a discussion on P=NP

https://en.wikipedia.org/wiki/P_versus_NP_problem

There are still NP-hard problems that aren't covered by np-complete and so creativity is still alive and kicking.

https://en.wikipedia.org/wiki/NP-hardness

 

[QuantumQuest]

If it turns out that there is, indeed, an algorithm that can solve any NP problem (now also P) in polynomial type, so that P=NP, creativity would prove worthless with such algorithm. Thereby, AI will not be needed as any intellectual inquiry we would ever had could be easily solved with this algorithm. Hence, we should try to solve the P=NP problem rather than research in Artificial Intelligence. Right?

Not so in my opinion. First I would ask you, from which perspective do you examine the impact of (a potential) P = NP?
From a programming perspective it is a search problem And it would eliminate the pains of trial and error. Is this technologically feasible in the near future?
I don't think so but even if it is, its impact on AI will be in terms of a subset. Some problems (e.g. some optimization problems, natural language processing and a variety of others) will still be open. Surely, the impact of P = NP will be huge. Programs, computers, networks and what not, will run faster. It may also bring unforeseeable improvements at various levels and sectors but at least as I see it, it can't solve anything and everything in AI.

As a second point, I would ask what would be the impact for AI at the human level? Our brain works the way it does even with approximations and it has proven to do very well. Optimized (or even better optimal solutions) are always welcome but whether P = NP or not our brain will be needed.

 

[Rive]

If it turns out that there is, indeed, an algorithm that can solve any NP problem (now also P) in polynomial type, so that P=NP, creativity would prove worthless with such algorithm.

Such math would sure turn many things upside down, but since many application of AI is not exactly about optimal, but 'good enough' responses, I think the impact on practical level would not be a really dramatic one.

Just look around: the whole NN, big data and AI stuff is about being lazy and trying to spare the effort for digging up the optimal/exact algorithms...

 

[sysprog]

What are the implications if P=NP in Artificial Intelligence?

If P=NP, and I am the first to prove that and/or find/invent a general algorithm that can exploit NP being in P, I freely promise that I will NOT publish the proof/algorithm, and I'll be careful to create a plausible fiction for how I made such a grotesquely large amount of money so fast.