it does but only if it is originated from a intelligent programmer. AI is as smart as the people who did its modeling. an artificial intelligence cannot be more intelligent than its origination.
presuming AI will be helping us securing cyberspace is like presuming we will have a accurate Accounting system or flawless GL just because we purchased a new fancy accounting software.
Artificial Intelligence is a product of human, it is result of how a programmer has defined the way Machine can Learn from patterns, so it is still a system limited to the creativity and imagination of a programmer. AI with its underlying ML is certainly depending on algorithms and logic of originator.
Most of products in the field of information security in the market today are not even close to an AI system. even if they were, still a system is as intelligence as the originator. do not confuse intelligence with other factors like speed, accuracy and precision of calculation, of course computer can be faster in calculation, but ability of a machine learning logic in terms of analysis is limited to the ability of programmer of the system.
machine learns the way we program it and will not have the complex multi dimensional thinking capability of a human unless we program it exactly the way human can analyze situations. but before that. a programmer shall first understand the problem andt has the ability to solve the problem. for example, how can I write a code to solve x in Algebraic equation x^2+2x+8=76 if I am not able to solve it on the paper? can I solve similar problems if I only memorized a particular answer?
one of the main reasons basic detection systems are not capable of finding security vulnerabilities in a network is that code cannot go further that what programmer defined for it, so how do we expect a detection engine understand a security issue when originator (programmer) did not understand the matter to begin with?
AI and machine learning are no exception, they are piece of code just like your first “hello world!” program. we cannot expect a “hey how’s it goin’?” be printed without an extra line of code with logic on when to print on screen.
we cannot expect AI solve security issues (detection for example) when original programmers do not understand the issue. you will find tons of security programs or “solutions” which do not solve simplest infosec challenges because just because programmer did not know how to approach the problem like how to analyze a common attack pattern.
I think we still need to remind ourselves that computer is not intelligent, not more than the code running on it programmed by a human with limited intelligence, and it does not matter how fancy is the term behind that piece of code like Artificial Intelligence, or Machine learning.