Well its about time AI starts fighting ransomware.
Back when Intel bought McAfee, they made some announcements about it automatically detecting hackers/malware/etc. However, the problem with any sort of automated detection is that there will always be false positives. So, what you can actually
do about it is quite limited, because you don't want to be randomly halting or killing programs that aren't true positives.
As the article says, this is the true weakness in IBM's scheme. They state that it will
detect ransomware, but then actually
doing anything about it is left up to the operator. That means you basically need skilled 24/7 operations staff that are capable of assessing the threat and taking the appropriate action, or else the feature isn't so useful. Also, the false-positive rate needs to be
exceedingly low, or else the feature will be likewise virtually unusable.
So, I'd put almost no stock in such a feature. At the very least, it needs to be independently tested, including deployment at scale.
What's more interesting is their data vaulting system. However, that's really just software and nothing fundamental to their CPUs. So, you could potentially use similar solutions from other vendors (or, for all I know, maybe even comparable open source solutions exist).