We keep hearing about all these AI discoveries recently. But is it AI or is it ML? What's the difference?
AI is when a machine thinks for itself outside of its programming. At this time, we do not have true AI. We have ML. ML is where large data sets are used to train and improve the responses the system generates. The system has been trained how to grow its knowledge. It is acting within the scope of its programming. However, there have been significant advances with how the solutions learn. Large amounts of data are continually fed back in, so the answer you get one day may be a different answer then the answer you get 30 days later.
The solutions accuracy is going to depend on the data it is being trained with. Garbage in will mean garbage will be put out. If the solution is trained on raw data from the internet, accuracy may be an issue. However, it is the truth according to the system, so you will get the confident wrong answers. ML also does not understand human boundaries. The things we inherently know by being human. We know that as adults are bodies will not continue to grow taller. However, if this is missed in a constraint when the ML system is being trained, it may think that adults can grow to 20 feet tall.
What does any of this have to do with cyber? ML and AI if and when we finally get there, will drastically change the cyber threat landscape. We don't yet know all the things that will happen with it. With some of the new chat AI programs, users can ask it to create a malware. As long as its done within the constraints of the programming trying to prevent users from using it for illegal actions, the code will be handed to a potential attacker. However, as time goes on and more and more solutions come online the ML can start inventing new attacks that we have not even considered. It will be an interesting future!
Comentários