How are algorithms created?

There are two types of algorithms. First are pre created by humans and fed into machines that perform different tasks based on these algorithms. These algorithms consists of different routines and modules that call each other and work in collaboration with each other to perform different tasks.

The second type of algorithms are evolutionary algorithms. I have heard about them but was unable to find a credible source to quote here. What I understand from them is that they learn by themselves and improve themselves without human interventions. They are the future in my opinion.

What factors decide how to teach a computer to do disaster recovery? 

There could be a number of factors that influence the decision to teach computer to help in disaster recovery. First is the fact that humans do not have the physical capacity to do so, it becomes necessary to teach computer to learn disaster management in such situations. Second, in my opinion is the fact that computer might perform well in making certain decisions compared to humans. For example humans have emotions, fears, likes and dislikes that may limit their capacity to respond objectively while computers do not have these issues, yet, so their responses would be objective and calculated.

How do personal assistants learn?

When we talk about personal assistants in the AI context, I think that we should distinguish between personal assistants that help us in performing different computer or other “soft” tasks and those who might help us in performing our physical tasks, such as those shown in the iRobot movie starring Will Smith. How they learn is based on their capability to understand our computer usage habits and identifying the areas that we might need assistance with. These assistance are programmed to present us with plausible solutions. In case of the iRobot king of assistants, that might be a part of the human future, they would learn from our gestures, verbal and non-verbal cues and from our emotional responses. They would then translate them to algorithms and learn similarly.