hplus0603: We have algorithms that can solve problems (even complex problems) once the problems are understood and formulated.

that`s a good point

hplus0603: What we don't have, is a solid language for expressing “universal” problems to computers, nor do we have algorithms to create such expressions from detected world stimuli

` language for expressing “universal” problems to computers`, that sounds like we need to invent a separate language (other than what we currently have) that deals/operates with new concepts a supposition that IMO is false. C++ is just fine. The idea itself of `universal problem` is misleading because it throws the issue into unstable sands/unknown waters. There is no difference between the problems a terminator like robot would have to solve and the problems your laptop is currently capable of solving. It`s all about figuring out what goes into the decision making process in a given situation.

PS. An object is `universal` if you can list at least two items that it can be used for, an universal screw drive is universal because it can be used for a variety of well defined types of screws. List me two types of problems that a computer which understands `universal problems` should be able to solve.

JoeJ said:
the situation itself needs to be mapped

mapped in terms of what? if we take a random location, any object at that location could mean a million things and could be there for a million reasons. So we`re narrowing it down based on the reasons we are there. In forensics when the crime scene is mapped and interpreted what they`re trying to do is to find out `what was in the mind of the killer`, they are not trying to figure out what the objects on the crime scene meant for a guy and a girl thinking about having sex found 100 meters away. When some terminator like robots are on a battlefield they will map the battlefield area based on the `enemy profile, what the objects mean to the enemy. If the enemy is solders on foot some objects may provide cover from small arms fire etc. , if it`s tanks those objects will change their `meaning` to something else.

Terminators be like

The military type terminator is the teaching book example because it`s the simplest to build. IRobot type/gentle terminators are harder to build. To operate in an urban area they need to be safer/more reliable and operate longer times. One problem is that in order to be safe they need to take into consideration cooperation between people. A terminator needs to mirror only objects in the subjects mind. IRobot needs to mirror in peoples mind not just objects but also other people. example: Anny, John and an apple. IRobot imagines Anny noticing the apple, IRobot also imagines John thinking about Anny noticing the apple. Terminators for peace times aren`t quite the teaching example. Setting loose 50-100 kg of metal that moves on its own in a populated area isn`t something to be treated lightly.

JoeJ said:
Anny might then go over the street to grab the apple, but John sees a bus coming, and he might decide to save her

Or if there are only Anny, IRobot and the apple across the street. If IRobot has the job of taking care of Anny he might take Anny by her hand knowing she might act foolish because she likes apples a lot. But a typical scenario were `terminator` mind would apply is predicting the trajectory of people walking on the sidewalk. Like if a person is carrying an object taking a lot of space in one direction the `mind` will foresee the trajectory of the people coming from the opposite direction.

I often think about constructing a simulation where such things can happen

that will most probably turn out dull (there needs to be a fun factor) My goal is to build a RTS with an AI that can read your mind while you are playing against him. That`ll take some time though as I`m quite slow at building a conventional RTS (which is a prerequisite).