Properties of intelligent software depend on multiple parameters such as agents.software ls can be intelligent agents based or agent enhanced or multi agent. kindly refer http://groups.engin.umd.umich.edu/CIS/course.des/cis479/projects/agent/Intelligent_agent.html to get info about intelligent agents properties

Intelligence, as I'm told, is the ability to explain evidences. Off the calf, explanation can be logical or a plausible cause-effect chain of events. The simplest mechanism for such explanations would be, I guess, crowd sourcing and pointing on similar past events, or a little more "cooked" solutions such as summing up events' common factors or association parameters.

Other methods of deriving causal factors would require in addition to digging up previous similarities also analytic and computer performances.

The following elements may then indicate in my view the level of potential intelligence of a software:

- relating to past events,

- practical definition of similarities,

- recognition and association,

- rules base which can be easily accessed by the user,

- identification of patterns of behavior,

- hierarchical clustering,

- learning and adjustment mechanism,

- retrieval driven storage structure,

- fast I/O and decision templates for real-time operations,

- intertwined user assistance,

- automatic adjustment to a variety of preconditions,

- self diagnostics in case of failure,

- self correction,

- intuitive user communication and activity traceability,

- being a swift-small-smart system :)

yes Edith Ohri, the properties mentioned in your reply need support of agents in software models. Through agents all the desires intelligence can be achieved,

We say that a software is intelligent if it sucess the Turing Test http://en.wikipedia.org/wiki/Turing_test for the task concerned.

Hello Benammar, thanks for the reference. Yet one must consider the long time since Turing laid down the foundations. In today's practical data mining the aim is to surpass human understanding and be able "to find hidden phenomena".

Who said that computer intelligence should be limited to what human can express?

Hello Edith Ohri,

What I've said is what I've learned in the university (in the course of Introduction to AI) where the teacher have presented to us 4 definitions of AI and said to us, what I've inderstood, that we say that a system is intelligent if it sucess the Turing Test. (this defintion exists since the 50's)

And for me, what you have said is more and more logic and I hope that the classical definitions will be updated for new learners.

Thank you.

Best regards.

Benammar, thanks.

It looks to me that the question was about the SW qualities that are needed to produce AI results, not the test. Anyway, it is interesting to continue along your answer:

If computer intelligence is not bound by human intelligence, then it means that

(a) the Turing test needs to be reviewed for the general case.

(b) a human can in principle create a machine more intelligent than himself, hmmmm...

How can such a thing happen?

Intuitively it is impossible that a product will be more intelligent than its program. But, machines have an advantage in searching new COMBINATIONS of rules and procedures, some of which may be better. So, although we set up the machines, their potential for reasoning exceeds ours.