Starting small with induced artificial intelligence.. the real life game approach to small assistant level tasks.

To start with this topic, I want to go in directly into a scenario which can serve as a simple case of intelligence in a program. We will talk about it and later will see how we can develop a program to solve for that scenario.

Suppose you have a group of people, let's say you have friends or you joined a new team and now you've got a discussion going on using mobile phones and group sms technology. Now consider what you have access to here is just the phone numbers flashing on the screen as soon as someone types a message and the communication text which the person sent. So in other words, you don't have any idea who the person typing is other than their last text.

Question is -

A) how long will it take for you to identify all the members of your group correctly by their texts?
B) how long will it take for you to identify at least half of the members of your group?
C) would you be contributing in any way leading to the identification of the members? (you can't ask them directly - who's this? - most of the times you have to take a passive approach).

The concept used above can be given any complex terminology name but essentially it's a combination of two factors -

1. behavioural prediction recognition.
2. identifying direct pointers between members talks (for example somebody addressing another person by their name or someone in authoritative position asking for a timeline based result).

It's some time for a human mind to identify all the people - for example when I did this experiment with a group of 25 odd people - it took me 5 days to reach half level identification mark when participants were communicating for 12-16 hrs a day with me on passive end most of the times.

If done correctly this concept uses phrases between 2 sets of people to understand their relativistic positions and roles and it also takes into fact the burst of communication, the quickness, the urgency or the timeliness, if this combined with a plethora of other factors in a working environment - it can expedite this identification process exponentially but is not preferred as this is a typical case of self training.

Why develop something like the above?

Think about it - any communication established with identification parameters is subject to that communication being reliant or based on those parameters and the scope of the output governed by those parameter roles, here were are removing the roles and letting the system decide the roles based on communication, essentially reversing the identification process. This process can be used to -

A) identify a pattern.
B) maintain that the pattern is uniform.
C) detect any changes on the pattern in due course leading to an indication of a security flaw.

This is just one basic use of above kind of area where this concept can be applied, another area can be new environments or unknown environments, the possibilities are vast.

In my next post, I will try to put an algorithm and develop a program to perform the above identification.

Well - now at least with such a program - I don't have to spend time guessing who it may be I am talking to or being impolite to ask that question in a group? ;).

Let's talk about some recent events & how intelligent computing could have helped..

A lot of us might have heard the news of earthquakes which rattled the western regions in America's.

When we think about it & wonder what was the worst part of these events -
    1) the fact that they occurred.
    2) the fact that they didn't give a prior notice before occurring.
    3) the fact that they might have been prevented.

Being an earthquake prone region, an occurrence of earthquake shouldn't be surprising but what is surprising is that the prediction models have just been constrained to the occurrence and not the simulated reality around it. Difficult to understand?

Let's take an example here - you wake up or maybe you are in the midst of doing something and you get an 'alert'(you can call it 'notification' - not usual time wasters but something like this)
- titled - 'seismic activity prediction' - (when expanded reads something like this) -

'there is a 75% probability of a seismic activity measuring 6 - 7.5 on the richter scale today between 2 & 3 pm EST. You might want to consider moving the following items - X, Y & Z - to a safer place to avoid losses or consequential damage costing upto 20% of your monthly earnings for the next 12 months.'

This is what current technology can do - they can simulate learning with prediction models to evaluate the risk of an earthquake and then predict the asset losses which can occur when such an activity occurs.

How - ask the right questions to the earth & earth's core & subsequent questions for changes in climatic conditions of atmosphere of different places on the seismic or fault zones and then establish a relation between them, perform deep learning using each model to arrive at most probabilistic activity.

Ask the right questions to know a  persons key items which they would not prefer loosing. The trick here is not a direct question but a question which forms a part of an answer to the direct question. If done correctly - the system should be able to predict the key items the consumer would not want to loose.

Then the only other thing left to do is to establish a relation between activity A & activity B, A being a seismic event & B - the affect it can have on that person.

Above appears to be simple when put forward but needs a lot of deep learning & artificial intelligence  arriving at conclusion when estimating the potential of a factor hampering the user.

Let's take another example - a famous company took a decision one day to remove the charge light or indicator which indicates a lap is charged or not from their chords, they also took the chords back 2 steps by keeping them non-magnetic. In short - however they want to interpret it but they lost the battle of savings vs indegenuity (their product turning backwards from a quality focussed one to an ordinary one) which will not be easy to get back unless they do something creative and out of the world.

Would this decision have been taken by an intelligent system - it wouldn't have gone backwards, as that system would have been able to relate to the consumption pattern accurately.

This way an intelligent system could help to aid in design decisions for crucial design areas and non-crucial but significant quality areas.