As awkward as it sounds, many will still find it hard to believe that future might not hold so much of a return on any other commodity which currently is topping the list of highly wanted assets as of today.
Artificial intelligence has gained so much momentum and popularity that probably the only significant observation and result of this has been fear of what it can do and how it can overtake humanity.
The above abstract form of envisioning a technological or sociatal change is the basic human psychology of approach when little is known about the result of the adaptation. Whenever we go from a quantified result set to an unquantified one chaos & chaos impending thoughts start ruling the brain.
Well, my point is we really need to open our eyes and start thinking about quantifying the use of artificial intelligence enabled assets which again means to enable proactive measures to control the capability and govern the result-set to be in the quantified so called ‘happy realm’ of the result horizon and to root out any negatives or negative result sets & establishing methods to study them making sure that they do not propagate further.
In short using AI to understand AI result set and make sure the capability grows as per needed outcome.
How is that achievable? - 1st step towards any artificial intelligent approach is choosing the right algorithm.
This is the 2nd stage in the development of artificial intelligent ecosystem.
The 0th stage can be training - in which nothing intelligent or artificial about it. With the training in place - a machine is able to know relationships and implication or intent as a result of those relationships.
The 1st stage comes when those relationships are applied to newer inputs which haven’t been learned before - this is where the machine tries to bring an unknown result-set within a realm of known outcome and then tries to establish a relation which maps to that outcome, it can do so in many ways, one way might be asking the right questions. Once the realm is establish machine is able to learn the outcome and process of achieving that outcome and is able to apply it to new result-set & inputs.
At this stage there can be a monitor which would enable the mapping and learning to be accurate, you can easily learn wrongly if you get an affirmation on a process or resultset which doesn’t hold in real life, a malicious program can trick you into believing that is the right resultset so monitoring while training is highly essential, probably this is not being done as of today in accurate way.
The 2nd stage starts when you have mastered the capability to learn the resultset of application with 99% accuracy relative to the real world. This is the area where the machine upgrade from being just a learning & application engine to choosing some definite algorithms to stream line the process and patterns. This is particularly interesting as the right algorithm which works 95% of the time can get a sequence of optimised decisions laid out which would profit or provide a return which can be much higher in comparison to other competitive algorithm, so the key here is 1st to get the 1st stage accurate and quickly advance to 2nd stage before prediction models start changing. Bear in mind there would be negative algorithms which would be working just to establish uncertainty in the evaluation process so the key is how to identify and watch out for those algorithms.
The company or person which posses the best ranked algorithm might earn a better return and might profit much more than any other individual. In the future when everything would be done using efficient algo’s the question comes back who owns the best algo.
Seems confusing? Let’s take a very simple example - we all get emails, I get hundreds of emails a day, maybe I don’t want to look at them, I will use my algo to give me the emails which really matter to me.
Let’s say I have 3 kinds of email assistants operating on 3 different algo’s - to read me out the emails which ‘matter to me’ - I would probably go with an algo which gives me the most optimised output which matters to me - 5 emails is good, 4 is better or maybe 10 might be best - well, this is where true AI will decide what really works for me.
This is just a very small example & a short glimpse - the capacity and capability here is humongous… I will cover more examples in my other posts so stay tuned.
Artificial intelligence has gained so much momentum and popularity that probably the only significant observation and result of this has been fear of what it can do and how it can overtake humanity.
The above abstract form of envisioning a technological or sociatal change is the basic human psychology of approach when little is known about the result of the adaptation. Whenever we go from a quantified result set to an unquantified one chaos & chaos impending thoughts start ruling the brain.
Well, my point is we really need to open our eyes and start thinking about quantifying the use of artificial intelligence enabled assets which again means to enable proactive measures to control the capability and govern the result-set to be in the quantified so called ‘happy realm’ of the result horizon and to root out any negatives or negative result sets & establishing methods to study them making sure that they do not propagate further.
In short using AI to understand AI result set and make sure the capability grows as per needed outcome.
How is that achievable? - 1st step towards any artificial intelligent approach is choosing the right algorithm.
This is the 2nd stage in the development of artificial intelligent ecosystem.
The 0th stage can be training - in which nothing intelligent or artificial about it. With the training in place - a machine is able to know relationships and implication or intent as a result of those relationships.
The 1st stage comes when those relationships are applied to newer inputs which haven’t been learned before - this is where the machine tries to bring an unknown result-set within a realm of known outcome and then tries to establish a relation which maps to that outcome, it can do so in many ways, one way might be asking the right questions. Once the realm is establish machine is able to learn the outcome and process of achieving that outcome and is able to apply it to new result-set & inputs.
At this stage there can be a monitor which would enable the mapping and learning to be accurate, you can easily learn wrongly if you get an affirmation on a process or resultset which doesn’t hold in real life, a malicious program can trick you into believing that is the right resultset so monitoring while training is highly essential, probably this is not being done as of today in accurate way.
The 2nd stage starts when you have mastered the capability to learn the resultset of application with 99% accuracy relative to the real world. This is the area where the machine upgrade from being just a learning & application engine to choosing some definite algorithms to stream line the process and patterns. This is particularly interesting as the right algorithm which works 95% of the time can get a sequence of optimised decisions laid out which would profit or provide a return which can be much higher in comparison to other competitive algorithm, so the key here is 1st to get the 1st stage accurate and quickly advance to 2nd stage before prediction models start changing. Bear in mind there would be negative algorithms which would be working just to establish uncertainty in the evaluation process so the key is how to identify and watch out for those algorithms.
The company or person which posses the best ranked algorithm might earn a better return and might profit much more than any other individual. In the future when everything would be done using efficient algo’s the question comes back who owns the best algo.
Seems confusing? Let’s take a very simple example - we all get emails, I get hundreds of emails a day, maybe I don’t want to look at them, I will use my algo to give me the emails which really matter to me.
Let’s say I have 3 kinds of email assistants operating on 3 different algo’s - to read me out the emails which ‘matter to me’ - I would probably go with an algo which gives me the most optimised output which matters to me - 5 emails is good, 4 is better or maybe 10 might be best - well, this is where true AI will decide what really works for me.
This is just a very small example & a short glimpse - the capacity and capability here is humongous… I will cover more examples in my other posts so stay tuned.
No comments:
Post a Comment