Comparison of inbuilt AI with networked AI.
There are two religons in the world of developing a capability of artificial intelligence in a component, 1st which talks about developing the capability with the resource parameters of the system where the capability would be used and the 2nd one which talks about developing the capability of using resources which are primarily available on the web, the intent of both schools of thought being the same, end up with a decisioning system which does the same with high probability of it being accurate.
Let's talk about the 2nd school of thought first, build with networked component on a web, this gives a lot of advantages - better compute power, experience over the web being transmitted leading to more accurate decisioning. This model definitely serves the purpose of quickly developing a system which will provide a machine learning approach which is much faster.
Now let's consider the 1st school of thought, having an AI chip inbuilt within the environment where it's invoked, this means the chip should have a greater processing capability and needs more resources other than speed to make the same decisions quickly as done by the 1st environment.
If you look at both approaches, the 1st one would seem to be the obvious choice but then let's think about it a little bit ...
Consider a scenario where there is no network - the 1st one would fail as it always needs a connected network, the 2nd one would still work and decision, also another instance where there is a corrupted program in network and it is trying to infect all the AI devices which come in contact - the 1st first scenario would have a bigger risk of contamination coming on this program. So now - if we think more - we see why the 2nd one now seems to be more advantageous.
Now let's look at the decisioning capability environments - for example secluded environments where decisioning doesn't have access to internet, like extraterrestrial environments for e.g. moon or mars or places on earth within earth where internet is not accessible. These again serve as likely candidates for places where the 2nd school of thought seems to prevail.
Although - it's upto the consumer - which utilisation pattern a consumer is more comfortable with but over the period of time the 2nd school of thought seems to be a more reasonable approach if developed correctly.
Let's talk about the 2nd school of thought first, build with networked component on a web, this gives a lot of advantages - better compute power, experience over the web being transmitted leading to more accurate decisioning. This model definitely serves the purpose of quickly developing a system which will provide a machine learning approach which is much faster.
Now let's consider the 1st school of thought, having an AI chip inbuilt within the environment where it's invoked, this means the chip should have a greater processing capability and needs more resources other than speed to make the same decisions quickly as done by the 1st environment.
If you look at both approaches, the 1st one would seem to be the obvious choice but then let's think about it a little bit ...
Consider a scenario where there is no network - the 1st one would fail as it always needs a connected network, the 2nd one would still work and decision, also another instance where there is a corrupted program in network and it is trying to infect all the AI devices which come in contact - the 1st first scenario would have a bigger risk of contamination coming on this program. So now - if we think more - we see why the 2nd one now seems to be more advantageous.
Now let's look at the decisioning capability environments - for example secluded environments where decisioning doesn't have access to internet, like extraterrestrial environments for e.g. moon or mars or places on earth within earth where internet is not accessible. These again serve as likely candidates for places where the 2nd school of thought seems to prevail.
Although - it's upto the consumer - which utilisation pattern a consumer is more comfortable with but over the period of time the 2nd school of thought seems to be a more reasonable approach if developed correctly.
Comments
Post a Comment