San Francisco – California








Economic Models




The future is proclaimed to be in the hands of computers (either machines or robots) since we are not sufficiently sophisticated or mentally acute to handle accounts and plans efficiently (or optimally).  More and more companies are outsourcing there sales and cost information to computer scientists to optimize sales and income.


Well I am not so sure.  I have found that the output of a machine is no better than the input.   In fact I found in one case, at a cost of $25,000, that the output was simply one of the inputs.  Most of the input dropped out in the manipulation of the data.  And in another case the output was wrong.


In these cases the authors could not explain why the output was bad.  What happened was in the manipulation of the data, essential elements were not integrated correctly.


Why does this happen?  Why is a machine unable to make good judgments?


I see the problem as the piling on of formulas and data that exceeds the logic of the system.    Some inputs are unknown and left out, others are inappropriate, or the weighting is wrong.


A person senses when a result is wrong because of observations that were not included in the formula as being either not thought of or considered minor or insignificant.


One other factor. 


I created a program to use for bidding that had 55 inputs and many formulas to arrive at a price.  I know when the program is not relevant, when inputs exceed or are too small to work accurately.   But those using it later do not recognize the program does not work well in these cases.


So I am confident that human judgment will still be needed to use the results of Artificial Intelligence.


What do you think?












Comments are closed.