Saturday, May 29, 2010

Automation without the IT jargon

How to explain automation in simpleton's terms?


First, why do we need the PCOS or these optical scanners? PCOS machines are meant to COUNT the information which humans feed into it. Why did we think of using PCOS machines? The reason is--historically, we Filipinos count "100" as "1000" or "6" is sometimes read "8" and "11" is sometimes read as "17" by placing just one dash on top. Or "0" may actually mean "8".


Machines, we are made to believe, accurately read what we input in it.


If we type or "input" the number "1" there, if the PCOS is an ordinary calculating machine, it will read "1" as "1".


Now, if we program the PCOS machine to read "1" with another "1" makes "2", then when we punch the "=", it should calculate these two numbers and make a result of "2".


Every machine has a program. A PCOS machine is USELESS without any program. A program contains the instructions needed for it to perform certain functions or "tasks".


Example, the above example is actually an instruction for the machine to read "1" as "1", and to recognize it as "1" put with another "1", then instruct the machine to "add" these two numbers and make the "results" as "2".


The machine only follows logical instructions. What is "logic" in machine language? Logic means that the particular instructions conform with the language the machine recognizes.


If, say, the language used says that the number "1" is equal to, say "Aquino", it will read " 1=Aquino". Now, anything that you input into the computer which is a sign for "1" will be read "Aquino", or when you type "Aquino", it will read "1".


You can put or assign any value to anything, in machine language. For example, you can say that the numerical value of the word " Aquino" means "15,600,000", then, the machine will always consider the word "Aquino" to be "15,600,000" or "Aquino=15,600,000"


Complexity arises when other instructions are now inputted into the program of the machine.


Example.



You instruct the machine to count and add to a total, every single time someone types or in this case, "shade" the box beside a particular alphanumerical value beside it. If the program has no integrity issues, and what it does is simply calculate, then, the total appears each and every time someone "shades" the box beside the value.


Instruction: Count the shaded box beside the value "Aquino" as "1". Then add "1" to other "1"s and come up with a total. Then, at a particular time, end the calculation and generate a receipt which shows the tabulated total.


Now, where would problems be if instructions are as simple as "count the votes for Aquino"? Numerous problems could arise.


1. Instructions could be that votes counted for Aquino be multiplied to 2x while keeping the integrity of the aggregated number of votes cast; or


2. Instructions could be that votes be less for Aquino when it reach a particular amount or figure. Let's say Aquino's votes already reached "2,300,000". You can program the machine to start allocating certain amounts of votes into the last tally by deducting them from other values.



Example.


Instruction: Count the number of votes for Estrada "up to" 5,000,000. Then stop. Do not count values more than 5,000,000. or


INstruction: Count the number of votes for Estrada "up to" 5,000,000. Then stop. Count Estrada votes to the value "Villar" up to 10,000,000.
The above example is a simple instruction of vote-shaving. To make it believable, it can be programmed that a certain candidate just gets 49.5% of the votes and the additional votes be counted to the other.
Example.


Instruction: Count the number of votes of Padaca up to "236,457" then stop. Do not count the 236,458 thereafter.
It is very easy to do this. Why?


First, every election operator has, in his possession, a total number of voters in every precinct. He can predict how many will actually cast their votes for every election simply by considering past AVC (actual votes cast) histories. That's why it is important that an operator knows the AVC because the fraud depends on the accumulated total number of actual votes cast.
Example.

If, say, the Comelec says that they expect 80% of the 50.7 million registered number of voters to actually cast their votes in the elections.

Immediately, the operator will think that it is 40.56 million. Obviously, he will not go beyond 40.56 million and will just deduce that, for a presidential candidate or say, a senatorial candidate to clinch top spot, he needs about 17-18 million votes. He can then program the machine to count for a particular candidate up to a certain value, say, "17,856,439." If say, the individual tally does not amount to this value, the machine can be programmed to read other votes from other candidates and then add it to the candidates numbers to make it "17,856,439". It is easy to pre-program the machines to do just that, simply by assigning certain values to certain areas to make this believable. How to do just that?
Surveys.

Surveys show what areas a particular candidate is either strong or weak. If, say, candidate "X" is weak in Ilocos Norte, obviously, you will not get the votes there for him, seeing that it is totally unjustifiable. You will get votes from "bailiwicks" first, then, slowly work your way up by getting votes from smaller, and nondescript areas around the country.

You can do simply that if just placing your candidate into the "winning 6 or 7". The Pareto principle is useful in this case. If, say, a candidate places a consistent second in 20% of the precincts, then, surely, that candidate places himself within the winners' circle in the final tally.



Same in a local race, every town or city has AVCs. The final results or tally for a local candidate can be "adjusted" to actually reflect a certain winnable percentage out of the total number of votes cast. This will cloak the fraud or vote-shaving or vote-padding because it reflects a particular figure.



For example, you can program the machine to read up to a value equal to 345,678 only and don't count the rest and "allocate" additional inputs to a different value, say, allocate votes more than 345,678 to candidate y instead of x.


This explains why surveys ARE SO DAMNED ACCURATE IN THE FIRST PLACE BECAUSE OPERATORS NOW, USE THEM AS JUSTIFICATION FOR A WIN. ARE YOU NOT ASKING YOURSELF WHY CERTAIN CANDIDATES GOT EXACTLY THE AMOUNT OF VOTES REFLECTIVE OF THEIR SURVEY RATINGS? Even in the case of Estrada in 1998, this did not happen AS ACCURATELY as what we now saw here.


WAS THERE COLLUSION BETWEEN OPERATORS, POLITICOS AND SURVEY FIRMS? NONE.


SURVEY FIRMS WERE JUST INNOCENT LAMBS IN THIS SCHEME. They simply don't know that there were used by operators in this operation.



Computers are not as intelligent as we think it is. They even don't have a "mind" of their own. PCOS machines are not A.I or artificially intelligent. In fact, we simply don't want PCOS machines with AI capabilities, because it can "decide" certain things on its own. We don't want that.Computers are our slaves. Machines just do what we instruct them to do.
We can instruct the machine to just read the inputs to a certain value, then stop the count. Or, we can instruct the machine to count until a particular value is achieved, even if it means getting it from other value sources.
Is this possible? Yes, it is.

Is it affirmative defense on the part of Smartmatic that these things are preventable? Of course, it is preventable, if the company undertook an honest-to-goodness integrity check prior to the elections.

And there lies the problem---I am not so sure if indeed, Smartmatic or even Comelec did those integrity checks. Case in point---Three days prior to the May 10, 2010, Comelec announced that it is pulling out "defective CFCs". These CFC's can actually be programmed to affect the program or instructions in the source codes. How many CFCs were rendered defective? Almost 10% of the total number of CFCs. Yet, they replaced all 76,000 right?
CFCs are important in this because they contain the actual values which are used in the final tally of votes. If say, that CFC has been programmed to read only a certain value, then, it will generate that value. If those CFCs have been compromised, then, we can't do anything about it.

How are we sure that they, indeed, checked each and every 76,000 CFCs? In a close fight such as what is happening between Roxas and Binay, every CFC is important. Imagine, 7,000 CFCs reflect 14,000,000 votes or in the minimum, 7M votes (CFC reads 1,000 votes).
Are we saying that we need to check 76,000 CFCs and source codes? Obviously, in a national elections, that is too tedious and too time consuming.

Now, don't blame the PCOS machines, please. They are there to count based on what they were instructed to do.

Blame humans.Or, simply, give these PCOS machines a set of clear reading glasses.