<blockquote id="quote"><font size="2" face="Verdana, Arial, Helvetica" id="quote">quote:<hr height="1" noshade id="quote"><i>Originally posted by tvanflandern</i>
<br />Life is about forming hypotheses. <hr height="1" noshade id="quote"></blockquote id="quote"></font id="quote">
I can't argue with anything you just said. Except maybe to point out that I've already seen many instances of image "blanking" that is obviously <b>not </b>hiding something, so it's very difficult to imagine what rule we would use in the hypothesis. Some selective blanking could easily be lost in the noise of real blanking.
As far as the image request part of it goes, let's just say I think we already discussed how seriously JPL takes the issue of high resolution public research requests. Didn't you say something about how they'd rather spend the time letting Kindergartners run the controls?
Regarding testing controls, I agree in principle, but I also believe there are cases where an intelligent human can look at the results at the end of the test and "interpret" them, above and beyond the obvious "black box" point of view.
Let me give you a simple example. In the early years of automation, when Autofocus, AutoMeasure, Pattern Recognition and Recipes where first being developed, it was not uncommon to still run large measurement characterizations manually by a human moving stages, pressing buttons, turning knobs, etc.
When the human did it, if something went wrong with the measurement, he saw it, and didn't count that one. Remember, these things were in their infancy at the time, so we were asking two questions: How good were the real measurements, and how and what were the nature of the bad measurements?
Ok, now fast forward about 1 year, to a similar test run on our equipment by the Japanese, using the first fully automated testing, with no human intervention. Same algorithms, just all automated. When the test of say 1000 measurements were done, we went in and looked at the statistics, and concluded which were the bad measurement outliers (outside 3 sigma), and gave the Japanese the results.
They said, no, there not interested in the good measurements, they wanted <b>all </b>of them, including the bad, to be included in the "values" of what we were measuring. We admitted that <b>full automode</b>, still had some problems, but we were able to explain the exact nature of the problems, and show them how they could tell a good measurement from a bad one, and give them statistics and values that were 99+% correct if compared with SEM measurements.
We asked them what was their goal, to get the real measurements, or to wait until our tool no longer made any bad measurements, and we might as well had been talking Greek to them.
In my opinion, that is an example of where "controls" gets in the way of "common sense". I guess I still believe there's a little "art" in everything.
rd
