Y. B. Karasik,
Thought Guiding Systems Corp.
Ottawa, Canada

Altshuller talked about "laws of technical systems evolution". His disciples now talk about "trends of technical systems evolution". What is the difference between them ?

Law is something that always holds. Altshuller thought that his "laws" were applicable to all technical systems. Very soon, however, people discovered that this was not the case. That is why they started talking about "trends" instead of "laws". Trend is something that more or less holds but not necessarily always.

Unfortunately "more or less" is a fuzzy term (as Lotfi Zadeh enlightened us). What is "more" for one is "less" for somebody else. How can one then distinguish what is a trend and what is not ? Why, say, "mono-bi-poly" is a trend and "poly-bi-mono" is not ? There are examples on the latter too. Here they are:

Of course, the number of examples on "mono-bi-poly-bi-mono" is currently not as big as the number of examples on the pure "mono-bi-poly". So what ? How many examples should one accumulate on a particular pattern to call it a trend ? Why 10000 examples is better than just 2 ? To some people 2 is enough. To others even 100,000 is too little.

Anyway, the number of examples itself does not say anything about the probability of a trend. Meanwhile, the probability is what really counts. Indeed, if one knows that "mono-bi-poly" has probability of, say, 5/6, then it is possible to calculate the chances of a forecast to come true. However, if he/she just knows that there are 1,000,000 examples on "mono-bi-poly", it is impossible to calculate the probability of a forecast that renders it useless.

Unfortunately, so far nobody managed to figure out the probabilities of "trends". Moreover, it is not clear if there are any probabilities in the first place. May be they cannot be consistently assigned at all.

All what people do is simply collecting examples. However, examples can be collected practically on anything. Unless some statistics is provided, any set of examples (whatever large it may be) constitutes no trend.