3.31.2010

Mlodinow

Excellent article, thank you for sending it along. Reminds me of Tolstoy's point about the fog of war. I would only come back on one sentence in the whole article, in red:

In addition to the intelligence reports that in hindsight seem to point toward a specific attack, there is also a huge background of useless intelligence, each week bringing new reams of sometimes alarming or mysterious messages and transcripts that would later prove misleading or insignificant. In advance of the event, you can’t tell one sort from the other.

If I were standing on line at the movies chatting with Leonard Mlodinow, in response to his assertion that 'in advance of the event' one can't discern useful from useless information ...

Mlodinow: Now hold on: I didn't say 'information'; I said 'intelligence.'

Montenow: Professor Mlodinow, my apologies. You asserted, after giving the specific example of Pearl Harbor, that 'in advance of the event, you can't tell [useless intelligence] from [intelligence reports that in hindsight seem to point toward a specific attack]." Have I got it right now?

Mlodinow: Yes, that is what I said.

Montenow: Are you saying that this is a hard-and-fast rule ... that no amount of processing or analysis of information available in the present will ever yield up predictions that are significantly better than chance? Such that one might never utilize such predictions to evolve strategies that are effective in controlling for risks and hassles and challenges that life seemingly randomly tosses up?

Mlodinow: No, no, nothing so strong as that was intended. As I said, "... I try to relax, and work on making the best of whatever develops, rather than worrying about how awful it is." So, the way I "work on making the best of whatever develops" -- that is, the strategies I calmly employ as I make my way through life -- are ones that rely on reading the present and forming judgments about how to act, always seeking to 'steer' circumstances toward 'solutions'; that is, toward futures in which risks and hassles are controlled, and happiness (among other non-zero-sum benefits) is promoted.

Montenow: Nice to hear. Because at the moment it appears that "evidence from computational models hints at the potential existence of a large equivalence class of adaptive behavior."

Mlodinow: Miller & Page, page 83. Yes, that caught my attention too. What does that mean to you, "large equivalence class of adaptive behavior"?

Montenow: Well, for starters, an 'equivalence class' of 'adaptive behavior' simply means, for example, that common strategies arise out of divergent implementations of, for example, genetic algorithms. "Implementations of such algorithms often differ in a variety of ways; for example they may use very different kinds of representations, selection mechanisms, and choices of specific operators. Notwithstanding these choices, the resulting algorithms tend to perform more or less identically." M&P, App. A, at 240.

Mlodinow: Go on, I'm with you ...

Montenow: So this tells me that
there are reasons to believe
that it is worthwhile to investigate
the extent to which
one can beat chance in predicting
(based on analysis and processing
of information available in the present)
that the state-space of the world
will fall into some such equivalence class.

Because if so, one might use this ability to evolve strategies that will be effective in controlling for the risks and hassles and challenges that life tosses up.

Mlodinow: Hmmm. I don't know what to say.

Montenow: I'm saying, it seems to be possible to gray some otherwise black swans. And it certainly seems like a good idea to investigate further whether or not it is possible. But apart from that one small quibble, I am very thankful for the points of view you have so eloquently articulated in the NY Times interview.


¡Salud!