Page 413 - Deep Learning

396 Notes to Pages 14–15

34. Lukacs (2001) describes the five desperate days in May 1940 when Churchill’s
insistence that Britain stand up to Germany was translated from personal stance
into national policy. The societal level of analysis that attributes the Allied vic-
tory to, among other factors, the ability to mobilize mass production for the war
effort, is developed by Overy (1996).
35. Gaddis (2002), p. xi.
36. See Levinson (2006, pp. 272–273) for the quote about shipping containers. Some
analysts believe that economic systems like markets exhibit a turbulence of their
own; see, e.g., Foster (2005), Mandelbrot and Hudson (2004), Ormerod (1998)
and Taleb (2007). For reflections on the unpredictability of events in politics and
international relations, see, e.g., Bernstein, Lebow, Stein and Weber (2000), Blyth
(2006) and Kuran (1995). Regarding the failure to predict the fall of the Berlin
wall in 1987, specifically, see, e.g., Meyer (2009).
37. Exactly what is and what is not innate has been the subject of controversy ever
since the publication of Francis Galton’s Hereditary Genius (Galton, 1892/1962)
and Charles Spearman’s groundbreaking 1904 article in which he proposed the
concept of general intelligence (Spearman, 1904). Recent entries into the debate
include the politically inflammatory The Bell Curve by R. J. Herrnstein and
C. Murray (Herrnstein & Murray, 1994), as well as The Blank Slate by Steven
Pinker (Pinker, 2002). The evidence for an innate component to general intel-
ligence is thin but consistent. Performance on tests of cognitive ability correlate
to a statistically significant degree even when the tests are very different in char-
acter; see Carroll (1993). Genetics add another piece of evidence; see Bouchard
and McGue (1981) for a review of several hundred studies that relate genetic
distance to similarity in IQ test scores, and Posthuma and de Geus (2006) for
molecular correlates to IQ test performance. The problem with all studies of
this sort is that there is little or no reason to believe that IQ tests measure any-
thing significant. The empirical evidence for innate components is stronger with
respect to a small number of specific cognitive domains, primarily language
(Crain & Pietrovski, 2002; Laurence & Margolis, 2001), number knowledge
(Starkey, Spelke & Gelman, 1990) and, especially, music (McDermott & Hauser,
2005). But the strongest reason to believe in an innate basis for any cognitive
ability is not the research evidence but the sheer implausibility of the idea that
all brains are born exactly alike; why would they be? Whether innate differences
are of such a sort and such a magnitude that they matter for everyday life is a
separate question.
38. To highlight the contrast between animal and human learning, compare Moore’s
(2004) review and synthesis of the mechanisms identified in animal learning with
the hypotheses regarding human acquisition of skills proposed to date (Ohlsson,
2008a) and human acquisition of declarative knowledge (Chi & Ohlsson, 2005).
See also Tomasello’s (1999a, b) attempts to explain the contrasts between ani-
mals and humans in terms of the ability of infants to observe, understand and
imitate the actions, physical as well as symbolic, of other humans. With respect
to the qualitative difference between human and animal language learning, see
S. R. Anderson (2004).