Page 425 - Deep Learning

408 Notes to Pages 45–55

50. The concept of punctuated equilibrium was first advocated in paleontology by
Eldredge and Gould (1972); see Eldredge (1989) and Gould (2002) for further
developments.
51. Imbrie and Imbrie (2002).
52. Callander (1978).
53. For the two-stage theory of the evolution of flight in insects, see Kingsolver and
Koehl (1985); for the two-stage theory of the evolution of flight in birds, see Dial
(2003) and Sumida and Brochut (2000). For the shrinking-body explanation of
the Kiwi bird’s large egg, see Gould (1986); notice that this explanation has been
challenged by Dickison (2007). For a review of the argument for the kin selec-
tion explanation of the evolution of altruism, see Foster, Wenseleers and Ratnieks
(2006). The general notion of theory articulation was explicated in Ohlsson
(1992b).
54. Cognitive scientists are beginning to tackle this scaling task by running their
simulations to mimic long-term learning; see Anderson (2002), Kennedy and
Trafont (2007) and Lebiere (1999).

Chapter 3. The Production of Novelty
1. Hadamard (1949/1954, p. 29).
2. The story about the failed death ray concept that gave rise to the idea of an air-
plane detection device is told by Buderi (1996, pp. 52–59).
3. Harman (2001) and Nersessian (1992, 2002, 2008).
4. For a history of the crucial decade in the creation of the impressionist style of
painting, see King (2007). For a more sweeping history of the movement, see, e.g.,
Broude (1994). For a conceptual analysis of how the impressionists contributed to
the gradual withdrawal of purely representational art, see Blanshard (1949).
5. Rothenberg and Hausman (1976).
6. Although logic, the systematic study of arguments, was begun in antiquity by Ari-
s totle in his Prior Analytics and continued through medieval times (Kretzmann,
Kenny & Pinborg, 1988) and beyond (Arnauld & Nicole, 1662/1996), developments
in the 19th century brought logic closer to mathematics by depicting deductive
arguments as sequences of manipulations on abstract symbols (Boole, 1854/1958).
Arguments were said to be valid or invalid solely on the basis of their form, inde-
pendent of their content. In 1910, the British scholars Bertrand Russell and Alfred
Whitehead published Principia Mathematica, a complete formalization of logic
as it had developed until that time; an abbreviated version of this work is avail-
able in Whitehead and Russell (1962). The development of logic up to that point
is described in detail by Kneale and Kneale (1962/1984). Russell and Whitehead’s
formulation is the basis for what is taught in introductory college courses in for-
mal logic, so it is formulated and reformulated in logic textbook after logic text-
book (see, e.g., Hurley, 2006; Suppes, 1957). In the course of the 20th century,
logicians proved theorems about the Russell-Whitehead logic and expanded the
system by incorporating new rules to represent arguments about what is possibly
the case, what ought to be the case, etc. (modal logic; see Blackburn, van Bentham
& Wolter, 2007; Hughes & Cresswell, 1996) and by introducing rules for what