A FACT360 BLOG SERIES – PART 6
Professor Mark Bishop,
FACT360, Chief Scientific Adviser
[email protected]
In his latest post in ‘The Science of AI’ series (read the first in the series here) FACT360’s Chief Scientific Adviser, Professor Mark Bishop explains how new approaches to cognitive science are replacing GOFAI – Good Old Fashioned AI…
Throughout the 1970s and 1980s classical computational approaches to Artificial Intelligence (aka GOFAI: Good Old Fashioned AI) and concomitant ‘Representational theories of mind’ – both of which, at heart, consider intelligence and cognition in terms of the explicit manipulation of symbol-structures (representations) – appeared to be ‘the only game in town’; as Allen Newell and Herb Simon famously put it in 1976:
“The principal body of evidence for the symbol system hypothesis… is negative evidence: the absence of specific competing hypotheses as to how intelligent activity might be accomplished – whether by man or machine”.[1]
In 1990, Allen Newell further emphasised the continued role of computationalism in cognitive science:
“…although a small chance exists that we will see a new paradigm emerge for mind, it seems unlikely to me. Basically, there do not seem to be any viable alternatives. This position is not surprising. In lots of sciences, we end up where there are no major alternatives around to the particular theories we have. Then, all the interesting kinds of scientific action occur inside the major view. It seems to me that we are getting rather close to that situation with respect to the computational theory of mind”.[2]
In stark contrast to Newell & Simon’s computational optimism, in the previous blog I outlined three ontological arguments that purport to show that computation is neither necessary nor sufficient for cognition; specifically, that computation cannot instantiate ‘genuine understanding’, ‘human [mathematical] insight’ or ‘phenomenal consciousness’.
In the early nineties, along with other critiques of computationalism, such critiques prompted a move away from classical AI via (a) a resurgence in interest in connectionism (Artificial Neural Networks; ANNs) and (b) the emergence of two new cognitive paradigms: ‘dynamic systems theory’ and ‘swarm intelligence’[3]. The latter approaches suggested that:
- cognitive systems may be better understood as ‘continuous dynamical systems’ rather than ‘digital computers’, with cognitive processes realised by state-space evolution in specific types of dynamic system, rather than by the abstract execution of a particular sequence of computation.
- intelligent behaviour can arise from the actions of many relatively simple agents acting collectively as a swarm. In this conception (a) ‘mind is social’ and (b) the resulting ‘swarmic’ behaviour offers a practical, decentralised, methodology for intelligent problem solving[4].
Although each of these alternative paradigms – ANNs, Dynamic Systems and Swarms – can be interpreted as a retreat from the view that cognition is best conceptualised as the manipulation of symbols, all three still exhibit an underlying commitment to computationalism, in that the common vehicle underlying their practical realisation remains the digital computer. It is the digital simulation of a neural network that learns a particular mapping; the digital simulation of a dynamic system that approximates a particular state-space evolution and the digital simulation of a swarm that exhibits a particular intelligent behaviour.
In contrast, the three ‘a priori’ arguments outlined in the previous blog suggest that in order to deeply grasp the nature of animate cognition, we are obliged to radically reject the ubiquitous computational metaphor, and instead reflect on how meaning, teleology and human creative processes are, ultimately, grounded in the human body, society and the world. This, in turn obliges us to take fundamental issues of embodiment – the body and our social embedding – more seriously.
In this way, the new embodied paradigms – emphasising the embodied mind and its interactions with environment and society – bridge the ontological gaps exposed in the last blog. Together they open a new, shared reality – mind lived-in the body; body lived-in society; society lived-in the world – exemplified by four recent research directions, the ‘Embodied, Embedded, Enactive and Ecological’ traditions – the so-called 4E approaches to cognitive science[5].
I will begin to unpack the 4Es in more detail in a subsequent blog, but I am introducing them here to show how the science is developing in very exciting ways and how it opens up new techniques for us at FACT360 to fathom the real meaning that exists within communication networks.
Read Part 7 here – “FACT360: eDiscovery in Action”
Professor Mark Bishop is FACT360’s Chief Scientific Adviser and to see how these leading-edge scientific techniques can be applied in your organisation download our White Paper “The Science of FACT360”or get in touch [email protected].
[1] cf. ‘Computer Science as Empirical Inquiry: Symbols and Search’.
[2] Are There Alternatives? in W. Sieg, ed., Acting and Reflecting, Boston: Kluwer, 1990.
[3] Indeed, in 1989, at an IEE conference in London, I published details from my PhD thesis which described the world’s first ‘swarm intelligence algorithm’, ‘Stochastic Diffusion Search’ (SDS).
[4] In a series of papers Bishop et al., showed SDS to be Turing complete; any Turing Machine program can be reduced to an appropriate SDS ‘swarm system’, and further that, because SDS can be recast on a connectionist – NESTOR – architecture, alternative insight into how ‘computation’ may be realised via ‘neural-like’ structures.
[5] I chose not to include Andy Clark & David Chalmers’ ‘Extended’ conception of mind within the aegis of the 4Es as, in extremis, I believe this reduces to a merely weakly embodied, and ultimately computational, account.