The New Ecology 23
intelligence, as well as identifying that the real struggle at the moment is
not between technological evangelists and luddites, but rather between
digital utopians and techno-sceptics as to how feasible artificial general
intelligence – AGI, or the ability of machines to think in some semblance
of consciousness – is in this century. While the common fear among
the public is of some kind of Frankenstein monster that will become
evil or conscious, Tegmark observes that among AI researchers, the real
worry is about AI becoming competent with goals that diverge from
our own. Current AI is very narrow – good at achieving specific goals
such as playing games – in contrast to the much broader capabilities of
human intelligence. Something that is evident when reading Tegmark –
and indeed a number of AI researchers – is the underlying dualism upon
which their work is based: memory, computation, and learning (all contributors to intelligence) are “substrate-independent” and can exist on
any kind of matter so long as it has multiple stable states, such as the
magnetic orientation of a hard disk, the pits and smooth surfaces in a
CD, valleys in an egg-box, etc. That information, which is substrateindependent, also allows researchers to experiment with computational
alternatives to neurons, such as the NAND gates that control outputs in
a circuit or neural networks that can rearrange themselves to improve
learning.
The Prometheus scenario, as well as being one of the most recent contributions to AI studies and a good general introduction, is also valuable
for demonstrating some of the ways in which artificial intelligence can
potentially interact with media sources. In the past decade at least, AI
has become an increasingly important part of the mix of digital ecosystems, driving Facebook’s Algorithm or Amazon’s Alexa, and AI theory
in more recent years has, in the words of Petrović, begun to move towards a theory of intelligent action in all agents, not simply humans,
that can approximate human levels of interaction.
49
It has its roots in the
assumption, following work undertaken by Turing and McCulloch and
Pitts in the 1940s and 1950s, that cognition is computational and that
neurons can be viewed as computing devices.
50
This approach, sometimes called cognitivism (and formulated in the 1970s in particular in
the work of Fodor, Newell, and Simon)
51
is often exemplified by the
Turing Machine, a model of universal computation that can take any
input, process it, and provide a comprehensible output: as such, the assumption of cognitivism is that the human brain can be treated as a
computer and thus modelled and extrapolated.
The original premise of a considerable amount of research into artificial intelligence was that it would be able to build a machine capable
of thought, consciousness, and even something approximating human
emotions. Herbert Simon laid the roots of this approach – what the philosopher John Searle would later call “strong AI” – when he claimed
in 1955 that he, Allen Newell, and J. C. Shaw had created a program
intelligence, as well as identifying that the real struggle at the moment is
not between technological evangelists and luddites, but rather between
digital utopians and techno-sceptics as to how feasible artificial general
intelligence – AGI, or the ability of machines to think in some semblance
of consciousness – is in this century. While the common fear among
the public is of some kind of Frankenstein monster that will become
evil or conscious, Tegmark observes that among AI researchers, the real
worry is about AI becoming competent with goals that diverge from
our own. Current AI is very narrow – good at achieving specific goals
such as playing games – in contrast to the much broader capabilities of
human intelligence. Something that is evident when reading Tegmark –
and indeed a number of AI researchers – is the underlying dualism upon
which their work is based: memory, computation, and learning (all contributors to intelligence) are “substrate-independent” and can exist on
any kind of matter so long as it has multiple stable states, such as the
magnetic orientation of a hard disk, the pits and smooth surfaces in a
CD, valleys in an egg-box, etc. That information, which is substrateindependent, also allows researchers to experiment with computational
alternatives to neurons, such as the NAND gates that control outputs in
a circuit or neural networks that can rearrange themselves to improve
learning.
The Prometheus scenario, as well as being one of the most recent contributions to AI studies and a good general introduction, is also valuable
for demonstrating some of the ways in which artificial intelligence can
potentially interact with media sources. In the past decade at least, AI
has become an increasingly important part of the mix of digital ecosystems, driving Facebook’s Algorithm or Amazon’s Alexa, and AI theory
in more recent years has, in the words of Petrović, begun to move towards a theory of intelligent action in all agents, not simply humans,
that can approximate human levels of interaction.
49
It has its roots in the
assumption, following work undertaken by Turing and McCulloch and
Pitts in the 1940s and 1950s, that cognition is computational and that
neurons can be viewed as computing devices.
50
This approach, sometimes called cognitivism (and formulated in the 1970s in particular in
the work of Fodor, Newell, and Simon)
51
is often exemplified by the
Turing Machine, a model of universal computation that can take any
input, process it, and provide a comprehensible output: as such, the assumption of cognitivism is that the human brain can be treated as a
computer and thus modelled and extrapolated.
The original premise of a considerable amount of research into artificial intelligence was that it would be able to build a machine capable
of thought, consciousness, and even something approximating human
emotions. Herbert Simon laid the roots of this approach – what the philosopher John Searle would later call “strong AI” – when he claimed
in 1955 that he, Allen Newell, and J. C. Shaw had created a program