24 The New Ecology
that could prove mathematical theorems, and that they had “solved the
venerable mind/body problem, explaining how a system composed of
matter can have the properties of mind.”
52
Strong AI has since evolved
into the notion that is more commonly referred to in this book, artificial
general intelligence or AGI. When referring to AGI, this means that machines will be able to perform the same intellectual tasks as humans: it is
more often than not the stuff of science fiction or futurism, and there is
a veritable specific genre of scientific researchers who produce what are,
frankly, fantastical applications of AI along the lines of Tegmark’s Life
3.0, whether resulting in dystopian situations such as the “AI takeover”
envisaged by Stephen Hawking or Elon Musk, or a fruitful singularity that promises huge benefits for humanity – the line adopted by Ray
Kurzweil in The Singularity is Near.
53
This book will not concern itself
much with AGI, but instead with what is more frequently referred to as
“weak” or “applied” AI. The 1950s to 1970s are sometimes referred
to as a “golden age” of AI, but the dream of a universal, general intelligence that can compete with humans seems to be receding even as the
tasks of specific automation become incredibly successful. In early 2018,
researchers at Carnegie Mellon announced that they were ceasing work
on human-like AGI to concentrate instead on refining and engineering
particular elements of automation, following calls by figures such as pioneering AI researcher, Geoff Minton, to “throw [current research] away
and start again”.
54
This is not for even a moment to deny that automation and applied
AI can achieve remarkable things. While not necessarily bringing about
either the kind of visionary utopia imagined by Tegmark or Kurzweil,
nor leading into a bleak future for humanity ruled by Terminator-style
robots, automation can be both incredibly beneficial and destructive for
people. Concerns about the role of highly automated machines in warfare
are already widespread enough for more than 2,400 scientists to have
signed a pledge opposing autonomous lethal weapons at futureoflife.org/
lethal-autonomous-weapons-pledge. Less dramatically, automation is
set to take over, or is already taking over, a wide range of cognitive tasks
that were previously considered the preserve of humans – including
many writing tasks. The linguist and philosopher Noam Chomsky assumed that humans are born with a biological predisposition towards
language based on the “poverty of the stimulus”, that is, there are not
enough empirical data in our environments to allow us to learn language
solely through experience. We cannot, for example, learn grammar
simply through hearing sentences, which led Chomsky to assume –
in the language of AI researchers – that human intelligence was not
entirely substrate-independent.
55
Functionalist approaches to AI, that
the mind and body are separable and that the former can be modelled
by software, do not as yet appear to be able to make the leap towards
true AGI.
that could prove mathematical theorems, and that they had “solved the
venerable mind/body problem, explaining how a system composed of
matter can have the properties of mind.”
52
Strong AI has since evolved
into the notion that is more commonly referred to in this book, artificial
general intelligence or AGI. When referring to AGI, this means that machines will be able to perform the same intellectual tasks as humans: it is
more often than not the stuff of science fiction or futurism, and there is
a veritable specific genre of scientific researchers who produce what are,
frankly, fantastical applications of AI along the lines of Tegmark’s Life
3.0, whether resulting in dystopian situations such as the “AI takeover”
envisaged by Stephen Hawking or Elon Musk, or a fruitful singularity that promises huge benefits for humanity – the line adopted by Ray
Kurzweil in The Singularity is Near.
53
This book will not concern itself
much with AGI, but instead with what is more frequently referred to as
“weak” or “applied” AI. The 1950s to 1970s are sometimes referred
to as a “golden age” of AI, but the dream of a universal, general intelligence that can compete with humans seems to be receding even as the
tasks of specific automation become incredibly successful. In early 2018,
researchers at Carnegie Mellon announced that they were ceasing work
on human-like AGI to concentrate instead on refining and engineering
particular elements of automation, following calls by figures such as pioneering AI researcher, Geoff Minton, to “throw [current research] away
and start again”.
54
This is not for even a moment to deny that automation and applied
AI can achieve remarkable things. While not necessarily bringing about
either the kind of visionary utopia imagined by Tegmark or Kurzweil,
nor leading into a bleak future for humanity ruled by Terminator-style
robots, automation can be both incredibly beneficial and destructive for
people. Concerns about the role of highly automated machines in warfare
are already widespread enough for more than 2,400 scientists to have
signed a pledge opposing autonomous lethal weapons at futureoflife.org/
lethal-autonomous-weapons-pledge. Less dramatically, automation is
set to take over, or is already taking over, a wide range of cognitive tasks
that were previously considered the preserve of humans – including
many writing tasks. The linguist and philosopher Noam Chomsky assumed that humans are born with a biological predisposition towards
language based on the “poverty of the stimulus”, that is, there are not
enough empirical data in our environments to allow us to learn language
solely through experience. We cannot, for example, learn grammar
simply through hearing sentences, which led Chomsky to assume –
in the language of AI researchers – that human intelligence was not
entirely substrate-independent.
55
Functionalist approaches to AI, that
the mind and body are separable and that the former can be modelled
by software, do not as yet appear to be able to make the leap towards
true AGI.