MIND AND MACHINE INTELLIGENCE: John Lucas finds helpful information and some degree of comfort in James Hodgson’s primer for those trying to get to grips with AI and the issues surrounding it
Mind and Machine Intelligence
James Hodgson
Greenwich Exchange, 2024
ISBN 978-1-910996-78-2
64pp £10. 99
The back-cover blurb for this short, vividly written, and amenable work deserves to be quoted in full.
Artificial Intelligence (AI) –the latest form of machine intelligence – is profoundly affecting
how we feel and think about mind, and thus ourselves. It looks like a dystopian threat – but
this book takes it as an opportunity to refresh our understanding of mind, intelligence, and
what machines can, and cannot, do.
Reading this, any techno-phobe will, I imagine, begin to feel a measure of reassurance. And as the book itself (one that pans out at not many more than fifty pages) is both surprisingly wide-ranging and enlightening, and given that it is written with the kind of unsparing clarity that forbids the most demanding of readers to complain of lack of plain dealing, it is the ideal primer to give to anyone wanting to understand what the present fuss concerning AI is all about.
First come some definitions. Machine intelligence is, Hodgson remarks, ‘the performance of tasks by a machine that are human tasks considered to be intelligent,’ e.g complex calculations … playing chess … learning … remembering.’ As for mind, the author, quoting from the O.E.D., defines it as ‘The seat of consciousness, thoughts, volitions, and feelings.’ He allows however that ‘mind’ is something of an umbrella term, and in pages that immediately follow he notes that in Western thought, from Plato to the present, there are ambiguities and uncertainties surrounding the definition of mind.
Hodgson deals briskly with these matters before noting that it might seem that ‘mind/consciousness is analagous to what appears on the screen of the computer or in the activities of the robot.’ After which he poses what he calls ‘a Turing-type question,’ which is whether a computing machine ‘could be suitably programmed’ so as to be ‘as intelligent as what we see as human intelligence.’ And from this comes a further question, one that, if I may judge from my own experience, often feels to be at the heart of our unease with AI. Might it be that ‘machine intelligence could converge on … human beings and, in time, overtake human intelligence?’
Dealing with this question prompts what for me is the most intriguing and perhaps reassuring section of the book under review, the consideration of ‘Tasks and Machine Intelligence.’ Here, I need to quote at some length. Hodgson wants to assert that the mind outsmarts, as it were, the machine. And so, he says,
taking a simple ‘intelligent’ machine – the jacquard loom controlled by punched cards
weaves, say, a cloth or a carpet, and then the thread snaps. The machine supervisor (a
human being) steps in to remedy the situation. This intervention could be designed as
an algorithm developed to do what the machine supervisor does. Out of feedback comes
more intelligent design, in these cases by a human agency. The interesting thing about the
jacquard loom example is that it produces cloth … or a carpet, let us say with patterns as
the finished product. The loom operates in such a way as to look intelligent: compared to
a human weaver it is faster and makes fewer mistakes; thereby it appears to be more
‘intelligent’ than the human weaver. [But] the reason the jacquard was constructed was to
make cloth or carpet, which is in turn appraised and valued – all by human agency.
Not that Hodgson is content to leave matters there. Far from it. Hence his remark a few pages on, that ‘only a few human abilities need to be migrated to machines for there to be an impact upon people. The fear of artificial intelligence arises from the judgement of what people with the ability to design and/or deploy these machines could well do – and that is not an irrational fear.’
Saying this then leads Hodgson to note that machines are designed to perform defined tasks, and these can include behaving with a speed and accuracy – ’infallibility’ – far greater than humans can themselves manage. And this in its turn leads on to a consideration of ‘Evil’. Yes, Hodgson allows, there can be such a thing as a Just War, but as he adds, ‘War is a system, with objectives, targets, strategies and tactics, which are just those characteristics that machine intelligence has.’ And, crucially, the machines are created by human intelligence. Besides, whether that intelligence is used for benevolent purposes or not, the fact remains that ‘short-term self-interest, greed, the desire for dominance … are expressed through the design of machine intelligence.’ The bombs which I designed and which my machines constructed can kill more people than the bombs you made. And this being so, ‘it is the human authorship of systems that we are talking about.’
From this implacable consideration Hodgson inevitably turns again to the question of what constitutes human intelligence. Here, at the outset, he notes that such intelligence ‘is the ability to make assessment, especially of outcomes of processes and activities.’ And he follows this with a very funny excursus concerning his childhood memories of an elderly neighbour, a woman who kept a large green parrot which learnt to imitate her ‘Shut up, you idiot,’ which were the words she shouted whenever the parrot’s meaningless cries annoyed her. As Hodgson comments, there is no point in blaming the parrot, any more than we should blame any inanimate thing for behaving as though it, too, has agency, or propose putting a car operated by AI on trial for having caused a crash. And if this so, it follows that human agency is and has always to be the ultimate cause of machine behaviour.
It also follows that, in Hodgson’s words, ‘Intelligence in its fullest sense is defined by the activities of reflection,’ and that such reflection can (perhaps best) operate in dialogue with other human beings though not with machines, however ingeniously created or complex the machines may be. For when all is said and done, machines are incapable of ‘creativity, innovation, problem solving, and, above all, the notion of agency.’ In short, the ‘carrying out of tasks, defining what they are, improving on them,’ necessary for the functions of life though these all may be, ‘are not the definition of intelligence, nor of thinking, nor of the mind.’
For which reassurance, we might say, many thanks.
Feb 18 2024
Mind and Machine Intelligence
MIND AND MACHINE INTELLIGENCE: John Lucas finds helpful information and some degree of comfort in James Hodgson’s primer for those trying to get to grips with AI and the issues surrounding it
The back-cover blurb for this short, vividly written, and amenable work deserves to be quoted in full.
Artificial Intelligence (AI) –the latest form of machine intelligence – is profoundly affecting how we feel and think about mind, and thus ourselves. It looks like a dystopian threat – but this book takes it as an opportunity to refresh our understanding of mind, intelligence, and what machines can, and cannot, do.Reading this, any techno-phobe will, I imagine, begin to feel a measure of reassurance. And as the book itself (one that pans out at not many more than fifty pages) is both surprisingly wide-ranging and enlightening, and given that it is written with the kind of unsparing clarity that forbids the most demanding of readers to complain of lack of plain dealing, it is the ideal primer to give to anyone wanting to understand what the present fuss concerning AI is all about.
First come some definitions. Machine intelligence is, Hodgson remarks, ‘the performance of tasks by a machine that are human tasks considered to be intelligent,’ e.g complex calculations … playing chess … learning … remembering.’ As for mind, the author, quoting from the O.E.D., defines it as ‘The seat of consciousness, thoughts, volitions, and feelings.’ He allows however that ‘mind’ is something of an umbrella term, and in pages that immediately follow he notes that in Western thought, from Plato to the present, there are ambiguities and uncertainties surrounding the definition of mind.
Hodgson deals briskly with these matters before noting that it might seem that ‘mind/consciousness is analagous to what appears on the screen of the computer or in the activities of the robot.’ After which he poses what he calls ‘a Turing-type question,’ which is whether a computing machine ‘could be suitably programmed’ so as to be ‘as intelligent as what we see as human intelligence.’ And from this comes a further question, one that, if I may judge from my own experience, often feels to be at the heart of our unease with AI. Might it be that ‘machine intelligence could converge on … human beings and, in time, overtake human intelligence?’
Dealing with this question prompts what for me is the most intriguing and perhaps reassuring section of the book under review, the consideration of ‘Tasks and Machine Intelligence.’ Here, I need to quote at some length. Hodgson wants to assert that the mind outsmarts, as it were, the machine. And so, he says,
taking a simple ‘intelligent’ machine – the jacquard loom controlled by punched cards weaves, say, a cloth or a carpet, and then the thread snaps. The machine supervisor (a human being) steps in to remedy the situation. This intervention could be designed as an algorithm developed to do what the machine supervisor does. Out of feedback comes more intelligent design, in these cases by a human agency. The interesting thing about the jacquard loom example is that it produces cloth … or a carpet, let us say with patterns as the finished product. The loom operates in such a way as to look intelligent: compared to a human weaver it is faster and makes fewer mistakes; thereby it appears to be more ‘intelligent’ than the human weaver. [But] the reason the jacquard was constructed was to make cloth or carpet, which is in turn appraised and valued – all by human agency.Not that Hodgson is content to leave matters there. Far from it. Hence his remark a few pages on, that ‘only a few human abilities need to be migrated to machines for there to be an impact upon people. The fear of artificial intelligence arises from the judgement of what people with the ability to design and/or deploy these machines could well do – and that is not an irrational fear.’
Saying this then leads Hodgson to note that machines are designed to perform defined tasks, and these can include behaving with a speed and accuracy – ’infallibility’ – far greater than humans can themselves manage. And this in its turn leads on to a consideration of ‘Evil’. Yes, Hodgson allows, there can be such a thing as a Just War, but as he adds, ‘War is a system, with objectives, targets, strategies and tactics, which are just those characteristics that machine intelligence has.’ And, crucially, the machines are created by human intelligence. Besides, whether that intelligence is used for benevolent purposes or not, the fact remains that ‘short-term self-interest, greed, the desire for dominance … are expressed through the design of machine intelligence.’ The bombs which I designed and which my machines constructed can kill more people than the bombs you made. And this being so, ‘it is the human authorship of systems that we are talking about.’
From this implacable consideration Hodgson inevitably turns again to the question of what constitutes human intelligence. Here, at the outset, he notes that such intelligence ‘is the ability to make assessment, especially of outcomes of processes and activities.’ And he follows this with a very funny excursus concerning his childhood memories of an elderly neighbour, a woman who kept a large green parrot which learnt to imitate her ‘Shut up, you idiot,’ which were the words she shouted whenever the parrot’s meaningless cries annoyed her. As Hodgson comments, there is no point in blaming the parrot, any more than we should blame any inanimate thing for behaving as though it, too, has agency, or propose putting a car operated by AI on trial for having caused a crash. And if this so, it follows that human agency is and has always to be the ultimate cause of machine behaviour.
It also follows that, in Hodgson’s words, ‘Intelligence in its fullest sense is defined by the activities of reflection,’ and that such reflection can (perhaps best) operate in dialogue with other human beings though not with machines, however ingeniously created or complex the machines may be. For when all is said and done, machines are incapable of ‘creativity, innovation, problem solving, and, above all, the notion of agency.’ In short, the ‘carrying out of tasks, defining what they are, improving on them,’ necessary for the functions of life though these all may be, ‘are not the definition of intelligence, nor of thinking, nor of the mind.’
For which reassurance, we might say, many thanks.