Imperial College London Centenary
 
About Imperial
About ImperialContacts/getting hereAlumniResearchCoursesAbout this site
Select your text size  for this site here: Small Text Normal Text Large Text Extra Large Text

Note: Some of the graphical elements of this site are only visible to browsers that support accepted web standards. The content of this site is, however, accessible to any browser or Internet device.

 

Conscious Robots with Impossible Minds: an introduction to the mind of a conscious machine


19 December 1996

Dillons and the Science Museum present a lecture by Professor Igor Aleksander, Imperial College of Science, Technology and Medicine

at 1pm in the Science Museum, South Kensington on Wednesday 22 January

Professor Aleksander is known for his design of neural machines which are inspired by the operation of living brains, and his books on this topic. How could a machine build up a point of view of its own? How could a machine see and know what it is looking at? How could a machine develop a freedom of will?

This lecture is intended for anyone interested in computation and consciousness, or who is just curious to know more about computers of the future. It will describe in simple terms how such machines work. Then, for the first time in public, there will be a demonstration of a PC version of the MAGNUS machine.

The implications of the effect of such work on explanations of human consciousness and the future development of such machines will be discussed with the audience. 'There is no need to fear the conscious machine of the future', says Igor Aleksander. 'It will just be friendly, easy to use and a joy to know.'

In his new book Impossible Minds: My Neurons, My Consciousness (IC Press: £17.00 hb), Igor Aleksander explodes the myth that human consciousness is too mysterious and complex for anyone to understand. A discount on the book will be offered to ticket holders at the lecture.

Tickets for the lecture are free by prior reservation only. Please telephone Dillons bookshop at the Science Museum on 0171 938 8255.

Background notes
Igor Aleksander is Professor of Neural Systems Engineering at Imperial College and the first person to design a computer that could recognise a human face. He lectures widely and is a regular contributor to television and radio science and documentary programmes. His books include Reinventing Man and Introduction to Neural Computing.

Lecture summary

A group of colleagues and I have a neural computer MAGNUS, with which we can demonstrate some of the many and complex facets that are said to make up consciousness. Following the recent publication of the book Impossible Minds: My Neurons My Consciousness (IC Press, 1996), this lecture demonstrates some of the elements of the 'consciousness' which forms the backbone of the book.

By looking into MAGNUS' brain it is possible to have privileged access to its thoughts (albeit somewhat embryonic) and this leads to some insight into how MAGNUS works. The lecture contains the simplest explanation of how 'artificially conscious' neural systems are made and why they are so very different either from conventional computers or some well-known neural networks used in industry.

The demonstration shows what MAGNUS is 'thinking' and perceiving and this leads to the conclusion that the allegedly mysterious link between conscious experience and neural activity, whether in artificial systems or the real thing, may not be as hard to understand as some commentators have suggested.

The view expressed in this lecture is that neural activity can be so vivid in the neural machinery of an organism that it cannot but BE the sensation itself.

An agenda of engineering
As an engineer, I have based my approach on the question: if consciousness were to be required of a robot, what artefact would have to be designed which would allow a consciousness to develop? The answer to this is a 'Basic Guess about Iconic representation': the firing patterns of neurons become meaningful to the system which contains those neurons (ie. make that system conscious) because they are created BY sensory experience itself.

So if I see a cat (taking into account that the muscle movements which make my eyes dart about are also sensory experience) firing patterns resembling the perception of that cat as a mix of signals from the retina and the eye muscles are created in the inner parts of a net. Similar 'iconic' representations can be formed by any of the senses. These iconic representations may be triggered even if the cat has gone, and together with other perceptions form a structure of mental imagery (state structure) which is the sum total of the 'thoughts' triggered in the system by perception or spontaneous activity.

So, as Wittgenstein first thought (but then rethought) we do have pictures in our heads (and other sensory memories). There is no need for anyone to 'observe' these as they are crafted by perception and, indeed, ARE a wonderful compendium of complex memories and plans which we can sense inwardly as part of what we call the process of thought.

Consequences
The above would seem unbelievably simplistic were it not for the fact that the consequence of the suggested existence of such iconic representations appears to answer most of the questions which arise in connection with consciousness. In particular, it nullifies the major objection which many sages have about computer models: 'they don't understand, it's only the programmer who does'.

There is no programmer, the understanding is a resonance which the system learns to develop with the world it inhabits. In 'Impossible Minds' it has been shown that iconic representations explain how memory forms such a major feature of consciousness; how memory may be involved in the perception of the present and a prediction of the future; how the ability to attend emerges from the operation of these neural nets; how a sense of free will can arise and how abstract sensations (ie. qualia) become automatically coded by the system itself. Even emotions and instincts are not sacrosanct ...

What can be demonstrated?
One of the reasons for the lecture is that I personally was struck by a sense of what consciousness might be while working with these neural systems. In the lecture I wish to share with the audience some of this experience. A special version of MAGNUS has been developed so that the nature of this experimentation and the impressions this provokes can be made available to the audience. The key feature is that it is possible to see precisely what goes on in the 'brain' of MAGNUS in a way that cannot be done with a living organism. Seeing this happen makes it not too difficult to think that something similar may be going on in our own brains.

The iconic transfer process is demonstrated and shown to explain that both long term memories and short term ones may be created. This turns out to be a process which is a strong emergent feature of neural circuits and goes beyond what cognitive psychologists had previously imagined. An aspect of this is the ability of the systems to 'attend' to some sensory information and extract it from a mixture of other things.

Concept learning, naming (ie. word acquisition), category formation and the emergence of an understanding of logic may be demonstrated in this context. How the system copes with a changing, dynamic sensation may also be shown. Experiments will be carried out on the way the system can plan on the basis of experience and how this might lead to an impression of free will.

Whither MAGNUS?
The lecture concludes by showing how the simple mechanisms which have been demonstrated may be extended to an explanation of the emergence of an awareness of self, the understanding of natural language, qualia, emotion and instinct. A discussion with a 'conscious' robot in the year 2030, a robot which is only endowed with elaborations of what has been discussed in the lecture ends the presentation.

Some call this process reductionist. If it is, I do not understand the meaning of the word. The system demonstrated in the lecture is quite puny in comparison with a human brain (by a factor of a million or so). But seeing such a puny system develop powersofa simple form of understanding of the toy world in which it lives, helps us to ponder the wonders and richness of the living brain and its conscious activity. This enhances and does not diminish the awe we may have for living organisms. I hope to discuss this issue with the audience.

[up]