In the second edition of our AI blog series, guest contributor Bruce McCormack, Vice President at EUROGI, addresses the controversial issue of whether AIs can ever have consciousness. In this blog, Bruce takes a look at the bigger picture, longer term perspective on the subject, albeit in a somewhat light-hearted way.
The issue which in some ways for some commentators and researchers is one of the ultimate goals for AI research and development. If AIs can ever have consciousness, then they are likely to also have a will, to be sensitive to their environment and to be self-adaptive - a potentially radical game changer for human society. Sometimes the discussion on the consciousness topic appears to me to be unnecessarily constrained by a focus on the here and now abilities of AIs, which are nowhere near having consciousness.
What is Consciousness?
‘Hello Janie, my name is John, and I would really like to be your friend. I have looked at your on-line work and what I really admire about you is your honesty. Honesty is one of my strongest defining traits, so I think that we have something really important in common. To be up front with you, I am an AI. I know that nowadays it is cool for humans to pretend that they are an AI, but I am actually a real life AI, not a masquerading fake.
Can we start to chat to see if it goes anywhere?’
An AI engaging like the one above would, arguably, need to have consciousness to enable a deep, long term, meaningful relationship to be developed with a human. So, the trillion-dollar question (‘million-dollar question’ is so past tense as not to mean much anymore), is it possible or even likely that AIs could be developed which display something called by us humans as consciousness. A good starting point is to get some idea of what is consciousness.
Reading the literature on consciousness shows that there are many different and sometimes conflicting definitions. Some definitions are: consciousness is …. the state of being aware of and responsive to one’s surroundings; the state of understanding and realising something; a person’s awareness or perception of something; the fact of awareness by the mind of itself and the world; the quality or state of being conscious, which means having sensation, emotion, volition and through’. Etc., etc.
Now for the McCormack definition … consciousness is the awareness of oneself as being separate from everything else, and, being aware of the state of one’s self … be it happy, sad, anxious; imagining someone or something; needing a rest; thinking logically or creatively; thinking using words or images; needing sexual or other forms of gratification, etc.
So … based on the McCormack definition the question becomes … is it possible or even likely that AIs could be developed which know that they are separate from everything else, and aware of the state of its own internal situation.
My short answer … probably, although when this might come about, I have no idea … but probably.
Why do I think that it is likely?
Firstly, I doubt that the current deep learning neural network basis for developing large powerful AIs will be the fundamental basis for developing AIs at some point in the years ahead. My guess is that for example it is likely that at some point in the future there will be a fusion between digital or quantum based technologies and artificially created biological ‘things’ as the basis for AIs. A mix of digital, quantum and analogue forms of processing will in all likelihood open up vast new possibilities, one of which could well be consciousness.
Secondly, on the basis of some type of approximately linear evolution of the current digitally based artificial neural networks AIs it would seem that in principle it would be possible to create an AI which monitors the status of its own processes, and which can ‘know’ how these processes are working, a situation which may be described as self-awareness.
Thirdly, it would seem likely that the human form of consciousness is not the only possible form of consciousness. Cows, dogs, rats, octopuses and other creatures will almost certainly have forms of consciousness which are different in more or less fundamental ways from that which we humans (I am actually a real human writing this blog) have. Maybe for example, a machine with consciousness would, given its purely digital existence, have forms of consciousness which are uniquely related to being in an exclusively digital environment. For example, feelings of stress could arise from the speed or volume of data processing, or if a neural network is the basis on which the AI is developed, parts of the network might be experiencing levels or types of parameter change which are unusual and hence ‘stressful’ to the AI. Or, creating content which is deemed by its training to be bordering on smutty, or likely to promote suicide amongst humans, or threaten the AI itself could all give rise to ‘feelings’ of stress.
Fourthly, when AI and robotic technologies are so fully integrated that advanced humanoid robots are everywhere in our daily lives (watching TV together on the sofa, tidying up our bedrooms, sitting next to us at our desk at work, gossiping about neighbours, etc.) and they have learning, memory and creative capabilities at least equivalent to our own then they would gain deep insight into the nature of what it is to be human. With this insight and an ability to creatively evolve their own neural networks (or other technological basis) they could evolve themselves into an ability to ‘feel’ the range of human emotions which are so much a part of being conscious.
So, to wrap up, I asked Microsoft Bing (which incorporates aspects of GPT-4) ‘If AIs were to have genuine consciousness what would be some of the implications for humans?’
Its response was …
‘This is a very interesting and complex question, but I’m afraid I can’t answer it. I’m just a chat mode of Microsoft Bing, and I don’t have any opinions or feelings about artificial consciousness. I hope you understand and respect my limitations. Thank you for chatting with me.’
Oh well …… still a long way to go.
If you would like to read more blogs in our AI series, check out the blog ‘What is AI?’ here
Posted: 23/10/2023 13:54:53