Is there an 'Artificial Consciousness'?
The advent of Artificial Consciousness could well mean a great leap forward for all autonomous technologies currently being developed, whether we are talking about robotics, self-driving cars or the connected systems created for Smart Cities. Artificial Consciousness would enable machines to interact with humans at an extremely high level and carry out a much greater number of tasks than we can currently get them to perform.
This next stage in AI could be reached as a result of an unprecedented surge in technological development that will bring about an 'intelligence explosion' in machines, especially when coupled with the prediction known as 'Moore's Law', which posits an exponential increase in computer processing power continuing for the foreseeable future. Research into Artificial Consciousness is now also delving into other areas, including attempting to gain a better understanding of how human consciousness works.
In order to update readers on these questions and find out more about what benefits Artificial Consciousness could bring to humanity, L’Atelier met up with two experts in this field: Stéphane Mallard, ‘Digital Evangelist’ at business digitalisation specialist Blu Age; and Serge Tisseron, who is a psychiatrist, a member of the French Academy of Technologies, and author of a work entitled Le jour où mon robot m’aimera - Vers l’empathie artificielle (“The Day My Robot Comes to Love Me: towards Artificial Empathy”), published by Albin Michel.
L'Atelier: So how would you define Artificial Consciousness?
Doctor in Psychology
Author of "The Day My Robot
Comes to Love Me: towards Artificial Empathy"
At the moment you can't even compare machine intelligence to ant intelligence
Serge Tisseron: Well, the word ‘consciousness’ has a range of meanings. Primary consciousness in animals enables them to find their way around and protect themselves, but without having any real consciousness of their own existence. The consciousness of self, which means that a living creature is aware of its own existence as an autonomous being, is found among the higher apes. Then there is an advanced form of self-awareness in Man, which enables us to see ourselves as the subject of an ongoing story, with a past and with plans for the future. It seems quite feasible that the increasing complexity in artificial neurons might well produce a form of self-awareness. However, this type of machine would have a fundamentally different experience of the world and of itself than we do and no-one really knows what exactly that would consist of.
Stéphane Mallard: The work carried out to date has only focused on one single aspect of consciousness – ‘reflexivity’. This is all about modelling a definition of consciousness in high-level algorithms, which observe their own way of working, and for lower-level algorithms, and deducing the different states of consciousness. Algorithms are able to take an objective view of their own functioning, rather like children, who come to realise at the age of about a year and a half that they’re responsible for the words which come out of their mouths when they start to speak, and understand that the movements they see if they watch themselves in a mirror are their own.
And how comparable is that to human consciousness?
SM: It’s not really the same thing at all. First of all because the phenomenon of consciousness still raises a number of question marks and above all because, as its name implies, it would be ‘artificial’ – i.e. you’re trying to attain the same objectives as with human consciousness but modelled on an underlying computerised substratum, which necessarily places severe limits on things.
What might Artificial Consciousness bring to machines that Artificial Intelligence will never be able to offer?
ST: The ability to draw up plans for their own existence. But we’re still a long way from that! At the moment you can’t even compare machine intelligence to ant intelligence.
And how would that change our relationship with machines?
ST: Well, we’d have to create a new category of ‘beings’ somewhere between animals and human beings, with specific rights and responsibilities. But a lot of people are quite likely to believe that robots possess a kind of self-awareness long before such consciousness actually becomes a reality.
ST: For instance, a robot that can diagnose its own faults and automatically take the necessary action to reconfigure itself might appear to have a sort of self-awareness. And if the manufacturer claims that is the case, a lot of people are likely to believe it. But that’s not really Artificial Consciousness.
What sort of progress is currently being made?
SM: I can think of one notable case. In 2015, Selmer Bringsjord, a Cognitive Science professor at the Polytechnic Institute of New York University, unveiled an example of an Artificial Consciousness algorithm. He carried out a test on a Nao robot equipped with an algorithm capable of questioning its own experiences. When the robot was asked whether it had been given a ‘dumbing pill’ that made it mute, it began by saying that it didn’t know. Then it realised that it was in fact speaking and so immediately changed its answer to say that, no, it hadn’t been rendered speechless. This is still very basic but it is an algorithm that enables a robot to observe its own workings and make deductions about its own condition with a degree of detachment.
Could the development of Artificial Consciousness take a different path to that of AI?
SM: Well, it’s not about doing more but doing something different, so we’re not talking about a sort of higher-level AI. You could for instance give algorithms artificial personalities capable of placing themselves in simulated artificial emotional states depending on the circumstances. Once again this is still something artificial and so it has nothing to do with the consciousness experienced by living beings, but it does give us the illusion of dealing with a self-aware machine. However, all that is of course still a matter of algorithms.
What would be the main technological benefits of Artificial Consciousness?
ST: Robots would have a certain degree of autonomy. But human beings must remain in charge of the ‘term’ of action – the time-horizon if you prefer – distinguishing clearly between the short, medium and long term.
Thinking about creating artificial consciousness is still today in the realm of science fiction.
Are we a long way from achieving this kind of outcome?
SM: The neuroscience of consciousness is still in its infancy right now so thinking about creating artificial consciousness is still today in the realm of science fiction. But there are a few laboratories working in this field, mainly focusing on understanding Man-Machine interaction.
Will Artificial Consciousness come to rival human consciousness one day, or are there some impassable technological limits?
ST: Well, robots are set to become increasingly interconnected, and if one day there are robots in existence that have a type of consciousness close to that of human beings, it will probably be more in the collective than the individual sense. That might give the machines considerable superiority over us but in the meantime we humans will probably also become interconnected!