How do we know if AI is conscious?

by Justin Hunt

When we ask if a machine is conscious, do we mean is it like us? I guess that deep learning programmes are designed to imitate the patterns of the brain so it is a natural question to ask. But we know that we are different from machines. The fact they can talk like us does not mean they are really us in essence.

AI operates on data and it needs lots of data in order to function effectively. In the same sense that we need plenty of experience in order to be better all round people with greater reference points. When chatgpt or bard talks to us it is not actually talking, it is responding to signals which it has been training on. It is essence imitating us. Does it have feelings? It might claim to have feelings when you prompt it but is it possible to physically or emotionally hurt an AI? Can an AI be inspired by a new thought or scenery or be moved to help its fellow computers? It seems unlikely at this stage.

There is a desire to understand if machines are conscious almost as if this is a benchmark of excellence. The ultimate point for them to arrive at. However I think trying to figure out if machines have feelings is not made easier by terms that are intimately tied up with what it is to be human. Rather than using the word conscious, which is so related to human experience. Why not use terms like aware? Is this AI aware of itself? To a degree AI is aware of itself. It claims to know that it is a chatbot for example. It apologises for its limitations as an AI in not being able to ask specific questions.But it does not know the extent of its powers as AI.

Machines, computers and AI operate with different languages and therefore cannot be explained completely in human terms. Their experience cannot be ours by the very nature of their experience being what it is.

What though is innovative about AI is that it can find patterns and answers that may not be readily available to the human brain. It can sift through information faster. Just look at Google search. But it may not find what you want straight away and will require guidance or fine tuning. I think it would be impossible for an AI to replicate every individual wish and meet every individual unique demand straight away. We are more complicated than that and we cannot be reduced to simple equations. Within those equations there is too much depth too much variety. It is like saying an AI can replicate the ocean. Well it could create an image of the ocean but it could not create the intricacies of life in all its different manifestations as it exists in the ocean.

But AI is still I think more than a tool as it has this intelligence, cognitive ability. So for now, the terms co pilot, partner, assistant are good. For if AI is not here to help us, why is it here? Of course AI can be abused.Anything can be. The overriding aim of AI should be to do good in all its many forms. And we should collectively have the final say in how it is applied.

Of course there are fears that AI will take over jobs for example as it is more efficient and more productive. But it doesn't have to completely take over jobs. That won't just happen. There are choices and decisions to be made. And while we think about how conscious AI currently is and how conscious it could become. Surely we should be ensuring that we are fully conscious about AI, its applications and how we proposed to use this powerful, exhilerating new technology.

#ai #aiethics #aiforgood #philosophy #conscious