This morning I caught the tail end of an interview on the BBC’s HARDtalk program with Sir Roger Penrose. I missed most of it, but what I did catch was absolutely fascinating. Roger Penrose is one of those physicists I admire, up there with Stephen Hawkin and many of that crowd.

The part that caught my attention was that I came on just before the discussion turned to AI and computers. Roger Penrose is well known for having turned his attention to AI in the late 1990s and has produced some phenominal work in that field. In 1990 he released what has come to be considered one of his best works, The Emperor’s New Robe in which he postulates that consciousness transcends modern science because we do not yet understand what it is, and that for this reason we will never be able to truly mimic it with a computer. Stephen Sackur of the BBC brought this subject up with Sir Penrose and I was instantly hooked.

Ads by AdGenta.com

In the interview, Sir Penrose stated that one day we will be able to create a conscious entity. However, this entity will not be a computer or any other mechine as we know it to be now. But the interesting part was that he also said that this conscious entity may not necessarily be a living being or creature. Something I’m struggling to get my head around.

It is well known that Sir Penrose is heavily critical of our current state of understanding in regards to gravity and quantum mechanics. The LISA project is an attempt to help us understand gravity further by placing 3 satellites in to space that will attempt to record and measure gravitational waves from deep space. Sir Penrose states quite emphatically that until we understand gravity more fully, quantum mechanics is missing a huge piece of the puzzle. The understanding of gravity will fill in much of that puzzle and fundamentally alter our understanding of quantum mechanics, which in turn will be the beginning of our steps towards understanding the consciousness of a living creature. Not just the human mind, but of all living creatures known to mankind.

In short, he says still still don’t know squat and what Newton discovered when an apple fell on his head is still about the same level of our overall understanding today.

What interests me is his emphatic statement that computers will never become artificially conscious. Strange as it might seem to many people, I have to fully agree with this sentiment. Not just because we really have no understanding of what consciousness is, but also because of the constraints of computers themselves.

Even if we were to master creating a 1 mega-qbit quantum computer, we would still not be able to create an artificial consciousness. Roger Penrose says it is because we still do not fully understand or even comprehend gravity completely, let alone quantum mechanics which will rely on that gravitational physics we’re yet to grasp. I say there is a lot more to it than that.

I should point out that I’m absolutely not going to get in to a discussion of mathematics and physics. They’re not my specialty and I cannot discuss them authoritively without spending years researching first. However, I have a basic understanding of both, and spend a lot of time following the fields of quantum computing from a security/encryption perspective.

Oh and just because I feel like sharing it, I just wanted to say that in no way do I subscribe to the hypothesis that because an observation alters the state of the observed particle, alternate universes must exist for each possible state. I consider that to be in the realm of religion and a matter of faith. We have no way of proving or disproving that hypothesis. Much the same way that we have no way of proving or disproving God. As the Vodafone Warriors slogan says, its just a matter of faith.

Okay, back on topic now.

In the simplest of terms, computers are a machine now consisting of millions of switches. The state of all those switches has a defined and predictable result. We use binary to represent the state of those switches in a way we can understand. 1 being on, 0 being off. We refer to each of these switches as being a bit. The software we run on our computers does nothing more than turn those switches on and off in a predefined manner to help us get answers. The resulting state of those switches then gets returned to the software which presents that information to us in a way that we can interpret and understand ourselves.

It is the software that controls and manipulates the state of those switches. Software written by people who had a predefined idea of what they wanted to achieve and thus created the instructions necessary to obtain that result.

Now, when I say software, I don’t just mean the programs you use on your computer. Software such as your operating system or your browser, or your email client are all examples. I also refer to the software that is stored within the hardware of the computer itself. When one device talks to another device, they need to know how to not only talk to each other, but how to interpret the results they receive. For example, a telegraph machine can interpret the signals sent from another telegraph machine which produces the beeping sounds. However it couldn’t interpret the same message from a telephone. This is an absolutely rudimentary form of software, almost at the point of not qualifying, but it is valid.

The difficulty for computers is that they cannot be presented with something they known absolutely nothing about and learn it. I cannot plug in a device to my computer and just expect it to know how to use that device. I also need to give it the software instructions that tell the computer what to do with the information from the device. This is where computers start to become unglued.

If I put a piece of paper in front of you with text on it in a language you’d never read or seen before, you have the ability to go off and research and learn how to interpret that text for yourself. All I have to do is give you the piece of paper and you can do the rest for yourself from there. If you struggle with that, or you cannot do it yourself, you know how to obtain help from others that might know. And from doing this you learn not only what is written on the piece of paper, but also things such as what language it is written in, where to find more information about that language. If you wanted, you could even learn the whole language yourself. All without me doing anything at all.

Not so the case with computers.

A computer cannot do anything at all that it is not told how to do. I cannot plug a device in to the computer and have it learn how to use that device on its own. I cannot give the computer a piece of disparate information and it know what that information is, or how to use it, what to do with it, or any other number of things that a person would be able to do. If I had given the computer the information on that piece of paper, nothing at all would happen. I would have to write instructions for every single step of what it needs to do with that information. I would have to tell it to search all its known resources to discover the language and interpret it correctly, or to use the resources of another computer to achieve the same. Even then, once its interpretted that information, it still means absolutely nothing at all to the computer. Its just data.

We will achieve a facsimile of artificial intelligence simply by giving a computer massive amounts of data and telling it exactly what to do with all that data. We might one day be able to hold conversations with computers, but the computers responses would purely be preprogrammed instructions of what to do with the data it already has, and how to treat new data it might receive in those conversations. It would be artificial in the truest sense because we would have told it what to do and given it instructions for everything we could think of. But if it came up against information or data we had not predicted, it would not know what to do with that data. It wouldn’t be able to learn anything about the data simply because we hadn’t told it how to.

Artificial consciousness is a different case again. That is where a computer or machine (or construct of any kind) will be able to mimic the mind of a living creature. Whether that be something we consider as simple as an insect or as complex as a primate of some kind. However, not even the most powerful computers forseeable today are able to compare to even the brain of an insect. An insect is at least able to learn on its own and has instictual behaviour.

Computers will never be able to do anything other than what we teach and tell them to do. There will never be a day of SkyNET or Seth coming awake and taking over. The Matrix will never exist. We cannot even define consciousness properly yet, let alone being to speculate on a hypothesis of how to reproduce it artificially.

Will we create a conscious entity? I completely disagree with Sir Roger Penrose on this issue. No, we never will. At least not a completely new and unique entity of our own making and design. However, might we one day be able to interface computers with our own minds to expand the capabilities of our brains? That I think is a lot more feasible. I think cybernetics is the closest we’ll ever come to artificial consciousness. Combining a living brain with a computer. But even then we’re still a long way from Artificial Intelligence or true Artificial Consciousness.