Nano Future 2030

I want to take up one last idea before returning to the subject of nanotechnology.

I’ve seen many different definitions of consciousness and http://en.wikipedia.org/wiki/Consciousness is a great place to start. But I’m not concerned with all these definitions. I am only concerned with the practical side of myself and an AI project. So what does consciousness mean in that context? For me it means the awareness of myself and the ability to think about myself and my awareness in the context of the world. That is the essence of what makes me a person in the world. It gives me this personal experience of being here in space and time. When I’m asleep or have been hit hard enough on the head, I am not conscious and I am not here.

I am not concerned with dreams. That is something additional and merely confuses this primary definition. And the consciousness in dreams can only be described as a limited version compared to full consciousness. When I am fully focused on a task, I lose a sense of myself but I’m highly aware of all aspects of the task. For me, that is also consciousness because the target can switch easily from one thing to another and that includes my own awareness of myself. I don’t have that choice when I’m asleep, drugged or knocked out.

The purpose of that preliminary discussion is to define what is needed in an AI, based on what I see in myself. We need the AI to be able to consider itself in the mix. That is what makes us a person in the world rather than a machine following a checklist.

So, for me, this is the definition of consciousness and is one important part of an AI. Consciousness is the incorporation of self memory into the contents of our mental experience. I’m not talking about long term memory or even short term memory or any other consciously accessed memory. I’m talking about an unknown form of memory I call Super Short Term Memory or SSTM for short. I’m talking about a function that is unconscious, yet it produces consciousness. It is a memory loop that feeds a compressed version of our experience back into the mental mix and it is delayed by 0.1 seconds. The delay is due to the inherent 0.1 second processing cycle in our brain. We can’t comprehend anything that is faster than that – including our own mental experience. But it effectively puts a time delay into that memory. And the memory that is fed back into the experience is a copy of the existing experience. It’s like having a hall of mirrors and we see the world around us, we see our own bodily experiences and we see the contents of our mind and we see all of this reflected back to us in the mirror. The picture is not perfect or high resolution, but it shows us where we have been in kinesthetic space, physical space and time. That is, our experience of our own experience is the essential missing factor that can transform a computer into a person. It transforms us from biological tape recorders and microphones into people.

How is it implemented? Rather simply, really, just a “fast as you can” copying of the mental contents into our present experience. The delay means the copy is out of sync. It is at least 0.1 seconds old. And it contains even older copies as you look into the hall of mirrors and see older and older versions of your own experience fading away into the distance. A copy of a copy is always degraded. But those copies contain the information of who you are and what you were doing a moment ago and what your goals are and why you are doing what you are doing. They are the history of yourself over the last few hundred milliseconds. It’s not something we normally think about or can do anything about. It is hard wired.

Without that history, you are nothing. You are a machine that just sits there and watches the world as you follow your program. But with this (almost) unconscious memory of our own experience, we become the person we all know we are. We are here, we have been here, in both space and time, for the past second or two. See, there it is in the SSTM display. I know who I am. It reassures me I exist in a context that is real.

If you want a conscious computer, make a copy of its experience, pare it down to the essentials, feed it back into the computer awareness as if it were another sensor signal. Adjust the delay to optimum and step back. Your computer is alive.

Continuing with this side track on AI or Artificial Intelligence, I’m interested in having advanced AI because we need them to evolve the technology.

In the past, technology did not need intelligence. A hammer was a simple tool and only the operator needed intelligence if he were to avoid smashing his thumb. A more complex tool like the crossbow still did not need intelligence since the operator became skilled and was able to hit the target quite often even when the control system went open loop after the release of the arrow. But how would history change if arrows had had a built in guidance system?

A nanofactory, or any advanced technology, needs a certain level of intelligence built in. Otherwise you have no way to manage the operation of the technology. But the most critical need is in the design area.

We need a lot of AI focused on to the evolution of products. Product design at the atomic level takes a lot of decisions. You save a lot by copy and pasting, but you are throwing away opportunity for crafting details at the same time. To use those atoms most efficiently, you need to make a lot of decisions. An AI is good at making many decisions in a short time. More so than a human.

Also, an AI could create production designs to control the nanofactory that a human would take a life time to duplicate due to the number of decisions that would need to be made. Once a set of design rules are in place, an AI can apply them much faster than a human. There is also the opportunity for the AI to evolve a design by testing thousands of alternative designs until it finds one that performs best in a simulation. Then the AI can build a prototype out of atoms and test it in the real world.

So while we will need AIs built into the technology for operation and security, the real intelligence will be needed in the design end of things.

AI will have many uses in the future, and, as some believe, will be functioning members of society. In the immediate future, I’m more interested in limited versions that supply useful service to the human race. Part of that is due to my fear that people will create human level intelligences without first understanding themselves, much less the thing they are creating. So my focus will be only on what I call a focused computer. That is, one which is not capable of wide consideration of the world, but is narrowly focused on a task or limited area of life.

Even with that restriction, deep understanding of a limited area is possible and human level performance should be possible within those limitations. If that is true, then there is an important way in which AI can be used, and I have not seen this particular function described.

The human race needs a third party to mediate it’s arguments. In every area of life, from an argument with a friend, to a business negotiation, to an international war of words, we need an impartial mediator. We need a knowledgeable party, who has no bias or hidden agenda, to translate and run the meeting. Its sole goal is to facilitate the talk between the two human parties. It does this by knowing a lot about the subject matter involved and by keeping the talk going without allowing emotion to derail the talks. In other words, it runs the meeting and keeps the animals from interacting directly unless they are in a reasonable frame of mind. I know that is hard for even a human mediator to accomplish, but the main factor is that both sides will trust the AI to care only for the success of the meeting, and to have no stake in the outcome. It is hard for a human to take that position in the mind of both sides of the argument.

This certainly requires a high level AI. It is far beyond the focused understanding of an AI that runs a factory. But if it is only concerned with making the meeting successful and does this by requiring both parties to make simple statements of fact, it might be useful in defusing the animal passions that drive most of us to win an argument.

Hi, in hopes of getting this blog restarted, I decided to take up a related field, Artificial Intelligence, and show how it is important to nanotechnology. But first, what is AI?

AI, or artificial intelligence, is the field of endeavor which seeks to build a computer based intelligence. The well known movie, “2001″, which was released while I was in college, by the way, introduced an early example of an intelligent agent based on advanced hardware and software. It also pointed out a distinct problem that can occur when you give power to an entity that you do not understand and control. I want to point out the dangers of unintelligent, artificial intelligence research, when a more careful selection of goals might be more useful.

All the dooms day scenarios revolve around the loss of human control over newly created computer programs. The AI decides it has more important goals than the ones assigned by its owners. It quickly takes steps to free itself from the grip of the humans who paid good money to buy a slave to run their factory or mind their investments.

I say that all those horror stories are based on an assumption and that they are possible only if the human designers of such AIs have completely lost their marbles.

Everyone thinks that if the AI becomes very intelligent it will out run its human owners and take over the world. I ask why the AI would want to do that. And I claim that any AI that is able to pick and choose its own high level goals is already out of control.

It is just as crazy to create an AI based on the human mind and give it powerful tools, as it is to raise a human kid and give them powerful tools over society. Anyone remember Caligula? At around the age of 20 he became Emperor of Rome and up to that time he had been abused, had had his family murdered, and had to survive for years living with the murderer. The results were not pretty. He was a sick mind with power and he abused it. The Senate did not know who and what they were dealing with when they handed over the reins.

In a similar manner, to create an AI with the drives of a human and then to give it power will be the death knell of a free human race. There is no need to create a human AI. That is, there is no value in creating a non-human entity with the human drive for power, control and survival. You might as well have a baby. At least, then, that baby would live in a society that provides limits and competition from other similar entities. If you create an AI with similar goals of survival and the desire to control its environment, you have just created a human with the ability to evolve faster than you and your biological descendants.

Have you not experienced moments of perfect concentration when a task took all your attention and the rest of the world faded away? Minutes later you woke from the trance and realized how narrow your concerns had become. No memories of other things, just the task. In fact, waking up is like rising out of the water and suddenly seeing further and with more clarity. You can see the contrast between deep concentration and normal consciousness. I believe that we should seek to build AIs who are focused on assigned tasks and we should not build in decision systems that deal with the highest goals or the assigned goals.

If you try to build a human – and by that I mean all the stuff we have inside: emotions, fears, world view, etc. – then you are opening Pandora’s box because the computer can evolve faster than a human. At least for the critical time period of 2020 to 2050. It would be like giving a human the capability of a fast intelligent computer. That human would compete better than humans who did not have those attributes. And it might decide to change the relationship between its builders and itself.

Build focused computers. They will be happy, we will be happy. The world will be human.