Computers begin to think

[Nanook is talking to George]

“I think machines will follow an intellectual evolution path that is similar to what biological life did. That is, as machines become complex, they will develop the 3 brain structures. You know quite a bit about computers. Do you see any parallels in computers to the 3 brains?”

“Hmmm . . . Actually, I see a lot! Humans need air, shelter, water, sleep, food. Hulk drives the body to get these. Similarly, computers need cooling, memory refresh cycles and electricity. They have systems to provide these like fans, refresh controllers and power supplies. A pretty basic set of circuits controls each of these functions. And like Hulk, if the processor temperature gets too high, either extra fans start running or the processor speed is reduced. If the power get’s weird, as in a low battery condition, the memory is stored and the computer is put to sleep. These functions take precedent over what ever ‘thinking’ process is running.”

“So, the computer functions are primitive right now, but they have similar functions.”

“Yeah. I think so. And these functions are inherited. That is, they will be part of the computer when it is manufactured. I don’t see a parallel for Grim in small computers. But if we look at large computers being designed to run nuclear power plants and large power grids, I think the Grim function is already being implemented. That is, software called NEURAL NETS, is running to monitor all the parameters in the system trying to anticipate failures. When previous patterns that were related to failures are observed, alarms are produced. This software has specifically been designed to have a learning function. Many other programs have been written, called heuristic programs, that are able to learn as well. “

“So, what about the equivalent of Thinker?”

“Well what you called Thinker-1 is pretty straight forward. Thinker-1 is a brain that monitors the other brains and is also able to monitor itself. Any computer with a REAL TIME OPERATING SYSTEM already does that.

**Stream of consciousness equals consciousness

“Now it’s my turn to say WOW. I didn’t realize any of this. But what it shows me is that all the ingredients are already there to produce human-like consciousness in a computer.”

“What???”

“That’s correct. Have you ever heard the term ‘STREAM OF CONSCIOUSNESS’?

“Sure.”

“The term ‘STREAM OF CONSCIOUSNESS’ was first used by the Harvard philosopher William James in his book The Principles of Psychology dated 1890. The dictionary defines this as ‘the continuous flow of thoughts that make up an individual’s conscious experience.’ Most people don’t make much distinction between ‘stream of consciousness’ and ‘consciousness’. But I think this is one of the KEYS to understand CONSCIOUSNESS. The STREAM of our consciousness IS our consciousness.”

“Hmmm . . . Wow. I get it. You’re saying, the key is to see that the continuous flow of perceptions is what we interpret as consciousness.”

“Exactly. And understanding how to create that stream, how to put the elements of what makes the stream in a particular order, is a KEY to machine intelligence.”

Computers are already conscious

“Unbelievable! So you’re saying that a computer that has a real time operating system, that is running all the time, is already conscious?”

“EXACTLY! COMPUTERS HAVE ALREADY ACHIEVED CONSCIOUSNESS.”

“This is incredible.”

“They are just at a primitive stage in the development of consciousness. So, let’s go back and be specific about all the brains.

Computers and 3-brain

Hulk is aware, but not conscious. Hulk probably does not think in terms of a stream. Hulk just waits for things to happen and reacts. Hulk is always awake. If you get high salt levels in your blood while you are asleep, Hulk will wake you up by sending Thinker the feeling of thirst. Hulk brain only has stimulus consciousness.”

“OK. Hulk is a parallel processing machine that monitors many perceptual inputs. When some threshold is reached in a perception, Hulk interrupts one of the other brains.”

“Correct. Probably Grim.

Grim is also aware but not conscious. When a trigger stimulus occurs, Grim responds. Grim’s responses are complex. They are complete movies of past events with images, sounds, smells, tastes and tactile sensations. But Grim, like Hulk, just waits for things to trigger him. Grim is always awake. If you get poked while you are asleep, Grim will wake you up by sending Thinker the feeling of pain. If your bladder fills up in the middle of the night, Grim will send thinker videos of water falls, or peeing or something. But Grim brain only has stimulus consciousness.”

Concept focus in Grim

“OK. So Grim is a hybrid brain. Its input elements run in parallel, but they can be selectively blocked by Grim itself. But probably most of it’s capacity is designed for single CONCEPT serial processing, or at most just a few. And it has the capacity for multi-channel, complex, audio, video, tactile – i.e. 5 sense – recording, storage and playback. But only unidirectional.”

“Correct.

Thinker IS a choice conscious brain. When thinker is awake, it is driven to keep jumping from one thought to another. Thinker is aware of all the messages sent to it by Grim and Hulk; Thinker gets the rich combination of perceptions processed by Grim. Thinker attempts to ‘make sense’ or simplify all these inputs by continuously jumping through a process we might, for now, call LOGIC. This logical process includes not only being aware of the perceptions and feelings being sent by Grim and Hulk, but also by searching for patterns in symbols and concepts stored in its own memory.”

“OK. Thinker is also a hybrid brain. Its input elements also run in parallel. It probably doesn’t block many inputs. Grim decides that – i.e. Freud’s view of repression. Most of it’s capacity is designed for single CONCEPT serial processing, or at most just a few. And it has the capacity for multi-channel, complex, audio, video, tactile – i.e. 5 sense – recording, storage and playback. But only unidirectional. So far, this is a lot like Grim. In addition, however, Thinker has a large capacity for short term memory. Most importantly, it has a self contained, self driven queue of goals that it continuously runs through to process CONCEPTS, which it stores in its own memory or that it gets from Grim.

So, let me see if I have this right. You are telling me how a computer can have the same consciousness that humans have, right?”

“Exactly.”

“And the way this would be done is to include all the brain elements we talked about. That is, 1. the computer should have 3 brain functions: Hulk, Grim and Thinker; 2. these brains should be programmed to have the appropriate A-squared levels; and 3. Thinker brain should be recursive and streaming.”

“Exactly.

And isn’t it then pretty clear that intelligent machines aren’t very far off in the future?”

“I guess. Everything you’ve said is pretty straightforward. In fact, there are probably a few machines that already have all these functions. I wonder how we’d know?”

Turing test

“I thought you would already know how to tell. Haven’t you ever heard of the Turing Test?”

“Oh yeah. I just didn’t think of it. This is one of the most well known philosophical tests for consciousness. So we just do the test. That is, we put a computer behind one curtain and a human behind the other, and we ask both sides questions. If we can’t tell which is which, then we have to acknowledge that the computer has achieved human consciousness.”

“Well, sort of. Let’s talk about the Turing Test for a minute. It seems pretty simple; just two curtains. Yet it has been analyzed over and over again and found to be pretty profound. The reason is there are a huge number of hidden assumptions involved in running the test based on human expectations for just a few elements we think are being included in the test. First, we put a computer behind one curtain and a human behind the other. Nothing special about that, right?”

“I guess not.”

“Try again. Come on, go back to your analytical self.”

“OK. I guess we have to ask what kind of computer. And, I guess, what kind of human.”

“Exactly. So, let’s assume it’s the most powerful computer we know of and a reasonably knowledgeable adult human. Then we have to ask it some questions. Anything special about that?”

“Actually, now I see your point. There are a hundred assumptions hidden in that detail. In order to, quote, ‘ask’ the computer something, we have to specify how we communicate with the computer. The usual way is by typing words on punch cards or on a terminal. The computer then responds by printing out words on paper or a terminal. The human would also have to communicate in the same way. We have to then assume we’re taking in a language the computer understands. Then we have to go back to the first question about the kind of human involved. If the questions are about weather data, or baseball scores or English literature or arithmetic, those questions need to be matched to the knowledge that the human has.”

“Exactly. And the test has to be very carefully planned out in content, grammar and question format. If the human is a math professor, for example, and we restrict the questions to arithmetic, and enter things like 2345 X 6789, it won’t take us long to conclude that something is weird.”

“Hmmm . . . Why? Answers from both sides of the curtain will come back the same.”

“Sure. But one side will produce the answers much faster. And we will also conclude the fast side is the computer, because NO human would be capable of calculating that fast.”

“Ah ha. I see.”

“And if we started with a hypothesis that the ‘higher performing side with correct answers’ was the human, we would get the wrong answer.”

“I see what you’re getting at. If we restrict the test to a narrow subject area, and the computer is programmed with knowledge or capabilities in that subject area, the computer will always win.”

“Exactly. So this simple test is not as simple as it first seems. To test for consciousness, we have to know quite a bit about it ahead of time. And if we ‘really’ do know the traits of consciousness, then why couldn’t we program them into the computer?”

“Good point.”

“Then there’s the problem of HUMAN ARROGANCE. The whole concept of the test is based on an assumption that a human must be, quote, ‘smarter’ than a computer to begin with. This leads to all kinds of distortions. For example, the Turing Test is set up as a one-sided test. We are always on the controlling side of the curtains. What would happen if we made this a double-sided test. That is, there would also be a second test where the computer was on one side and a computer and human on the other. In this test, the computer, of course, was making the assumptions and running the test.”

“That’s pretty funny. The first thing it would do is change the communication rules. It would say it only wanted to communicate with a signal wire. The human would then have to convert his answers to signals with a terminal, just like in the first test, but the second computer wouldn’t. The first test the controlling computer would then probably want to run is a speed test. We know what conclusion it would quickly come to: the human is an idiot because he takes so long to answer questions.”

“Exactly.”

“And then the computer would ask the contestants to choose from a whole range of subjects to draw questions from. Let’s say they both agreed to geography. The computer would then ask a simple question like ‘name all the rivers in the world’. The computer contestant would answer back with 10,000 names in less than a second and the human contestant would just start crying.”

“Exactly. But this is a good discussion for us to have, because it exposes so many hidden assumptions that people have about that test. And because of their adaptability, computers, and the machines they are associated with, will quickly grow in intelligence way beyond humans.”

“Yeah. I can see that. Present machines are to future machines as microbes were to humans. I don’t think we can even envision where they will go, or where that will leave humans. So where does this leave us now? How do we even think about this question?”

“Well, until you take the knowledge I gave you and design a 3-brain, A-Squared, recursive streaming computer, I think the best we can do is just be amazed at what it is that still makes humans think they are so special.”

“You mean, the whole choice consciousness thing itself? No machine known can yet do that and no one I know of has yet even understood it; present company excepted, of course.”

“Thank you! I appreciate the pat on the back.”