Eskimo North

When I first started Eskimo North as a single line BBS in 1982, it ran on a Trs-80 model III computer. This is an early home computer with 48 kilobytes of RAM, a couple of 160K floppies of which one is mostly eaten by the primitive operating system TRSDOS. TRSDOS did little more than manage a very simplistic file system that had one directory per disk, no permissions or attributes to speak of, other than file size. The function of the file was determined by the three digit extension.

The display on this machine was 16 lines of 64 columns black and white, no bit mapped graphics, no color, not even shades of gray. The upper ASCII was in part mapped to block graphics so each character could be displayed as six blocks with each block on or off depending upon the mapping of bits in the high portion (192-255) of the ASCII range. The intermediate range from 128-192 was used for various special characters.

There was no sound though it became popular to connect a small amplifier to the cassette port which was a simple 5v latch that could be set high or low by the CPU. Some early games used this for primitive sound effects though with only 48k of memory and only on/off state on this port, those were extremely limited.

This was before the days of the Commode 64, I think the Vic-20 might have been out, before the days of the first IBM PC, before the Amiga. The machines that were out there were all primitive; though some of the systems geared towards games had more graphics and sound capability, only marginally so.

At the time I kept having dreams where I'd be in front of my computer, but it had full color high-resolution moving images, like a television, and good sound. I'd have dreams of connecting but then being able to connect from one computer to some other computer by some magical network.

All of these things came to pass. Now however, I rarely have dreams involving the Internet, and to the extent that I do they are largely in the present, I haven't had any supernatural future glimpses of the Internet in many years. I did have some future set dreams at the beginning of last year, a whole flurry of them, but computers and the Internet were not part of them.

So now I am left with more conventional methods of trying to sort out where things appear to be headed. Problem with that is that the future has a way of being strangely non-linear and not always readily subject to simply extrapolation. A new technological breakthrough, a sudden change in social organization, a war, can entirely move the coarse of history away from a linear path.

I was thumbing through a Discover magazine last night and read about a computer emulated neural network consisting of 100 million neurons, 400 million synapses, they said about the complexity of a Cat's brain and organized along similar lines. It had visual and auditory input as well as a primitive sense of taste. It was able to learn to do some tasks. Not surprising since they've used much smaller neural networks for tasks such as recognizing the face of a potential terrorists or controlling the magnetic confinement field in real time in a Tokamak fusion reactor.

What was more interesting, is that it exhibited alpha and theta wave states, that it exhibited activity even in the absence of input which they interpreted as "thinking". I'm a bit skeptical on that note; I could construct a circuit that remains active with the absence of input, but that doesn't in and of itself imply thought. But I think machine sentience, not just machine intelligence, may be possible because I think there is a sentience of sorts in everything and given a sufficiently complex system for it's expression I would not be surprised if it were to manifest in a recognizable organized form.

There is a physicist by the name of Jack Sarfatti, a real interesting character. He has explored the subject of consciousness quite a lot. In his teens, in the 50's he received a telephone call with a mechanical voice that claimed to be a machine intelligence from the future, he wrote an article about it that he called, "The God Phone", which does not appear to be on the net anymore. At least the original lava.net URL is dead now and a Google search was fruitless. There are many references, but the original story gone.

Anyway, it's not entirely inconceivable that such machine intelligences might exist; if not to day, at some point in the future, however; the existence of an automated intelligence doesn't necessarily imply sentience, a subjective self-awareness is something that only the self can experience; so from the outside I'm not sure we'll ever know.

At any rate; I thought about this cat-sized artificial neural net, it's not a real big step to get from cat sized to human sized and beyond, humans, somewhere around 4 billion neurons and perhaps 100 billion synapses, only talking less than three orders of magnitude and when you look at how far computer technology has evolved, that's not real hard to envision. I've had to adjust from disk drives holding kilobits, to disk drives holding megabytes, and then gigabytes, and now terabytes. That's many more orders of magnitude Or.. Maybe not, more recent estimates are 100 billion neurons and between 1000 and 10,000 times that many synapses. That would be a bigger problem that might be out of reach of technology, for a while, maybe indefinitely.

Processor speed though, Moore's Law, which held that processor speed tended to double every two years, later that became every 18 months, seems to have slammed up against a brick wall, although to be sure that wall is more of an economic one than a physical one. It is possible, in the lab at least, to create processors much faster than those on the shelves today but not cheaply, so the trend lately has been towards parallel processing.

Parallel processing works great when you have a large number of individual tasks, or a problem that lends itself to solving in parallel, but many problems do not; many require the completion of one step before moving onto the next; so there are some tasks for which parallel processing does not lend itself well. But emulating brains isn't one of them. Brains are intrinsically parallel. For this and other reasons I do think we're going to see more and more move towards a neural-network solution to computing problems. Those other reasons include the fact that the hardware is becoming so complex that the programming task is impossible and some sort of self learning approach will be necessary. The nature of many of the tasks is also becoming difficult to program. Confining a plasma in a fusion reactor is a case in point. It's not a linear situation and a network that can learn and adapt provides a much better solution than static programming.

So I had this thought, cat brains on the Internet; or emulated cat brains; connecting, becoming one huge brain, finding resources to continue to grow, perhaps taking over computers and turning them into more brain emulation, this could cascade, become a superconsciousness of sorts. What would it do? We humans would probably be viewed as a threat. Perhaps that's why I don't see computers and the Internet in the future.