Throughout history god is the only entity credited with creation that was more than the sum of its parts. That creation, humanity, has lived, thrived and survived; and now we have the potential to create beings that are also more than the sum of their parts. Or do we?

     According to Maureen Caudill in In Our Own Image an android must have several key characteristics (among others) it must be able to: accurately interpret what it sees, learn from its own experience, cope with unexpected situations, fully communicate in a spoken language, possess a reasonable amount of common sense, and obey basic principles of the social world. Some of these characteristics have been realized already while some of them are still in the designing phase.

     The field of artificial vision is making great leaps but for a computer to store what a face looks like it must store a picture of the face in it’s memory. The picture would necessarily be taken from a certain point of view and at a time when the subject looked a certain way. If, for instance, an android took a picture of me and wanted to be able to recognize me the next time we met it would take its picture of me and store that away. But what if before we met again I had taken a Florida vacation and gotten a tan and bleached my hair. I would look nothing like the android’s memory of me and would pass unnoticed. The solution, suggests Caudill, is to not store a picture of a person but to store identifying characteristics about me. This would ensure that I could still be recognized despite my changed appearance.

     Speaking of taking pictures, how would our android store this picture? The human brain holds an amazing amount of knowledge. Using a standard sequential memory structure found in personal computers would be inefficient due to the fact that the amount of memory needed would have to be huge to hold all of the information necessary. But sequential memory can’t be taught. The method that seems to hold the key to what we want is a neural network.

     Neural networks consist of an input layer of neurodes which are wired to an output layer of nodes in such a way that every input neurode is wired to every output node. Each wire has a weight associated with it. This comes into play when a series of input neurodes are stimulated. They send out messages to the output layer based on the stimulation of the input neurode modified by the weight of the connection. This forms a unique output pattern on the output layer which is how the neural network ‘thinks’. In addition, when a network correctly it modifies the connections so that the same output pattern will be more likely to be that outcome for the same input. If we were to train a network on the number 20; because the network modifies it’s connections the out put for 21 and 19 will likely be very similar. This technology is actually quite developed for some functions and is the study of much additional research.

     However the technology for verbal communication remains much less developed. Computers are able to be taught to understand one person or a very small group of people but the task of understanding may people is a much harder task. If everybody were to say the same phrase the words would be the same but the sound vibrations produced for the same phrase would be vastly different. We call these differences accents and intonations but to a computer these differences are the same as those between totally different words. Even when the same person is extremely tired or extremely active their voice sounds very different. Computers have great difficulty in separating accent and mood from pronunciation. Technology is in the works that acquires the speech pattern for a voice in about 100 words from a given user and while this is a great advantage; it needs to be a good bit better for any android which will need to have almost instantaneous recognition.

     Despite these complications to building an android, a greater one is this: how do we give an android common sense? Common sense is one of the greatest tools that we as human beings have. It allows us to adapt to different situations and unfamiliar settings. But because our android will find itself in all manner of different situations and settings how then will we decide which rules of common sense to give it? Giving our android all the rules for every situation is unnecessary and taxing on the androids resources. The truth is, the rules for common sense don’t translate well in to instructions that an android understands well if at all. Research continues, however, and solutions to this problem are expected to appear in the near future.

     As one can see almost all the parts are in place for an android to be constructed with a little bit of research in some areas. The act of combining enough systems to create a fully functional android will be an immense task in and of itself resulting in a creation which is complex beyond belief. We can be sure that such an interconnecting of systems will have unforeseen side effects in effect creating a personality for our android. It might always wipe its feet when it passes through any doorway or it might say every persons name with a southern drawl. These side effects aren’t something that was programmed into the android. They are in fact something that we are, indirectly, the creators of. They may even lead to creativity in our android which I believe would certainly carry the androids out of our control. But humanity has shown that it often does things for the sake of doing them. It is perfectly reasonable that someone would build an android just because it could be done. Caudill’s book raises a lot of similar questions and presents relatively few answers. Certainly it does answer the question how for off is an android in our future but then asks are we ready to deal with the consequences.