Artificial intelligence, why you so stupid?
At this point in time I would have guessed we would have robots doing all sorts of work, making the quality of life for humans a lot greater. So why is it that we don’t have robot nannies and assistants yet? Why is it that when you play a video game the enemies rarely seem to exhibit any sort of realistic behavior in terms of their personality? Instructions.. A programmer has to program instructions for the computer and that is how AI works.
It would take forever for a programmer to not only create the instructions an adult human has learned, but also the intelligent associations those instructions have with each other.
To make matters more difficult, our language is still evolving. All the time there are new words people are coming up with, slang words, and new ways of describing things.
The instructions are relative to the context of an object, for example if you tell a computer a strawberry is red, but there are different colored strawberries, the computer won’t recognize the new strawberry until you write out a sort of definition for it that it can understand.
For a long time I haven’t viewed anything really promising, back in 2009 there was a tech demo from a well known game designer/developer Peter Molyneux. Milo was a promising tech demo, and I think its well worth a look if you haven’t seen it before.
Somehow they are able to create a character that seems fairly convincing, but how much milo can understand that is outside of the scope of his world is a very interesting thing to look into.
Perhaps when you start talking to milo enough, you will find that he has a number of default things he might say when he isn’t programmed to understand a subject. Say you start talking about 3D graphics to him in detail, explaining how programs work. Will he retain and be able to understand what your saying? Chances are he will only be able to intelligently respond to a few things in context.
Something else interesting to check out in the world of AI, is this little website called cleverbot.
cleverbot is an experiment in AI that can learn the more you teach it. Every time someone interacts with cleverbot through the web, the information is stored in its memory. How well it can create intelligent associations is of question.
We all know how hard it is to have your sarcasm be interpreted correctly through the internet, and if clever bot cannot associate the meanings of words and concepts correctly with all the little subtle nuances in the way humans express themselves, then it will fail ultimately as an intelligence and will simply be a database of human phrases.
When you ask it questions, it will simply reply with random unintelligible responses, which it seems cleverbot mostly does. Though it really does seem to be building some associations, but how much of that was programmed in rather than learned on its own.
Also, it must have some personality programmed in or it will be useless as a tool. If it was a huge database of human knowledge we could ask it to help us with questions, but if it had no personality it might choose to answer us with valid or invalid information. How it sorts information is viewed as a personality to us, does it answer with good information or does it answer with bad information.
If we asked the AI a question about how meat is processed in a factory, and it responds by explaining where we can find a delicious juicy hamburger at our nearest mcdongles, then it is possible the AI was programmed by someone to say that. If the robot was being sarcastic and this happened one time that would be interesting, but f it was consistently happening I would imagine someone programmed it to try to advertise products to you.
Well a huge database needs to be created, and that is the problem. No one has figured out how to make such a sophisticated database, that can actually have so many instructions it can know how to respond correctly.
Maybe someday soon though, we have come so far! Here is another video of Milo being shown at a TED talk.
