Human Categorization, Activation, Familiarity and Learning
As some might have guessed I am an avid reader and developer of many different types of artificial intelligence. While Darren was busy putting two and two together with his math refresher course he is putting himself through he found the time to point out an interesting article and make some possibly interesting comments Linking morsels of informational data... sometimes learning is like a tantalizing treasure hunt. What he noticed is that the more he learns about math in theory the more tangible links seem to develop in terms of application. In fact many things that he has probably taken for granted have turned out to be enabled through the math he is only now learning.
The reason he is making many of these realizations is the ability for human categorization. We'll call this human categorization because it is conceptually and functionally different from other categorization algorithms. For instance, human categorization has to be trained and trained hard to be truly effective while other categorization routines that run in silicon tend to have rules that ensure some level of categorization the first time around. The training isn't as important because it is based on fixed rules rather than rules that change over time.
What Darren was realizing as he is re-learning trigonometry and algebra is that there is already a familiarity there. It is easier for him to learn the same concepts a second time through because the memory from the first time was never completely lost. Instead human categorization and learning is based around activation energies or weights. As he relearns the old stuff the weights increase over time and they now have more of an influence over the decision making process. This can be a good thing and a bad thing as we'll discover shortly.
Now, as the activation for these old memories begins to rise and new memories are created the brain is re-indexing many of Darren's past experiences. Most links in the brain are created through focus. As you focus you can direct your mind that certain paths of thinking are either correct or incorrect. This is the training as the brain is able to throw out the bad connections and only keep the good ones. Now what happens as you begin to forget your old math but you being to implement things that would possibly be much easier if you could remember your old math? Well, the brain still makes links. The proximity is always there only the activations are smaller now and the focus is missing. Your brain makes all sorts of low activation links. This is the power of the human mind to connect things early on and make use of the massive parallel processes of neurons, but not bother our focused thought process with the information.
That means Darren is at a crucial cross-roads. As he relearns the math, all of the past links start to come to the front of the mind (using very non professional terms). Some of these links are detrimental and some are helpful. Remember the brain made these links without training and so the links are very low activation and they can either be true links or phantom links. The links now have to be examined because they come rushing into the mind the more math (or other subjects) that are refreshed. Hell, sometimes hundreds of different past experiences can come forward over a single evening of study.
But if some of the links are bad and some are good, what is happening in the mind? Well, you have to be skeptical of the thought process. Remember these are age old mental links and normally you trust what comes out of your own mind. As each link gains in weight because of the association with the math processes you are learning you'll have to either enforce the association or break it. Breaking the association doesn't mean getting rid of the link, it means linking some really bad weight with it (at least that is what turn of the century research tends to show). Wrong things always have the ability to come back and bite you in the arse and one of the things you'll realize especially as you grow older is that it takes a much longer time to unlearn something than it does to learn something new.
Good links continue to grow in the mind. A year in the future those tenuous links that weren't important in the past will be just as vivid as the actual memories and links of the process you actually used. Looking back I can point out many places where things I learned within the last three months would have helped me. This type of cross-indexing where I can associate events from many years ago with things that I'm learning now is ultra-powerful.
This all speaks volumes about the huge parallel process that is human categorization. As we continue to learn new things or use old things those memories continue to be enforced and the activation energies increase. The ability to focus on and retrieve those memories increases and more importantly the activation speed or recall speed is improved that allows us to think quickly. While human categorization is a huge database of valid and invalid links it is through learning that our familiarity with valid links improves and the activation speeds increase. Only so much can be available at once and focusing on things that are important at a particular point in time is crucial to being efficient at our current task. The process of switching the active mind to a new palette of information is slow and unlearning old techniques in favor of new techniques is very slow and difficult. The neurons used for those old techniques may only be re-tasked after hundreds of hours of focused unlearning.
While human categorization works well and is powerful, the true potential of the system becomes apparent when more than one person works together and shares ideas. Often times the choice is between five or six different options and by comparing these options in parallel with more people the more agreed upon option will emerge. This would be similiar to merging the results of a search from several different algorithms and weighting entries based on occurence. Some of the text-search engines make use of parallel systems like this to make things work. Human intelligence really is a team sport and it is the result of contributions by billions of people. While computer intelligence can endeavor to approach and simulate human intelligence I have a nagging feeling that we'll find more break-throughs in focusing problem solving using methods that are more geared to way a computer thinks rather than the way humans think.