It seems difficult for little human children to learn math, but if they learn numerical values first in tried and true tested learning sequence, they learn very quickly. Artificial Intelligence for the most part is very good at this because its entire bases of calculating is using math. I’d like to take this dialogue to a higher level if you’d care to indulge me for a moment.
You see, there was an interesting paper I read recently on a sub-topic of Learning Sequence titled; “The Logarithmic-To-Linear Shift: One Learning Sequence, Many Tasks, Many Time Scales,” by Robert S. Siegler, Clarissa A. Thompson, and John E.Opfer which noted the time it took for students, starting at the elementary school level to learn about numeral values, such as the how the number 6 related to six-objects, and how number valuations related to each other, for instance 4 is less than 6, and then how 6, 60, and 600 increased valuations and then how long it took students to recognize all the relations between numbers, objects and values into their adult years.
The uptake of information for a second or third grader was relatively quick at the outset, although looking at it as adults we’d assume it to be slow, but our understanding about values and numbers over our adult years didn’t match the rapid advances from knowing nothing to understanding a good deal.
Now then if we look at machine learning, all we can see that a machine can learn things faster provided there is input to a point of being able to make huge short-cuts, better yet, consider the use of quantum-mechanic strategies to discover relationships at a huge speed, almost to the point of hopping forth while computing using prior calculations plus the ability to compute so much faster, thus, it too jumps forward incredibly fast the more it is used, and is able to process much faster than using only ones and zeros.
Thus, it simulates a child’s learning of numerical values (after programmed) to fill in further calculations, but can continue its calculating without slowing down through a much longer “power band space” and yet, although continuing to move fast through iterations, eventually starts to slow due to the vast amounts of calculations. Best of all, it does this amazingly fast, perhaps milliseconds at first, bursting onto the scene like the big bang and continuing outward extremely fast but not similar to the acceleration of the starting of the process.
Still, it didn’t start as a clean slate with regards to numbers as a young child first introduced to the concept, but then again a young child just starting pre-school knows how old they are and holds up 3-fingers and states with pride; “I am this many” and pronounces the word three the best they are able. The relationship between computer learning and human learning when it comes to math and numerical value is similar and not – both at the same time. Almost like right-spin, left-spin, both, and perhaps this is due to context of how both machine and mind think.
Getting computers to think and reason by context, object, observation and orientation might be tedious and challenging, but could prove to be another way to better match the Turing Test goals than to continue forever with endless tables and data bases, continually creating more into infinity. Why not do both? Why does it take a human to decide what a robotic system should do? Why can’t the system figure that out, including the formatting of its own system – there has to be a better way – please consider all this and think on it.