Got to play with this on an interactive white board today. #Captivating #Want
(image via IEEE Spectrum)
The holy grail of machine logic since before the invention of the first rudimentary calculators has been the creation of something that could mimic the brain. It’s always been ‘right around the corner’ since the development of the first vacuum tube to the development of the transistor and we’ve made a lot of advances in Artificial Intelligence since the first computer chips were developed. We have computers which can recognize your face or parallel park your car or even sort out your email inbox based on what you read first when you open it up. Just the same, our computers and the programs they run don’t have the sort of versatility that we come to expect even from our average cheese-stealing rodent.
There are several reasons for this. Firstly, computers and software are organized around Boolean Logic. Any highschooler suffering through a tenth grade course on the subject will tell you that there’s very little in common between the way our brains process information and the way we program our computers using Boolean routines. The second hurdle for creating a good brain analog is purely a hardware issue. When one neuron sends information out, its target takes that information and processes it based on its own current state and how applicable the signal is to that current state. In simpler terms, a neuron acts as both a processing circuit and a memory storage circuit at the same time. The hoops you have to jump through to get a transistor-based architecture to do the same thing are incredible.
Modeling a human brain with current transistor-based technologies would take a supercomputer the likes of which our civilization has never seen. It’s inefficient and simply infeasible. Researchers at Boston University are working on an interesting alternative. They’re using a new circuit component, called a memristor, which can pass a current which varies with the voltage applied (like a normal resistor) but which has a resistance that varies depending upon the state it was in before voltage was applied. So, much like a neuron, it can both process and remember information all at once. The scientists at BU are combining this hardware with a specialized programming language that uses a multi-core processor teamed with memristors to simulate both neurons and synapses. They hope to soon be able to create a purely artificial analog to a simple brain like the one you’d find in a small mouse or a cat – all in a package the size of a shoe box.
This gallery contains 1 photo.
Conrad Wolfram on Teaching Mathematics in School
Loved this. Especially 3:20 to 4:50. His four step procedure is more or less the process of quantitative research, and Steps 1, 2 and 4 are really where human input is required.
I do think understanding calculation is important, but rather than teaching, say, long division in a rote fashion, getting students to write a program to do it would provide greater understanding.
As a kid, my first exposure to programming was turtle graphics. An excellent way for kids to learn programming and geometry at the same time.