Khan Academy

This article, really irritates me.

It leads with a good point: the problem with math education is that it fails to connect math to the real world, and the author claims that Khan Academy has this same flaw. Fair enough. I agree that the central problem with math education is that students see no reason to learn math. Learning is hard, it requires a certain amount of struggle, and trying to get anyone to learn something they do is unlikely to produce lasting learning. They may memorize some procedures for a test, but they won’t be able to use it in their life or build on it in the future. This has little to do with Khan academy except insofar as Khan academy tends to teach a series of techniques without really motivating them or answering the larger question of relevance. I don’t know that this is necessarily true.

This is where things go down hill.

The next critique is that Khan Academy hired computer scientists instead of teachers. The author thinks teaching should be improved by helping to hone the craft of teaching. But let’s get real, we have been honing the craft of teach for a few hundred years now, how much better is it going to get. Khan Academy is about the medium, not the message. They are not trying to reform the math curriculum, however desirable that would be. They are trying to change the process by which people learn.

Teaching is currently done in a non-scalable way, and software and video is completely scalable. A video, once shot, can be replayed a million times at any speed. A program can be executed almost for free. No one thinks that videos or programs can compete with the best a human being can do, but they don’t need to—most people don’t have Richard Feynman as their high school physics instructor. The point is that despite how inferior video is as a medium it is possible to take the best possible video of instruction that mankind can produce and let the poorest student in the world have access to it (yes, assuming they have a computer…but you get the point).

If someone has an idea about how to re-orient the math curriculum, they can make their own videos and programs and give them to the world. The point about scalability is that only one person needs to do this for everyone to benefit.

And the rest of the article is just the silly teacher’s union stuff about how teachers, unlike every other profession in the world, can’t possibly be evaluated. The fact is, teachers can be evaluated, and good teachers make a bigger difference than anything else in how much students learn. Standardized tests aren’t perfect, but they are the best tool we have for identifying good and bad teachers. I think this paper shows fantastic work to identify the impact of good teachers. The obvious fix, to reward the good teachers and replace the bad ones, isn’t possible in most places because of politics, which makes teaching somewhat unique in that respect.

One of the main sinks of energy in the “developed” world is the creation of stuff. In its natural life cycle, stuff passes through three stages. First, a new-born stuff is displayed in shiny packaging on a shelf in a shop. At this stage, stuff is called “goods.” As soon as the stuff is taken home and sheds its packaging, it undergoes a transformation from “goods” to its second form, “clutter.” The clutter lives with its owner for a period of months or years. During this period, the clutter is largely ignored by its owner, who is off at the shops buying more goods. Eventually, by a miracle of modern alchemy, the clutter is transformed into its final form, rubbish. To the untrained eye, it can be difficult to distinguish this “rubbish” from the highly desirable “good” that it used to be. Nonetheless, at this stage the discerning owner pays the dustman to transport the stuff away. Quote from the fantastic Sustainable Energy Without The Hot Air by machine learning research and physic professor David Mackay.

Engineers and operations people are complementary goods to datacenters

There is a common perception that “cloud” computing will make operations people unnecessary. The reality is the opposite, I would expect the demand and price for operations people to rise on the whole. It is easy to see why: the final product of “cloud”-style data center operations and software engineering is a website or service. Insofar as “cloud” computing is successful it will lower the cost of running a data center by removing direct interaction with lower-level infrastructure. When the price of making and running websites and service goes down you expect more of them (i.e. demand goes up). This means the demand/price for anything else needed to run a website goes up to. It turns out the other thing you need to run a web service is engineers and operations people. This is what is known as a complementary good (see the wonderfully readable textbook Principles of Economics if you are interested in this kind of thing).

But all operations people are not created equal. Some are actually being automated. Certain lower-level systems tasks are being replaced. The people who build out servers, data centers and networks are being replaced by centralized teams run by Amazon. The interaction with these things is now via an API.

So systems operations will be impacted, but by no means replaced. You still absolutely need a networking expert if you run on Amazon you just no longer need to build out networks on your own. The result of this is that I expect most operations people in those areas to move up the stack.

One other impact is that because many services like machine allocation that used to be manual are now behind APIs, I expect the importance of programming for operations people to increase. So if you are an application operations specialist and want to future proof your skills, make sure you are a solid python programmer.

Credit to Eric Sammer for reminding me of this.