select all

The End of Code Is Really the End of One Guy in a Garage

Did you hear? Code is dead. There’s no more code. If you were learning to code, you screwed up. Less than a year ago, Bloomberg was answering the question of “What is code?” and now we don’t even need it anymore.

The latest installment in Wired’s series of bold proclamations claims that increasingly capable artificial intelligence and machine learning signal the end of computer programming in its current state. (In 2010, the magazine announced that the web is dead, which was mostly a semantic dodge about how the web is just a subsection of the internet.)

Jason Tanz writes:

Over the past several years, the biggest tech companies in Silicon Valley have aggressively pursued an approach to computing called machine learning. In traditional programming, an engineer writes explicit, step-by-step instructions for the computer to follow. With machine learning, programmers don’t encode computers with instructions. They train them. If you want to teach a neural network to recognize a cat, for instance, you don’t tell it to look for whiskers, ears, fur, and eyes. You simply show it thousands and thousands of photos of cats, and eventually it works things out. If it keeps misclassifying foxes as cats, you don’t rewrite the code. You just keep coaching it.

Tanz is correct that machine learning is a new (for most people) and exciting way for humans to interact with computers. Instead of requiring coding knowledge, programs, functions, and algorithms can be improved more or less by training them. (Probably worth acknowledging here that this sort of “black box” coaching — in which the instructor cannot predict the program’s behavioral outcome with 100-percent certainty — is also very worrying.)

Large companies are already doing this, and the thing that has gone unmentioned in Tanz’s piece is that, well, only large companies can do this. Machine learning has been making strides in recent years “thanks in part to the rise of deep neural networks, massively distributed computational systems that mimic the multilayered connections of neurons in the brain.” That requires large data sets and a lot of physical resources — you can’t run a neural network from your phone. The implication is, then, that if this new type of code-less computer programming makes programming accessible to everyone, it is because Google or Facebook or Microsoft, or whoever, is providing the infrastructure to do so.

Silicon Valley is often built on the myth of some schmucks in a garage. Apple was launched in a garage, as was Google, as was HP. Most of these myths are falsehoods, but the notion that one or two people with grand ideas and enough coding skill can change the world and “disrupt” old ways still pervades. Wired’s conception of the end of code is also the end of a tech industry that doesn’t require substantial amounts of capital, physical space, and back-end hardware to simply get off the ground.

So, yes, more people might be able to create and “train” computer programs without any knowledge of coding required, but they’ll be doing it on the backs of the organizations that have built and control these neural networks.

Which brings up a further point: Who constructs and maintains these supposedly revolutionary neural networks? The answer is developers and engineers — people who possess the sort of specialized computer know-how that Wired argues will soon be a thing of the past. That’s not eradicating code as we know it; that’s sending it one level deeper into bureaucratic obfuscation. If Wired is right, machine learning will reduce the number of people who know how to code, but, at the same time, it will make those who do even more powerful.

What the End of Code Really Means