Small Numbers, Huge Changes

In a recent interview, Sundar Pichai of Google discusses improvements in the accuracy of their voice recognition:

Just in the last three years, we have taken things like error in word recognition down from about 23 percent to 8 percent.

That’s the difference between misunderstanding one word in four, to one word in twelve; the difference between completely unusable, and annoying.

Andew Ng, formerly of Google and now of Baidu, expands on this:

Most people don’t understand the difference between 95 and 99 percent accurate. Ninety-five percent means you get one-in-20 words wrong. That’s just annoying, it’s painful to go back and correct it on your cell phone.

Ninety-nine percent is game changing. If there’s 99 percent, it becomes reliable. It just works and you use it all the time. So this is not just a four percent incremental improvement, this is the difference between people rarely using it and people using it all the time.

It’s fascinating to see how these small numbers make a huge difference; you might think Google’s 92% accurate is only a little less than Baidu’s 95% accurate, but in practical terms there’s a big gulf. And it gives me pause to think about the money, human resource and computing power spent on trying to make those small huge increases in accuracy.

An interview with Andrew Ng

The Huffington Post series, Sophia, brings ‘life lessons from fascinating people’. Their latest interview is with Andrew Ng, a Stanford University professor, co-founder of Coursera, and key member of the deep learning teams at first Google and now Baidu. I really like some of the insights in his interview, the practicality with which he treats innovation and the easy way he explains hard concepts.

For example, here he talks about career advice:

I think that “follow your passion” is not good career advice. It’s actually one of the most terrible pieces of career advice we give people. Often, you first become good at something, and then you become passionate about it. And I think most people can become good at almost anything.

On retraining the workforce for a more heavily automated future:

The challenge that faces us is to find a way to scalably teach people to do non-routine non-repetitive work. Our education system, historically, has not been good at doing that at scale. The top universities are good at doing that for a relatively modest fraction of the population. But a lot of our population ends up doing work that is important but also routine and repetitive. That’s a challenge that faces our educational system.

On why machine learning is suddenly more popular, despite the technology being around for decades in some cases:

I often make an analogy to building a rocket ship. A rocket ship is a giant engine together with a ton of fuel. Both need to be really big. If you have a lot of fuel and a tiny engine, you won’t get off the ground. If you have a huge engine and a tiny amount of fuel, you can lift up, but you probably won’t make it to orbit. So you need a big engine and a lot of fuel. We finally have the tools to build the big rocket engine – that is giant computers, that’s our rocket engine. And the fuel is the data. We finally are getting the data that we need.

And the challenges of talking to computers:

Most people don’t understand the difference between 95 and 99 percent accurate. Ninety-five percent means you get one-in-20 words wrong. That’s just annoying, it’s painful to go back and correct it on your cell phone. Ninety-nine percent is game changing. If there’s 99 percent, it becomes reliable. It just works and you use it all the time. So this is not just a four percent incremental improvement, this is the difference between people rarely using it and people using it all the time.

It’s a really interesting interview, I encourage you to read it in full.

The Future Is Coming Faster Than We Think

Today I read a fascinating article in the London Review of Books. The Robots Are Coming, by John Lanchester, is about the rise of cheap automation and the effect it’s going to have on the workforce and society at large. In his introduction he talks about the Accelerated Strategic Computing Initiative’s computer, Red, launched in 1996 and eventually capable of processing 1.8 teraflops – that is, 1.8 trillion calculations per second. It was the most powerful computer in the world until about 2000. Six years later, the PS3 launched, also capable of processing 1.8 teraflops.

Red was only a little smaller than a tennis court, used as much electricity as eight hundred houses, and cost $55 million. The PS3 fits underneath a television, runs off a normal power socket, and you can buy one for under two hundred quid. Within a decade, a computer able to process 1.8 teraflops went from being something that could only be made by the world’s richest government for purposes at the furthest reaches of computational possibility, to something a teenager could reasonably expect to find under the Christmas tree.

This makes me think of IBM’s Watson, a deep learning system, ten years in the making at a cost in excess of $1 billion, with hardware estimated at $3 million powering it, and coming soon to children’s toys.