Image Recognition

Next time when you go to a Bank of America ATM you are up for a surprise. The envelopes are gone. To deposit a check you insert it into an opening. The machine makes a little noise and up on the screen you see a check. And not only that, but the correct amount has been filled in for you.

Even a seasoned techie like myself felt a sense of magic the first time I saw it in action. Sure, I remember taking the image processing class during my Computer Science masters. I know that you can use variety of techniques ranging from neural nets to cellular automata-based noise filters to "teach" computers how to recognize images. Yet, to see image recognition applied on a mass scale by Bank of America is nothing short of a marvel.

I had a bunch of checks with me ranging from printed ones to handwritten ones. Each time the ATM displayed the correct amount on the image of the check. It didn't make a single mistake. And the whole process felt like a big improvement: No envelope, no numbers to tally up. Bank of America used image recognition to simplify customer interactions and to truly raise the bar.

Touch Screens

Since the first day I got my iPhone I've been a happy customer. I've written all about why that on is this blog, and in many ways, the iPhone is a ground breaking and game changing technology. But the fact is that almost six month later, it just works. Everyday I use my fingers to operate its screen and every day it does what I intend it to do.

What other touch screen technology was ever this flawless? Remember Palm Pilots? They worked okay, but not nearly as perfectly, particularly when you used fingers instead of the "magic wand" (stylus) that came with it. The fact is that Apple is the first company that made interactive screens work so well that they could take it to the mass market and offer a touch screen phone, music player, and web browser in one device.

The simplicity and ease with which the Apple UI responds to our fingers is deceptive. Beneath the surface there is a sophisticated algorithm that took years to calibrate and perfect. It is like that with many other Apple technologies - the elegance and correctness are results of the years of a culture of perfecting every detail, every algorithm.

Voice Recognition

About ten years ago I got my hands on one of the first releases of voice-to-text software called Naturally Speaking. I was intrigued by the promise of software that could type my thoughts as I spoke, secretly hoping that I could dictate C++ code into it. Alas, it worked only about 85% of the time, which was frustrating enough to make me stop using it. I called up the company and they explained that my accent was harder for the software to understand and that they would be dealing with foreign accents in future releases.

Fast forward to 2007 and I am getting a tour of my new Acura from the dealer who just sold it to me. He explains that Acura has one of the best voice recognition technologies on the market. When I give it a try, it understands my commands with 100% accuracy. Literally, it never makes mistakes. I even had my 83 year old grandpa, who speaks English well, but with a much heavier accent than mine, speak commands. Flawless. The car just gets it.

And there is a substantial improvement in voice recognition across the board. Have you noticed that automated travel agents and customer support reps are not as annoying? This is because they now understand us better.

Biometrics

The next technology which is likely to hit the mainstream is biometrics. The idea behind biometrics is that people can be uniquely identified using a combination of bio markers - fingerprints, shape of the ear, voice, eye retina, etc. While there are definite privacy and safety concerns, like with many technologies, the benefits are likely to prevail.

With biometrics in place, plastic will become obsolete. For example, in the same Bank of America ATM you will be able to use your thumb instead of your bank card. In the supermarket, you will not need a debit card, and in the department store you will be able to buy things without a credit card. So it would, in fact, be okay to leave home without your Visa.

Conclusion

As the title of this article says - good things take time. Early implementations of most technologies are not stable and not suitable for mainstream use. Individuals, teams, and companies spend time perfecting, simplifying and polishing new technologies, and finally, after several years and iterations, the innovations hit the mainstream. Of course, by this time, we have all heard of the technology and take it for granted. We are no longer surprised by the name, but we are surprised by how well the technology works.

Image recognition, touch screens, and speech recognition have been long considered domains of Artificial Intelligence. To most of people, all this sounds like magic. But it is not. And it isn't AI either. These technologies are marvels of modern engineering and human ingenuity. It is our restless drive to innovation and business incentive that makes it possible to turn abstract ideas into mainstream technology.