Almost everyone has heard the term “machine learning” thrown around. It’s used in some vague context about how computers and/or technology are suddenly and magically going to make the world a better place. That sounds cool, but what exactly does “Machine Learning” mean? From 1959:
Machine Learning — a “Field of study that gives computers the ability to learn without being explicitly programmed”.
Computers (or code, or machines in general) are terribly literal devices. They do exactly what they are programmed to do and nothing else. While most people don’t realize it, this is one of the biggest sources of frustrations with computers. But what if they didn’t have to be so literal?
Rather than talking about machine learning in the conceptual let’s look at an average morning. Then we’ll come back and apply a heavy dose of theoretical machine learning.
- Set an alarm for the morning. Go to sleep.
- Alarm goes off, get up.
- Turn on coffee maker, make a pot of coffee.
- Get ready.
- Leave for commute to work.
Even trying to “keep things simple” I’ve provided an example that leaves huge gaps in most peoples routines. Sure you set an alarm, but when do you set it for? Do you live somewhere that has inclement weather? Maybe the occasional snow storm means that you need to leave a little earlier — therefor you have to get up earlier. How strict is your schedule? Do you have morning meetings? An extra early meeting? You might work at a relaxed startup where you can roll in 15 minutes late because of the rain, but if you’ve got an 8:30 AM meeting it does you no good rolling at 9:15.
Since I can’t possibly make up a sample schedule that applies to every single reader, stop for a few seconds and think about what’s in your morning routine… even those things that maybe you need to take care of the night before? There’s probably no shortage of “small” tasks that you think about, take into account or manage.
So let’s go to the machine learning version of the morning. For sake of argument, we’re assuming that every electronic device in the house is “smart”. Not only are the devices smart enough to learn for themselves, they are capable of talking to the other devices. Not Jetson’s level of a totally integrated house, just appliances that can talk to one another.
- Go to sleep.
- Your alarm clock checks the traffic and weather reports overnight and decides on the optimal time to wake you up based on you morning routine and estimated commute time.
- 15 minutes before your alarm goes off, the bedroom lights slowly light up and transition to a full glow, simulating sunrise (your lights asked your alarm during the night to let them know when wake up time was).
- 5 minutes before your alarm goes off, the coffee maker kicks on and makes you a cup-a-joe (the coffee maker has also talked to the alarm clock).
- Your alarm goes off.
- Right after you pick up your cup of coffee, your sound system starts playing quietly your headline news (the coffee maker lets the sound system know when coffee was picked up).
- As you start to get ready for the day, your cellphone chirps to remind you when you should leave and to take your umbrella — it’ll rain this afternoon.
-
On your way out the door your lights automatically turn off. You also notice that you’ve got an email from Amazon for a re-order of coffee pods (your coffee maker had determined you were running low).
Some would argue that it’s easy to do a fair amount of the above without the “smarts” today. An alarm clock, light timers and coffee makers with alarm clocks aren’t all that hard to come by. Those skeptics would be 100% right. This isn’t a suggestion for people to run out and replace every electronic in their house for a slightly more smooth morning experience — that’d be silly. However, it is an example of one of many possible benefits to the “smart” and “connected” home of the future.
Let’s be clear though, no one wants to program the time on the VCR. Nor does anyone (myself, a die hard techy, included) want to manually program every single device in the house. Machine learning in the home would mean that each device learns based on needs and usage. If I have a cup of coffee every morning as soon as I get out of bed, and then listen to the latest tech news — the house should learn that’s what I like and do it before I ask.
Oh and remember those NEST thermostats? The ones you don’t have to program because they learn your schedule and likes? Yeah, machine learning.