The future. The year 1997, or 2003, or 2011. Judgment Day. When Skynet wreaks havoc on the world, the machines rise up, and take us all out.
Or at least, that's how it happened in the Terminator movies and TV series.
As we know, it didn't quite work out that way. Machines are a large part of our world, just not in the post-apocalyptic sense. Don't believe me? Take a quick count of how many iDevices, Android, handheld, or other systems you own. I bet it's more than you thought. I myself own two laptops, a smartphone, a tablet, and an "intelligent" thermostat.
The thing is, we're not even close to having machines become smart enough to take over the world. Hopefully we'll get there in my lifetime: I'd love to see the day I could order my Butlertron to get me a coke, confident in his or her ability to do so. Before that day comes, before machines walk and bring and take action, machines need to do something important: they need to understand, like humans do.
The first thing machines need is advanced speech-to-text, what we humans like to call "ears". Machines need to be able to recognize the sounds coming at them, and differentiate between human sounds and other sounds. Then they need to understand the words themselves, and transcribe them to text with extreme accuracy. There's a ton of software out there like Apple's Siri, Nuance, Voci, and more, that already do this very well.
The next step involves understanding the context in which the text appears. Context, in fact, is perhaps the most important thing out there.
Why? Because context determines everything. This is where text analytics comes into play.
Let's look at an easy example:
- "Butlertron, can you pour me a glass of coke?"
- "Butlertron, can you score me some coke?"
In these two sentences, coke is a noun, and refers to an object of some kind. But they're two very different things, aren't they?
In the first, we're obviously referring to the beverage Coke, short for Coca Cola, and in the second one, we're referring to the drug, cocaine. Any human can look at those two sentences, and easily differentiate between them, based on the sentence.
For a machine, that's a lot of work. It's much harder, but very important. Imagine a future with robot policemen, hunting everyone that says the word coke. All of a sudden, half the world's population are drug addicts.
One of the fundamental aspects of Semantria is it's been trained to understand context. Semantria's powered by Lexalytics Salience, a system that ingested all of Wikipedia, and understands how its concepts are related to each other in real life.
Thus when Semantria is presented with the word "coke", it looks at its Wikipedia database, and finds all other words associated with "coke". For instance, it knows that "drug" is associated with "cocaine" aka coke, and "drink" is associated with "Coca Cola" aka coke. Just like a human digging into its memory banks.
It then looks at the words around coke. Is the word "drink" there? Probably a beverage. Is the word "snort" there? Probably cocaine (unless you like snorting fizzy drinks).
Back to our Butlertron. As you can imagine, text analytics and companies like Semantria will be crucial to their understanding of context. Once they can hear, and they can understand, what's left for them to do?
The final piece of our puzzle involves taking an action. This is where cognitive computing comes in.
Our Butlertron knows that there are two different types of coke. Context tells it to fetch the one associated with the beverage. It then makes a decision to go to the nearest storage place of this product: the refrigerator.
Good thing it understood right: otherwise I'd have created the first robot drug dealer ever.
So, we're still a long way off from an apocalyptic future where machines rule over us. We're still pretty far from robot policemen.
But I think Butlertrons are not so far-fetched. With cool technology like voice-to-text, and cognitive computing, and Semantria's contextual understanding and sentiment analysis, a future where machines think like people could be just around the corner.