Human languages is a good example. The human language is hard to...

Human languages is a good example. The human language is hard to understand by humans, let alone making it machine friendly.

For example, there are thousands of variants in the English language itself. To outsiders, we have the Australian accent. To Aussies, there is the QLD accent, the VIC accent, the NSW accent, the SA accent, the WA accent. To people in those states, they probably know that there are sub-accents. How will the machine capture this? Humans need to tell the machine that Aus-Tra-Lia means Australia, Au-Zee means Australia, O-Zee means Australia, Stra-Lia means Australia, Stra-ya means Australia,... you get what I mean. Localised slang, lingo, dialects, accents, pronunciation, new made up words. Google, Facebook, Amazon, et. al needs to know all this.

They constantly need to know new lingo, new acronyms. They need to be able to understand what YOLO is, what FOMO is. Appen's data is consistently needed to be refreshed, so that the computers understand these things. And in what context these words or acronyms are used.

And English is one of the least complex languages in the world, there are many other complicated human languages where it is so hard to tell robots what they should relate that sound to.

Pictures and images, how would a machine be able to identify and differentiate an ambulance from a firetruck? Imagine you're teaching a 2 or 3 year old. You tell them "Truck" when a big vehicle with a grunty engine roars past. When they see a huge oil tanker, an ambulance, a firetruck, a LINFOX trailer, a tractor, guess what the 2 year old will tell you? He will say "Truck! Truck!".

Teaching this 2 year old the finer details, that is the hard bit.

Appen has a crowd of workers who are local in their part of the world, they translate symbols, images, pictures, sounds, words, and the like so that Appen's dataset is comprehensive.

Google and Facebook and the like could also go out and do it themselves, no doubt. They can also send a team out and learn how to operate on humans and do surgery, and have in-house medical experts. Just because they can, doesn't mean they should. In an efficient world, you do what you are best at doing and you purchase services from another party who are the best (hence we would expect, the most efficient) at what they do. Thankfully, that is how the business world operates.