While AI has become the buzz word, it is still a mystery for most of us how machines can learn to be intelligent.
We are prone to look at it as some sort of magic that techie kids are performing, but the truth is Artificial Intelligence, at this point needs help, a lot of help from human intelligence, and in that process hides a miraculous solution for a problem that India desperately needs to solve.
Computers are often compared with human brain because both share a common trait. Both have capabilities of processing information. While computers have convincingly proven that they are capable of far great power of processing information, there is still one big problem that they can’t solve.
In simplest terms, computers don’t “know” reality. They need to be told what is what.
So, if AI has to grow and become capable of dealing with reality, the first requirement is to educate the computers about our reality.
A computer can process how many trees are there in a picture only of someone tells it what is a tree. Similarly, a computer can learn Urdu or French, provided someone sits and labels each sound. This list can go on, into every formation of information that we humans sense and cognise.
As rise of AI is inevitable, the task before humanity for next couple of hundred years is simple.
Humans will need to sit down and educated machines about the reality. In short, all the “data” of entire reality needs to be labelled by humans and communicated to the machines.
While machines themselves are also developing capabilities to learn the reality using their own “senses”, this process can be hugely augmented if humans help the machines, as humans are already reality-aware.
The beauty of this process is that it requires no technical skill to teach machines. Just being human is enough! And if you don’t believe it, mistype your password in Google Mail couple of times.
When Google asks you to prove that you are not a machine, all it does is to show you a set of images and asks you to identify an object therein. What you are doing there is data labelling, and by responding to the recognition challenge, you are actually helping Google’s AI to get more reality-savvy.
If we take this massive task of explaining reality to machines, it is actually a huge opportunity for India.
Data labelling can be a miracle solution to Indian unemployment problem as it is a humungous task and all that one needs to be employed in it is a human brain.
Data labelling needs no education, no qualification, no training. It just needs a human brain that can recognise reality and label it.
India had benefitted from last such large opportunity that was created by technology thanks to having a lot of educated kids that knew English. We are a nation that became visible in world technology space thanks to the Business Process Outsourcing (BPO) appeared on the scene thanks to internet and communication networks. It generated millions of jobs and the wave actually helped us develop a work culture and infrastructure.
Though the BPO model is now dying thanks to many other technologies that have emerged and will replace humans, it not only offers us a great example of the viability of such human-labour intensive work, it can become a foundation of yet another boom that can be a hundred times bigger, and that is Data labelling centres.
Data labelling as a business model comes with massive advantage to India as we have what it really takes, i.e. humans, but the same trait is a threat too.
While our English speaking skills allowed us to have an edge over other populous nations like China or Indonesia, in the data labelling space there is no such barrier. Any and every nation can join the game and compete.
The only advantage that we can have needs to be created by us, and that is moving in this space rapidly and building tech-infrastructure required for it.
By recognising and supporting data labelling business, India can give employment even to a kid sitting in remotest corner of India provided we can make internet and computers reach there. It is a business that can be taken down to grass-root level by building processes for quality control.
As our reality is a massive place, there is no dearth of data that needs to be labelled. With right tools and processes, humans can do this from homes or while waiting for a bus.
It is a massive endeavour with its own trials and tribulations. It may even lead to machines more reality-savvy than us and become our masters, so there is a Frankensteinian possibility embedded in this opportunity.
But, for India, there is actually no real option.
We need to join data labelling at a massive scale and use that as a foundation to enter the AI space. If we do and succeed, we have a future, but if we don’t, we are doomed in a future that belongs to machines.
DISCLAIMER : Views expressed above are the author’s own.