Search Engine Optimiser was not in the job lexicon a decade ago and a decade hence, probably earlier, Blockchain Validator might become a sought after job. Artificial Intelligence (AI), or computers that can learn themselves are changing the complexion of employment by becoming capable of doing tasks that could earlier be done only by a human, for example, driving a car in traffic.
AI is disrupting jobs all across the skills ladder. According to a McKinsey study, this is happening because, be it a low skill or a high skill job, if you deconstruct the tasks needed to be done for the successful performance of a job, you will find that in almost every job there is a component that can be done better by AI. In such a scenario people who have the skills that complement AI will thrive. Think of a good surgeon who can become great by using AI-based Augmented Reality glasses that overlay useful information while the surgeon is performing surgery.
I think the Flipped Classroom model, where teachers curate online learning resources for delivering the didactic part of teaching and use precious classroom time on learning activities like discussions and debates that ensure students have a deeper understanding of the topic of study, can also be applied to the use of AI in jobs. Ability to use AI to ‘flip’ a job will distinguish a good professional from a great professional. For example, doctors could use IBM’s Watson AI machine for better and faster diagnostics and use the time they save in solving more complex cases.
You may not realise it but AI is already very much a part of your life. It lives in your email spam filters, it is present in your smartphones, it drives Apple’s Siri, Amazon’s recommendation system and soon it will impact many more areas. Many tasks being performed by humans today will be better done by AI machines and in the coming decades, this will change the nature of employment. Just like you may not know how your smartphone works but you know how to make the most of it since AI is becoming an integral part of your life, it is better you understand it and know how to complement it with your skills instead of competing against it.
AI is a computer that can learn by itself. It does so by using Machine Learning, which is different from traditional rule-based programming and uses computers’ ability to analyse the big amount of data and decipher patterns from it.
Pedro Domingos, the author of the book The Master Algorithm, explains that traditional programming involved inputting data and an algorithm (an algorithm is a detailed sequence of steps or operations that tell the computer what to do with the data), the computer then processes the data based on the algorithm and outputs the result. In Machine Learning, the data is input along with the output and the computer generates the algorithm.
For example, we may input thousands of x-ray images of lungs and tell the computer which x-ray images reveal cancer and which are normal. The computer takes both these inputs and teaches itself how to detect cancer. If it makes a mistake in its diagnosis this data is again fed back and the iterative process helps the computer enhance its cancer detection algorithm. Computers taking data and figuring out the algorithm themselves is called Machine Learning.
Domingos explains that there are different approaches to Machine Learning. Some computer scientists take inspiration from first order logic to derive the AI algorithm. In this method the data that is input are specific facts and the computer works out the general principle. This approach, for example, is used in drug discovery like finding a new drug for malaria. Another group of computer scientists takes inspiration from the way the brain and the neurons work and create Neural Networks to extract rules and patterns from a set of data. This AI approach, also called Deep Learning, for example, is used by Facebook’s Deep Face facial recognition system, which identifies human faces in digital images. It has a nine-layered neural network and used four million images uploaded by Facebook users to train itself.
In another AI approach, uncertainty decides the probability of different possible outcomes and based on actual outcomes this probability is updated and iteratively the system performance improves. This AI approach is used, for example, in Google’s self-driving cars, and in email spam filtering. Another AI approach takes inspiration from ‘reasoning by analogy’ and uses what is called the ‘nearest neighbour algorithm’. Recommendation Systems (‘if you liked this song you may also like…’) and Collaborative Filtering uses this approach.
We can distinguish between two types of AI.
Weak or Narrow AI: is AI that is good at doing only one task. In 1997, IBM’s Deep Blue computer beat the world chess champion, Gary Kasparov. In 2011, IBM’s Watson computer beat the human champions of the American television game show, Jeopardy. Earlier this year Google’s AlphaGo AI machine beat the world champion of Go, a really complex strategy board game. These and other areas like spam filtering and recommendations systems are all examples of Weak AI and can do only one task. IBM’s Deep Blue computer was very good at learning and improving its chess playing technique but it was not much good at anything else, not even playing another type of game.
Strong AI or Artificial General Intelligence: is, as yet, a hypothetical machine that can think, learn and perform any intellectual task that a human being can perform. Strong AI can improve its performance by itself using what is called ‘recursive self-improvement’. Natural Language Processing and Computer Vision are examples of strong AI.
Some computer scientists believe that sometime in the future, not certain when, there will be a moment of ‘singularity’ when AI will exceed human intelligence. We could also come to a point where AI machines will create even more intelligent machines themselves – what is described as Artificial Super Intelligence. Although when this will happen is not certain, many prominent people like Bill Gates, Elon Musk and Stephen Hawking are of the view that we need to put safeguards in place because the ‘maker’ (us humans) will no longer be in charge of such machines. Swedish philosopher, Nick Bostrom, believes that Artificial Super Intelligence poses ‘existential risk’ meaning such machines pose the danger of annihilating humans.
Whether in the long run Strong AI poses an existential threat or not, what is certain is that in the shorter term Weak AI itself is disrupting our socio-economic future. Some experts argue that AI will lead to mass unemployment (leading to massive social unrest) while other experts are of the opinion that adoption of AI will lead to the emergence of new jobs, like repairing robots.
We don’t know which of this prognosis will come true but one thing is certain – the skills, competencies and dispositions needed to flourish in the age of intelligent machines will be very different. Creativity, ideation, large-frame pattern recognition, ability to solve unstructured problems, fine dexterity, and complex communication, along with the ability to complement these skills with the use of AI, such that the human-machine alchemy allows you to do tasks that were not possible earlier, will greatly enhance your employability and entrepreneurship potential.
We don’t know which of this prognosis will come true but one thing is certain – the skills, competencies and dispositions needed to flourish in the age of intelligent machines will be very different. Creativity, ideation, large-frame pattern recognition, ability to solve unstructured problems, fine dexterity, and complex communication, along with the ability to complement these skills with the use of AI, such that the human-machine alchemy allows you to do tasks that were not possible earlier, will greatly enhance your employability and entrepreneurship potential.