Artificial Intelligence (AI), often confused as just ‘machine learning’ or – in extreme cases – ‘robots’, is something that extends beyond the confines of a concise definition. AI’s expansive goal of recreating rational human thinking has led to endless debates over its precise definition. This ambiguity, perhaps, is the leading cause of the fear that some grapple with as we relentlessly ponder the effects of Artificial Intelligence in the future and on our jobs.
What exactly is Artificial Intelligence?
In 1950, “Can machines think?” was the question posed by Alan Turing, an infamous English mathematician and computer scientist, which haunts the mission and vision of the field of Artificial Intelligence to this day.
At its core, Artificial Intelligence is a branch of computer science that tries to replicate, or even better, human intelligence in machines. It is the intimate collaboration of many technologies to enable machines to comprehend and act according to the inputs they receive. Machine learning, speech recognition, language generation, etc. are just the different types of technologies that are used to make artificially intelligent tools.
AI is divided into two categories:
- Narrow/Weak AI: Artificial Narrow Intelligence (ANI) focuses on doing one task really well. It is not trying to mimic human behaviour and intelligence but is trying to simulate human behaviour. Most current tools, like Google Translate, Siri, and various image processors are examples of weak AI. Siri, Apple’s virtual assistant, has a task to predict what the human is saying (its input) and then process it to return results. Hence, it is focused on one task while using other tools to supplement its total functioning, like Google in this case for processing. Though AI tools seem very sophisticated, they operate within a specific dataset. Consequently, when Siri is posed with abstract questions, the answer returned is vague and disconnected because this lies outside of the Siri’s programmed range of the dataset. Further, it requires human thinking. Even a self-driving car will be classified as a weak AI because it does not think like a human but instead tries to predict what a human will do by having several ANI tools in it.
- General/Strong AI: These are the machines that can truly replicate consciousness – the thing that sets humans apart from current machines. They are able to think reasonably and abstractly, strategize and make judgements under uncertainty, plan, learn, integrate prior knowledge in decision making and be imaginative, innovative and creative at the same time. This would truly make machines into the robots that we see in sci-fi movies today (for example, Vision from MCU) where they not only focus on one task but think faster and better than a human.
To answer Alan Turing’s question: currently, we are far from the stage where machines can think. As an idea, Artificial Intelligence has been around for decades but it is only now ANI tools are growing in quality and quantity and creating paradigm shifts in every sector of industry.
How does ANI affect businesses?
ANI impacts two main things: processes and people – which in turn influences the strategy of a business. Prediction, data, judgement and action are the core influencers of executing a task. In the world of business, the main focus of any strategy is to reduce risk and increase productivity. I will concentrate on AI’s influences on prediction, data, judgement and action which impacts strategy through people and processes.
Prediction
The book ‘Prediction Machines: The Simple Economics of Artificial Intelligence’ written by three business professors: Ajay Agarwal, Joshua Gans and Avi Goldfarb, breaks down the effect of artificial intelligence for businesses in a simple statement: it brings down the price of prediction.
Prediction is defined as ‘the process of filling in missing information’. It takes the information you have, often called data, and uses it to generate information you do not have. When the semiconductor revolution occurred it gave way to the formation of computers, which brought down the price of arithmetic. Arithmetic could be done faster and accurately which added to the efficiency of tasks. Further, spreadsheets were invented which caused the fall in the demand for jobs like accountants and book-keepers. However, in the long run, it led to arithmetic being used in places that one would have not expected: photography, data storage, art, etc. Similarly, these predictive tools will bring about the removal of monotonous, repetitive tasks in the day to day workings of various jobs and will lead to time for the higher rewarding tasks for employees like decision making. They increase the efficiency of workflow and increase productivity due to their high accuracy rates, making them more dependable than humans in their specific tasks. They were first used in obvious, basic prediction problems like loan defaults and insurance risk before using it in demand forecasting. However, since the drop in the price of predictive tools has become much more pronounced, we see them being used in wider applications such as medical diagnoses or autonomous vehicles.
The key to building new artificially intelligent tools is the ability to see any issue and reframe it into a prediction problem. This is called ‘AI Insight’. Engineers and computer scientists are now focusing on obtaining this skill, which goes beyond simple mathematics to become an art that requires critical thinking, vision and a rounded understanding of processes and technology. On a more basic level, this will impact business strategy by just making the processes faster and more accurate (exactly how will be explained later).
However, some AI tools will affect the economics of a business so dramatically that it will lead to a change in the overall business strategy itself. For example, in 2013, Amazon obtained a patent for ‘anticipatory shipping’, that will change their business model from shopping then shipping to shipping then shopping. With the help of data and machine learning, they will be able to predict what customers want and send it to them before the customer purchases it. This will help them to increase sales as it is a step forward from conventional marketing. However, at the moment, they do not yet have the accurate predictive capabilities to send the right products which will result in more losses due to the logistics for returns. Hence, they use this time to gather more data and refine their algorithms via deep machine learning. Simultaneosuly, some other unexpected ventures will be converted into prediction problems to increase their accuracy. Like Google Translate, where translation was governed by linguistic rules (where words were translated word to word) but now an artificially intelligent tool is being used that uses deep machine learning to refine their returns and better predict the correct translation.
There will be a greater focus for new skill sets like diversity and roundness in knowledge, which helps to improve judgement of any decision. Further, a managerial role that will likely develop is the ability to envision where they can deploy the right AI tool to increase productivity with given constraints of capital and skills of the workforce. In the future, we can see the need for a workforce that is flexible with working with technology.
Microeconomic theory teaches us that, with the decrease in the price of prediction and the increase in the demand and use of predictive tools, the value of its complements (goods that are consumed together) will increase. The complements that aid prediction in the decision-making process are: data, judgement and action.
Data
With the revolution in technology that has been thrust upon us due to the growth of artificial intelligence, data is referred to as ‘the new oil’. It is the core ingredient of good predictive tools. With AI there are three kinds of data: input data, training data and feedback data. Training data is used to make the algorithm, input data is put into the algorithm to get a prediction and feedback data is what is used to improve the prediction. All these data types are different and do not overlap, making data one of the most expensive investments to produce these predictive tools currently.
Over time, after the initial formation of the algorithm, through feedback and deep machine learning, the technology finds structures and regularities in data which helps increase its accuracy. The large amounts of data it processes for one specific task makes it hard for any human to compete. Coming back to the impact on strategy, it is important for managers to balance the cost of investment for data and the benefit of the increased accuracy of prediction. This is because, from a statistical standpoint, there comes a point when more data has diminishing marginal returns and it comes down to each business to understand to what extent they want their predictions to be accurate.
In a recent talk, Sid Paquette, the Head of Technology and Innovation at Royal Bank of Canada, has said that old banks have mountains of data – and now with recent technology, they have the opportunity to develop new ventures to improve their services and stay ahead of the game. He added that they see a change in the banking sector in the next ten years with the new services and improvements on traditional operating systems. This boosts the ability to produce new jobs. He concluded that what distinguishes AI is the ability to distil the complicated data of banks which enables them to better their decision making.
Lastly, where human judgement really matters is in situations where there is not much data and intuition supplemented with a theoretical understanding and experience helps to determine the best path. There may be a rise of ‘human prediction by exception’ where humans are depended on for rare events which machines cannot predict, such as the economic recession and effects caused by COVID-19. There were only a few investors, fund managers and economists that were able to predict the impact of the spread of the virus from the city of Wuhan to the world economy as it stands now.
Judgement and Action
This takes us to the final component: Judgement. Judgement is the process of determining the reward to a particular action in a particular environment. It is about working out the objective you’re pursuing.’ Prediction facilitates the decision-making process by reducing uncertainty but judgement assigns the value to those decisions. As the value of prediction decreases, the value of human judgement will increase.
Often automation and artificial intelligence are confused with one another. Automation is when a machine takes over an entire task from prediction, to judgement and action. Here is where our problem lies. Technology is growing at a rapid stage where there is automation by programming the judgement. This is seen in autonomous cars. Autonomous cars are coded in such a way that they receive input from various sensors, predict and judge what a human will do – like whether they would pause in front of a block or take a left – and carry out the action. This leaves the question – where will human judgement be used? In the long term, human judgement will be the most important in cases or rare events where there is sparse data available and also when there are subjective objectives of a business. This includes events like mergers, acquisitions, innovations which are complicated and where each case has its own unique features.
However, we are not yet at the stage where there can be complete automation. Cars like Tesla have had customers recalling their emergency brakes hitting automatically before even the car ahead of them slowing down by sensing a problem in front of that car. That automation had helped save the customer from a fatal accident. Still, there is a real need for a human driver. This is because currently, we are at a stage where human-machine collaboration is the most effective. In fields of law where machines are now used to redact sensitive information or in medicine where scans can be read by AI tools, professionals assist the machines in a way to reconfirm their work. This helps to reduce their workload while they focus on the more evaluative and non-repetitive part of their work. Further, companies tend to take a conservative approach to customer engagement technologies because of their immaturity. For example, Facebook found that its chatbots couldn’t answer 70% of customer requests without human intervention. As a result, it restricts bot-based interfaces to certain topic domains or conversation types while humans tackle the rest.
What is its implication?
This shows us that machines are well on their way to learning our judgement and with an increase in training and feedback data, they will be able to take over more of our repetitive and non-engaging work. Often, 20% of non-routine tasks create 80% of the total value, and it is these roles that will now be the future human hires. Like with the arrival of spreadsheets where the need for manual calculations decreased, there was a shift to hiring people with an understanding of technology and accounts to run various scenarios were required. Hence, it was not the end of the world in terms of mass unemployment but simply a shift in the necessary skill sets. So, to answer the question on everyone’s mind: there will still be jobs.
Some of the highest paying jobs currently like data analysts, doctors and lawyers have predictions at their core. With the price of prediction decreasing we can see the decrease in salaries due to competition. This could perhaps help to decrease inequality. At the same time, new skills will be required to be able to exercise judgement and it is the highly educated who will be able to benefit from this. Further, just like in the revolution of arithmetic, we can see an increase in the returns to capital instead of labour. Thus, we cannot say whether inequality will increase or decrease. Only time will tell.
Big countries and MNCs already have a headstart in the field of artificial intelligence due to their large capital reserves and superior ability to access data. This could give them the first mover advantage, but one cannot definitively say that they will be the only ones to succeed. In conclusion, we can say that AI will definitely transform every sector of society and the most important skills that will be demanded will be critical thinking, diversity of knowledge and a technical understanding of the new tools around us.
Cover Image Attribution: Mike MacKenzie via https://www.vpnsrus.com/