Artificial Intelligence (AI) stakes a claim on productivity, corporate dominance, and economic prosperity with Shakespearean drama. AI will change the way you work and spend your leisure time and puts a claim on your identity.First, an AI primer.
Let’s define intelligence, before we get onto the artificial kind. Intelligence is the ability to learn. Our senses absorb data about the world around us. We can take a few data points and make conceptual leaps. We see light, feel heat, and infer the notion of “summer.”
Our expressive abilities provide feedback, i.e., our data outputs. Intelligence is built on data. When children play, they engage in endless feedback loops through which they learn.
Computers too, are deemed intelligent if they can compute, conceptualise, see and speak. A particularly fruitful area of AI is getting machines to enjoy the same sensory experiences that we have. Machines can do this, but they require vast amounts of data. They do it by brute force, not cleverness. For example, they determine the image of a cat by breaking pixel data into little steps and repeat until done.
Key point: What we do and what machines do is not so different, but AI is more about data and repetition than it is about reasoning. Machines figure things out mathematically, not visually.
AI is a suite of technologies (machines and programs) that have predictive power, and some degree of autonomous learning.
AI consists of three building blocks:
An algorithm is a set of rules to be followed when solving a problem. The speed of the volume of data that can be fed into algorithms is more important than the “smartness” of algorithms.
Let’s examine these three parts of the AI process:
Fed into machine learning (ML) models
Applied to business applications
The raw ingredient of intelligence is data. Data is learning potential. AI is mostly about creating value through data. Data has become a core business value when insights can be extracted. The more you have, the more you can do. Companies with a Big Data mind-set don’t mind filtering through lots of low value data. The power is in the aggregation of data.
Building quality datasets for input is critical too, so human effort must first be spent obtaining, preparing and cleaning data. The computer does the calculations and provides the answers, or output.
Conceptually, Machine Learning (ML) is the ability to learn a task without being explicitly programmed to do so. ML encompasses algorithms and techniques that are used in classification, regression, clustering or anomaly detection.
ML relies on feedback loops. The data is used to make a model, and then test how well that model fits the data. The model is revised to make it fit the data better, and repeated until the model cannot be improved anymore. Algorithms can be trained with past data to find patterns and make predictions.
Key point: AI expands the set of tools that we have to gain a better grasp of finding trends or structure in data, and make predictions. Machines can scale way beyond human capacity when data is plentiful.
Prediction is the core purpose of ML. For example, banks want to predict fraudulent transactions. Telecoms want to predict churn. Retailers want to predict customer preferences. AI-enabled businesses make their data assets a strategic differentiator.
Prediction is not just about the future; it’s about filling in knowledge gaps and reducing uncertainty. Prediction lets us generalise, an essential form of intelligence. Prediction and intelligence are tied at the hip.Let’s examine the wider changes unfolding.
AI increases our productivity. The question is how we distribute the resources. If AI-enhanced production only requires a few people, what does that mean for income distribution? All the uncertainties are on how the productivity benefits will be distributed, not how large they will be.
AI is about trying to get control, but it’s an illusion. AI will always be limited by the execution capabilities of the organisation. Can business owners deliver appropriate data inputs to take advantage of model insights and be agile enough to handle the changes necessary to respond to AI signals?Over-reliance on big data may undervalue hunches or experience.
CEOs may suffer data bullying: feeling obliged to follow the advice of managers armed with data. But where the data is always historical and ignores anything new that could be disruptive, the CEO is likely to side with the data because the decision can be justified, even when “black box” ML models aren’t fully understood. (Deep Learning models, the new frontier of AI, are even more obscure because they mimic neural networks of the human brain.)
ML is already pervasive in the internet. Will the democratisation of access brought on by the internet continue to favour global monopolies? Unprecedented economic power rests in a few companies – you can guess which ones – with global reach. Can the power of channelling our collective intelligence continue to be held by these companies that are positioned to influence our private interests with their economic interests?
Nobody knows if AI will produce more wealth or economic precariousness. Absent various regulatory measures, it is inevitable that it will increase inequality and create new social gaps.Let’s examine the impact on everyone.
As with all technology advancements, there will be changes in employment: the number of people employed, the nature of jobs and the satisfaction we will derive from them. However, with AI all classes of labour are under threat, including management. Professions involving analysis and decision-making will become the providence of machines.
New positions will be created, but nobody really knows if new jobs will sufficiently replace former ones.
We will shift more to creative or empathetic pursuits. To the extent of income shortfall, should we be rewarded for contributing in our small ways to the collective intelligence? Universal basic income is one option, though it remains theoretical.
Our consumption of data (mobile phones, web-clicks, sensors) provides a digital trail that is fed into corporate and governmental computers. For governments, AI opens new doors to perform surveillance, predictive policing, and social shaming. For corporates, it’s not clear whether surveillance capitalism, the commercialisation of your personal data, will be personalised to you, or for you. Will it direct you where they want you to go, rather than where you want to go?
How will your data be a measure of you?
The interesting angle emerging is whether we will be hackable. That´s when the AI knows more about you than yourself. At that point you become completely influenceable because you can be made to think and to react as directed by governments and corporates.
We do need artificial forms of intelligence because our prediction abilities are limited, especially when handling big data and multiple variables. But for all its stunning accomplishments, AI remains very specific. Learning machines are circumscribed to very narrow areas of learning. The Deep Mind that wins systematically at Go can’t eat soup with a spoon or predict the next financial crises.
Filtering and personalisation engines have the potential to both accommodate and exploit our interests. The degree of change will be propelled, and restrained, by new regulatory priorities. The law always lags behind technology, so expect the slings and arrows of our outrageous fortune.