Jane Zavalishina: Machine learning’s ability to replace human decision-making by analysing data to determine the best appropriate next step for routine, repetitive tasks means that it is ideally placed to lead the next generation of automation. Using data to determine the best course of action within a set of finite rules is one thing, but empowering technology to draw its own conclusions, by learning from empirical data – is quite another. This is where machine learning offers a competitive advantage through decision-making and automating intellectual tasks by providing greater precision compared to rule-based systems and human analytics.
To give an example – let us look at maintenance efforts that are performed for any equipment or machinery. Typically, the decision to execute maintenance or replace spare parts is guided by a pre-defined schedule and a set of conditional rules. Such rules are often guided by previous statistics of faults or breakdowns, but they are widely generic and tend to be suboptimal in terms of costs incurred. In fact, maintenance is always a balancing act, as both excessive support efforts and potential breakdowns and delays mean potential losses.
Equipment maintenance is a good example of a process that can be automated with the help of machine learning. Instead of simply following a predefined schedule or responding to alerts, a machine learning-based predictive model can be built. This complex model can precisely predict potential breakdowns before they happen or before they are detected through human analysis. This is achieved through the analysis of past maintenance data, performance and machine telemetry. This also allows for the ordering of spare parts to be done automatically. Eventually, this reduces costs and automatically streamlines servicing efforts.
When it comes to machine learning, it’s important to remember that the more relevant the data fed into decision-making, the more accurate and appropriate the decision made. But equally, the bigger the amount of data, the greater the need for machine learning and automation due to the sheer complexity of the calculations and impossibility to properly leverage the data without the help of machines. That is what we offer at Yandex Data Factory – the development of such machine learning models that help companies automate routine decision-making to achieve measurable business results.
Companies are investing in machine learning not because it’s a fad or because it makes them seem pioneering. They invest in it because they are seeing the opportunity for rapid positive return on investment. Being able to predict success more accurately or avoid potential losses preemptively are game changing benefits and something companies will always strive for.
What in the long run do you think will be a challenge for AI developers? What is the intellectual quality that humans have that robotics/automation/AI will not now or possibly ever be able to emulate?
The greatest challenge for AI is for it to explain its actions – why it has done what it has. While machine learning is capable of automating many intellectual tasks, AI developers are faced with the challenge of replicating the human brain and rationale. No small feat.
The main challenge lies in establishing trust towards AI – as it is often impossible to rationalise machine learning in the same manner we rationalise and justify our actions – simply due to complexity of decision-making and a vast number of factors taken into account. When it is impossible to understand the “thought process” behind the machines, the only way to build the trust is through experiments demonstrating and measuring the direct value brought by the smart algorithms. Establishing this experimental culture in business practice will lead to greater trust and, as a result, a greater division of labour, and increasing readiness to accept machines without fear, doubt or scepticism from humans.
However, there is one thing machines still cannot do: creativity and defining the best actions in the absence of past historical data. This strategy guidance and bold risk-taking is still reserved for humans, and will continue to be for a while yet.
What are the economic implications for AI – good or bad?
Data compiled by Deloittte from the census data for England and Wales, stretching back to 1871, has shown that technology has created more jobs than job losses. Nonetheless, some have come to fear AI due to a belief that as automation becomes more sophisticated, the need for human input will decrease and jobs will be lost.
In fact, while robots may replace some roles, this shift to relying on technology will enable people to focus more on creative and strategic tasks, changing the way they approach various processes and even changing industries altogether. The first step to this is for humans to assess which tasks machines will take on and which strategic and creative tasks are to be dedicated exclusively to humans.
This shift in the way humans and organisations work together with machines will ultimately see the economy boosted by increased efficiency. As tasks become automated and efficiency increases, we’ll see cost reductions on a large scale, improving entire industries and propelling the economy forward.
On the bright side, automation is likely to improve the quality of life, especially for low-income households. Many things will become more affordable, and personalisation, facilitated by AI across various industries, will become a commodity.
Will humans ever become slaves to the machines? Could the growth of AI have a negative impact on society?
AI and automation will continue to evolve and its sophistication will heighten but far from becoming slaves to machines, humans will be freed by them. With the onset of AI and automation, we will see machines take over the mundane which will leave humans with more time and energy to explore different goals.
Moreover, as costs are reduced, businesses will have more money to spend on things which may have otherwise been compromised in favour of profitability – for example, corporate social responsibility or ethical activities.
If humans want to make the most of AI, we must ensure that we learn how to use these new technologies to their full potential so that they support our transition into strategic beings.