Machine Learning: Automation Versus Augmentation in 2020
The artificial intelligence (AI) renaissance is here, and the numbers don’t lie. According to Gartner, the business value generated by AI will reach $3.9 billion by 2022. With the Gartner Hype Cycle for AI suggesting over a decade of transformative potential ahead, market maturation still has much in store.
In many circles, anxiety is the first response to this robust expansion. Since the birth of AI decades ago, humanity has feared a self-inflicted demise at the hands of autonomous technology. However, the reality of our relationship with AI is far less apocalyptic, at least for the foreseeable future. As McKinsey points out, one-third of new jobs created in the US over the last 25 years didn’t exist before the advent of data-driven tech.
Contrary to prevailing fears, AI and, more specifically, machine learning (ML) present an opportunity to do things better - while still employing human help. This perspective reflects the concept of ML augmentation, and to many, the next phase of ML evolution. However, in contrast to the well-entrenched awareness of ML automation, augmentation garners far less attention. So, what does each of these approaches have to offer the broader ecosystem?
To answer this question, we’ll explore the underlying principles of machine learning automation and augmentation, their natural correlation, and how the broader ecosystem will decide which takes the lead moving forward.
What is Machine Learning Automation?
Since the inception of ML, automation has been the buzz word of choice. But what exactly is ML automation? In short, the automation of processes aims to replace human decision making and action using a combination of hardware and software. According to McKinsey, AI and the automation of activities could raise global productivity growth by 0.8% to 1.4% annually.
However, while autonomous or semi-autonomous solutions such as self-driving cars and auto-pilot airplanes have seen enormous investment over the last decade, they remain far from perfect. Unfortunately, we’ve seen the tragic consequences of their current limitations. From the pedestrian death caused by a self-driving Uber to the MAX8 airplane crash that might have resulted from “confused” AI, it’s apparent that obstacles remain to achieve fully-automated ML integrations.
However, while automation alone may fail to fulfill the requirements of select use cases, it also remains suitable for others. For instance, automation presents many benefits in the retail industry, especially when it comes to handling the monotonous tasks employees would rather forgo. This type of automation has also generated impressive results in the realm of energy conservation.
What is Machine Learning Augmentation?
In contrast to ML automation, the concept of ML augmentation proposes that technology be used to support and improve human behavior, both in decision making and executing tasks. According to Gartner, decision support and augmentation will generate the majority of AI market growth through 2025. In 2021 alone, global AI augmentation is forecast to create $2.9 trillion of business value and 6.2 billion hours of worker productivity.
Given the versatility of these human-assisted solutions, this trend is understandable. By integrating machine learning with human skills, strategic decision-making is subject to further optimization. This functionality enables the consideration of otherwise unexplored variables, dramatically improving outcomes.
For the foreseeable future, human intelligence will remain a crucial component of machine learning applications. Until the safety of fully-automated solutions is proven over time, human skills will bridge the gap. In summary, automation alone remains insufficient in its current form; human inputs must be employed to augment automated systems.