Human-like robots, which have the ability to accrue abilities, skills and knowledge the same way humans do, will in future be able to perform and take over every task that can be done by human beings.
This is the word from Dr Suzanne Gildert, SingularityU faculty member in robotics and artificial intelligence (AI), and co-founder of Sanctuary AI, speaking at SingularityU South Africa Summit 2019, held in Johannesburg last week.
Sanctuary AI is a robotics company on a mission to create ultra-human-like robots called “synths” that will be indistinguishable from human beings.
Gildert explained that with current advancements in robotics, AI, machine learning and software, the contemporary human workforce will in the distant future be completely displaced by automation.
Without providing a time frame, she explained that in order for this to happen, four important technologies will combine, resulting in the creation of truly autonomous robots, which not only have the ability to perform any action or task accordingly, but also have the cognitive ability to make their own decisions.
These technologies are: robotics, tele-operation or remote operation, narrow AI and digital anthologies.
While research shows that in future, human-machine collaborations are expected to shape the future workplace, Gildert believes this is a short-sighted view of what can be expected.
“Could all work be automated? Each time I ask this question I get two reactions: either that’s crazy, not all work can be automated; or the second reaction, which is... actually that’s possible, but also very terrifying,” she noted.
“Over the next time period, we will see human-like robots, which can acquire different abilities, skills and knowledge, and use them the same way humans do, gaining traction. These abilities will slowly improve over time and then eventually we’ll create better and better robots, which slowly walk through the linear curve. Before you know it, they will be able to do all skills and any job you can imagine.”
Discussing the automation trajectory, Gildert explained that as part of civilisation, humans have been automating things for centuries, from the early phase of agriculture where machinery, in combination with the power of animals, was traditionally used for the purpose of ploughing.
Today, automation is used to aid work across all industries, such as the automobile sector, where assembly robots are used in the manufacturing process of vehicles; software systems used to automate work processes; self-service and self-check-out systems used in retailers; self-driving vehicles and autonomous delivery robots, which are gaining ground in urban landscapes.
“These are just a few examples of how technology is slowly but surely automating pieces of work. When robots eventually take over all work, I don’t know if this will be a good thing or not, but it’s inevitable.”
Defining the nature of work
Before we understand how work can be automated, Gildert highlighted the importance of understanding the term “work” as an activity or function, involving mental or physical effort done in order to achieve a purpose or result.
“There are over 2 000 individual tasks or work activities that people do every day at work and there are three different categories of work – abilities, skills/tasks and knowledge-based work.
“So if we think about the automation of work according to this framework, what’s been happening in the past decades is that we’ve been automating work in the first two categories, where we take a specific task and automate it. What will happen in the future is that we will stop looking at only the first two categories and focus on the deeper level where we will start automating knowledge-based skills.”
Developing human-robots
In order for robots to learn human-like skills, abilities and acquire knowledge, robotics – a branch of engineering that involves the conception, design, manufacture and operation of robots – must combine with tele-operation, narrow AI and digital anthologies, she pointed out.
Robots use AI to act and learn skills in the physical world while tele-operation allows robots to learn certain tasks.
“Tele-operation is important because if you just load robots with AI, they can make a lot of mistakes, which can prove costly in the real world. It allows robots to learn tasks on the job while watching humans perform those tasks. Narrow AI, on the other hand, is a specific type of AI in which a technology outperforms humans in some very narrowly-defined task,” Gildert continued.
However, in order for all of this to come together, digital anthologies, a collection of information about an array of subject matters, must be included to provide robots with conceptual understanding, creating intelligent machines that are able to make decisions.
“This means in future we will have robots that are the world’s best scientists, which help solve global challenges, which help us with house chores, and even provide us with healthcare advice.
“Many people find the idea of all jobs being automated very scary, but we need to open our minds to the fact that it might not be that bad. Humans will focus more on what gives them meaning and purpose in the world, such as making a positive impact on society and living a self-sustaining lifestyle,” she concluded.
Share