Kathy Gibson is at SingularityU Summit in Kyalami – Humans harbour a deep-seated fear that artificial intelligence (AI) will negatively impact our lives – to the extent that some people believe it will result in our extinction.

Alex Rubsaam, faculty member in post-human intelligence at Singularity University, believes that the people subscribing to these ideas have a very linear idea of intelligence.

However, Rubsaam points out that visions of technology reflect the time and space in which they originate.

The more interesting question, she asks, is what is the humanity that people believe is under threat?

“So much of what we as humans do – our behaviour and connections – are being represented digitally in ways that AI systems can interact with. We have to think about what this says about our humanity.”

She adds that the metaphor of technology of the day overtaking humans in intelligence or physical exploits has cropped up throughout history.

AI will only pose a threat to humanity if we define our humanity as computational, Rubsaam adds.

As a civilisation we have been automating things for centuries – from using the strength of animals in agriculture to autonomous machines today.

This begs the question of whether all work can be automated, says Suzanne Gildert, Sanctuary AI co-founder and a faculty member of Singularity university.

Work can be defined as occupations, narrowed down to tasks. “There are about 2 000 tasks that people do over and over in work.”

Workers are defined according to abilities, skills and knowledge.

These categories constitute about 120 “work primitives”.

In the past we have been automating work in terms of tasks, with some looking at occupations, Gildert explains. As we move forward, we will start automating work primitives instead.

Gildert is working on synthetic humans (or synths), human like robots that have the potential to do things that humans can do.

Among the technologies enabling synths is robotics, giving machines the ability to act and learn in the physical world.

“Robotics is getting easier,” she adds. “We are now able to 3D-print robots, which is allowing it to progress quickly.”

Tele-operation is another enabler. “If you just send you robot out there, they could make mistakes,” Gildert says.

We can now deploy multiple robots and teleop capability across the Internet, and robots can learn on the job, she adds.

Machine learning is not new in the narrow AI world, where robotics is used to collect big data.

Those data streams from human-like robots can be collated and used to make the algorithms better, quicker.

What’s missing in the equation, Gildert points out, is the understanding of natural language.

Even when the computer beat people at Jeopardy, it didn’t understand a single word of what it was asked or answered.

“I think physical robots will be able to develop a proper understanding of words,” Glibert says.

Ontologies are technologies that can make the connections between the robot, the AI and the physical world – and will unlock a wealth of information.

Because exponential growth is embedded in computing, Gildert believes these changes will come quickly.

“I think of automation as a form of abundance – the robots are giving us extra labour, intellectual abilities, knowledge processing abilities. So we have to find a way to use that abundant resource wisely.

“If we can do this properly this abundance of capability will be great for our civilisation.”