How brand new science will manage the fourth industrial revolution

(Image: Diagram by Christoph Roser at AllAboutLean.com (CC BY-SA 4.0))

Diagrams explaining the fourth industrial revolution, like this one by Christoph Roser, are OK as far as they go. Apart from the term "cyber physical systems". Ugh. What they mean is that physical systems are becoming digital. Think of the Internet of Things (IoT) supercharged by artificial intelligence (AI).

But according to Distinguished Professor Genevieve Bell, these diagrams are missing something rather important: Humans and their social structures.

"Now for those of us who've come out of the social sciences and humanities, this is an excellent chart because of the work it does in tidying up history," Bell said in her lecture at the Trinity Long Room Hub at Trinity College Dublin in July.

"It doesn't help if what you want to think about was what else was going on. Each one of those technological transformations was also about profound shifts in cultural practice, social structure, social organisations, profoundly different ideas about citizenship, governance, regulation, ideas of civil and civic society."

Another problem with this simplistic view is the way the Industry 4.0 folks attach dates to this chart. Steam power and mechanisation in 1760-1820 or so. Mass production from maybe 1870, but the most famous chapter being Henry Ford's work in 1913. Then computers and automation started being used to manage manufacturing from 1950.

"That time scheme works really well if you're in the West. It doesn't hold if you're in China or India or Latin America or Africa, where most of those things happened in the 20th century, many of them since 1945," Bell said.

Bell wants to know what we can learn from those first three revolutions. She heads the 3A Institute at the Australian National University, which was launched in September 2017 and is working out how we should respond to, and perhaps even direct, the fourth revolution.

Take the steam engines of the first industrial revolution. They were built by blacksmiths and ironmongers, who knew what they needed to build the engines. But they didn't know how to shape the industries the engines could power, or how to house them, or about the safety systems they'd need. These and other problems generated the new applied science of engineering. The first school of engineering, the École Polytechnique, was established in Paris in 1794.

The large-scale factories and railway systems of the second industrial revolution needed massive amounts of money. Raising and managing that money literally led to capitalism, and concepts like common stock companies and futures trading. And the first business school with funding from industry.

Early in the computer revolution, the US government had a problem. Nearly all of its computers relied on proprietary software from companies like IBM and Honeywell. So it asked Stanford University mathematician George Forsythe to create an abstract language for all computers. Two years later, his team developed a thing called computer science, and issued a standard 10-page curriculum. An updated version is still used globally today.

"So, engineering, business, and computer science: Three completely different applied sciences, emerging from three completely different technical regimes, with different impulses," Bell said.

"Each starts out incredibly broad in terms of the ideas it draws on, rapidly narrows to a very clear set of theoretical tools and an idea about practice, then is scaled very quickly."

With this in mind, Bell said that the fourth industrial revolution needs its own applied science, so that's exactly what the 3A Institute is going to build -- as the website puts it, "a new applied science around the management of artificial intelligence, data, and technology and of their impact on humanity".

And the 3A Institute plans to do it by 2022.

Nine months into this grand project, it's identified five sets of questions that this new science needs to answer.

First is Autonomy. If autonomous systems are operating without prewritten rules, how do we stop them turning evil, as so many fictional robots do? How do different autonomous systems interact? How do we regulate those interactions? How do you secure those systems and make them safe? How do the rules change when the systems cross national boundaries?

Or, as Bell asked, "What will it mean to live in a world where objects act without reference to us? And how do we know what they're doing? And do we need to care?"

Second is Agency, which is really about the limits to an object's autonomy. With an autonomous vehicle, for example, does it have to stop at the border? If so, which border? Determined by whom? Under what circumstances?

"Does your car then have to be updated because of Brexit, and if so how would you do that?" Bell asked.

If autonomous vehicles are following rules, how are those rules litigated? Do the rules sit on the object, or somewhere else? If there's some network rule that gets vehicles off the road to let emergency vehicles through, who decides that and how? If you have multiple objects with different rule sets, how do they engage each other?

Third is Assurance, and as Bell explained, "sitting under it [is] a whole series of other words. Safety, security, risk, trust, liability, explicability, manageability."

Fourth is Metrics.

"The industrial revolution thus far has proceeded on the notion that the appropriate metric was an increase in productivity or efficiency. So machines did what humans couldn't, faster, without lunch breaks, relentlessly," Bell said.

Doing it over again, we might have done things differently, she said. We might have included environmental sustainability as a metric.

"What you measure is what you make, and so imagining that we put our metrics up at the front would be a really interesting way of thinking about this."

Metrics for fourth revolution systems might include safety, quality of decision-making, and quality of data collection.

Some AI techniques, including deep learning, are energy intensive. Around 10 percent of the world's energy already goes into running server farms. Maybe an energy efficiency metric would mean that some tasks would be done more efficiently by a human.

Fifth and finally are Interfaces. Our current systems for human-computer interaction (HCI) might not work well with autonomous systems.

"These are objects that you will live in, be moved around by, that may live in you, that may live around you and not care about you at all ... the way we choose to engage with those objects feels profoundly different to the way HCI has gotten us up until this moment in time," Bell said.

"What would it mean to [have] systems that were, I don't know, nurturing? Caring? The robots that didn't want to kill us, but wanted to look after us."

As with computer science before it, the 3A Institute is developing a curriculum for this as-yet-unnamed new science. The first draft will be tested on 10 graduate students in 2019.

Bell's speech in Dublin, titled "Managing the Machines", included much more detail than reported here. Versions are being presented around the planet, and videos are starting to appear. This writer highly recommends them.

Related Coverage

5 ways machine learning makes life harder for cybersecurity pros (TechRepublic)

While many companies are turning to machine learning tools to fight hackers, they may not be as helpful as they seem thanks to a talent shortage and a lack of transparency.

Google Next 2018: A deeper dive on AI and machine learning advances

Google Cloud announcements bring deep learning and big data analytics beyond data scientists, but enterprises will want more.

Moving fast without breaking data: Governance for managing risk in machine learning and beyond

How do you resolve the tension between the need to build and deploy accurate machine learning models fast, and the need to understand how those models work, what data they touch upon, and what are the implications? Immuta says data governance is the answer.

When malware turns artificial intelligence into a weapon (TechRepublic)

In the future, your face could become the trigger for the execution of malware.

Hiring kit: Android developer (Tech Pro Research)

Companies are increasingly dependent on mobile platforms to power their business operations and to enable a productive workforce - and that means hiring topnotch developers to build the apps they need.