Leaders in Change | Allies not rivals - the human machine interlace

Be Kaler Pilgrim

Be Kaler Pilgrim

.(JavaScript must be enabled to view this email address)


The terms "Ai" and "automation" are cropping up more and more – not only in digital circles but in wider society too, which is making a lot of people anxious as they question their role in this age of automation.

There are many thousands of articles on the subject that talk about the end of work, the dissolution of millions of jobs - including concepts like dark factories. And even Silicon Valley leaders have spoken about the impact of automation on the workforce, with Elon Musk recently saying that "what's going to happen is robots will be able to do everything better than us...I mean all of us".

And, of course, dystopian robotic futures are everywhere in popular culture, from Terminator to Westworld.

A fear of the machine isn't a new phenomenon - the Luddites were a famous example, and even back in the 16th Century, Elizabeth I refused a patent for a knitting machine fearing it would deprive her subjects of employment.

For our last Leaders in Change event of 2018, we invited Kevin McCullagh, founder of Plan to talk through the reality of the age of automation. Rather than fretting over how easily and soon humans will be replaced, he gave the case for thinking about the future of automation in a positive way. While work will need to be redesigned, interlacing machine strengths with those of humans has the potential to unlock economic growth and fuel innovation. An excellent topic to spark some Christmas cheer!

Increased prosperity

Kevin explained that automation tends to raise employment and prosperity. One example of this is the role of the paralegal. In the 1990s, electronic discovery software went mainstream. However, instead of seeing an end to the role of paralegals and other legal support staff, there has actually been an increase - faster than the labour force as a whole, as the role shifted to focus less on reading pages and pages of reports, and more on summarizing and sensemaking, and their roles became more meaningful.

Something similar happened when ATMs automated the tasks of bank tellers and when barcode scanners automated the work of cashiers: Rather than contributing to unemployment, the number of workers in these occupations grew over time, and those roles became more varied

In fact, in a study of job titles on the 1950 US Census, it was determined that just one out of 270 listed was eliminated due to automation - lift operator.

And it's important to remember that the job market is not, and has never been consistent, and ebbs and flows due to all kinds of factors - with automation being just one of these.

Kevin pointed to a McKinsey study on the impact on productivity growth from previous automation advances, which puts the steam engine, for example, adding 0.3% between 1850-1910, and early robotics adding 0.4% from 1993-2004. By their measure, automation should add between 0.8-1.4%. It's going to have a significant impact, but it's not the only factor to consider.

The right tool for the job

Kevin explained that most jobs are performed most effectively with a mix of human and machine strengths. For example, Toyota, which is a famously progressive manufacturer, uses robots for less than 8% of its production, for a couple of reasons:

  • Much of a car is made of floppy components, like leather and cables. While machines can be effective for parts of the process, humans are better suited to these aspects.
  • Toyota prides itself on its philosophy of continuous improvement. While a robot is excellent at carrying out repetitive tasks, it has no ability to improve the quality of its work.
  • It's expensive.

Another car manufacturer who has famously championed the power of robots, Tesla, has also come out to say that "excessive automation at Tesla was a mistake...humans are underrated".

Humans are underrated

We have seen some amazing strides in Ai - but it's important to remember that these are in very narrow fields, such as in the board game Go.

Many people misunderstand the link between these examples, and of Ai with 'human level intelligence', seeing this as a continuum. In reality, there is an absolutely enormous gulf between the two points.

The reason crossing this gulf will be such a challenge - and perhaps not ever possible, is illustrated with Moravec's paradox. Computational skills like chess, or solving a quadratic equation, are relatively easy to train a computer to do well, but it's difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.

For example, you can show a young child a couple of pictures of cats, and they will then be able to recognise other cats pretty quickly. They will also be able to connect with wider ideas - you could then read them a story about a cat, and they'll be able to relate the two. This is very difficult to engineer in a machine because the process is unconscious - "we know more than we can tell" in the words of Michael Polanyi.

In this sense, machines are pretty unintelligent - indeed the head of Ai at Apple, John Giannandrea talked about his role as making machines "slightly less dumb".

So the reality of an automated future is very unlikely to see machines reaching human levels of intelligence, rather that they might be "ceded authority far beyond their competence" - a concept known as overtrust.

The problem is that humans often underestimate their fellow humans.

Human-machine collaboration

Humans and machines have different strengths. Broadly, humans are suited to lower data situations, where other skills such as judgement, empathy, leadership, creativity and improvisation are required, while machines are good at data-driven and repetitive tasks. With this in mind, Kevin walked us through some examples of collaboration between humans and machines. 

  • Assigned - where certain tasks in a human workflow are outsourced. An example of this could be when stuck in a traffic jam, a person could enable a traffic jam mode on their car so they can get some work done, and retake control when traffic moves again.
  • Supervised - where decision-making processes are automated, but under a human eye, such as a pilot in a commercial plane.
  • Coexistent - an intelligent machine, such as a delivery robot, that shares the same space as humans, that is able to negotiate this space while carrying out its task.
  • Assistive - a system that helps humans perform tasks better. Jaguar Landrover has demonstrated an alert system for their cars that allows better visibility of driver blind spots, combining warning lights and a haptic touch in the seat to alert of obstructions.
  • Symbiotic - this is an emerging mode of collaboration that is interactive and reciprocal - such as Adobe Sensei.

Redesigning work

Many of these are still in their nascent phases of development - but while human and machine strengths can seem almost oppositional, there's a big middle ground that us humans need to become comfortable with:

  • Training - teaching machines how to perform tasks or act more human. This might include cleaning or tagging data, improving machine language learning, or defining and developing a brand or personality.
  • Explaining - untangling machine decisions and translating this to stakeholders.
  • Sustaining - ensuring the proper use of Ai, overcoming setbacks, managing stakeholders and maintaining momentum.

There are also some important ways that Ai can enhance the effectiveness of human activity and decision making. For example:

  • Amplifying - smart glasses can help engineers assess a complex environment more accurately.
  • Interacting - AI agents with voice interfaces can help facilitate interactions at scale.
  • Embodying - Ai combined with sensors to safely and effectively physically augment human workers.

Redesigning design

Ai can (and is already) playing a role in design. Kevin gave some great examples:

Yossarian - a discovery tool that uses Ai to identify new data patterns and connections.

Crimson Hexagon - an Ai-powered platform to uncover insights from existing consumer or user insight data.

Project Dreamcatcher - Ai creates designs based on specified constraints.

Aimybox - an Ai platform to accelerate and democratise prototyping.

Adobe Sensei - helps accelerate iteration and unlocks new creative possibilities.

Descript - automates transcription.

Nike - using computational design to reimagine high-performance running shoe design

Stitch Fix - uses a combination of Ai and human stylists to customise options for their customers.

Marcel - Publicis Groupe hit headlines for pulling its marketing and awards budget for a year to invest in Marcel, an Ai-powered communications platform.

I'd love to hear your thoughts on these products - and any others you'd add to this list.

Automation has the power to drive economic growth, fuel innovation and help make work more human, but it's important that we focus on working with machines, not against them. We need to champion human strengths, and invest in supporting and upskilling the workforce to be ready for the redesign of work.

If you'd like to talk more about how to redesign work, or find out more about future events we're running, please get in touch at be@wearefutureheads.co.uk, I'd love to hear from you.

About Kevin McCullagh

Kevin is the founder of Plan, the product strategy consultancy. Plan helps mobility and consumer tech companies explore the early stages of the product and service development. Their clients include: Ford, Toyota, Yamaha, Deutsche Telecom, Carl Zeiss, Microsoft, and Samsung.

Kevin writes, speaks, and chairs conferences on design, innovation and society; and has been published in: The Wall Street Journal, The Telegraph, FastCompany, Unherd, Icon, Blueprint and The Design Management Review.

You can connect with Kevin on LinkedIn and Twitter. Kevin also sat down with us to tell us more about his career story as part of our FFS series.