Why HR should care about machine learning


HR should understand machine learning. Not because it will change work, but because it already has.

Depending on what sci-fi stories you grew up with, machine learning will either herald a Jetson’s-like utopia where robot butlers scoot around on wheels and predict our every whim, or mankind will be rounded-up, enslaved and exterminated by its own creations.

It’s not difficult to trace where fear of the latter stems from. Machine learning has already played a part in disrupting and revolutionising some industries, in turn feeding our pre-existing fears of becoming disempowered, obsolete and ultimately bumped off the top of the food chain. But Arnold Schwarzenegger clones aren’t exactly coming for us wielding shotguns in the dead of night.

“It’s fear of the Terminator, right?” says Anton van den Hengel, director of the Australian Institute of Machine Learning. “That fear is motivated by a lack of information. And actually those fears are not really founded.”

Describing a cat

Machine learning is a relatively simple idea, insists van den Hengel. Let’s say you want to code some software with a set of rules so it can recognise pictures of a cat. Sounds easy enough, right? Not exactly.

Ask a human to describe a cat, says van den Hengel, and they’ll likely describe it along the lines of small and furry with four legs. But those exact same rules apply to dogs.

“If I showed you an image of a cat and an image of a dog, you will instantly know what the distinction is,” says van den Hengel.

“But you can’t describe how you know that – nobody can describe how they distinguish an image of a cat from a dog.”

Which explains why it’s impossible to code a computer to make the distinction.

But there is a solution, or one that’s very close to a solution. It involves humans guiding the machine to create algorithms – or rules – for itself. This process is known as machine learning.

“If you want to solve that problem, you give the computer 5000 images listed as cats and 5000 images listed as dogs,” explains van den Hengel.

“Then you ask the computer to figure out what the distinction is, and it will create algorithms so it can make the distinction for itself in the future.”

The result will have up to 98 per cent accuracy. If that doesn’t sound great, van den Hengel says that humans only achieve about a 95 per cent success rate from the same test.

“Most of the things a human can do in less than a second, machine learning is more accurate at,” he says.

“Even if you classify yourself as a doomsayer who fears the rise of machines, chances are you have already been participating in your own demise.”

Neural networks

Neural networks are the underpinning technology for machine learning. They help deal with large sets of data and are very much inspired by the human brain.

“They are interconnected layers of algorithms, called neurons, which feed data into each other, with the output of the preceding layer being the input of the subsequent layer,” says Nick Heath in a September 2018 article for TechRepublic.

One of the biggest benefits of a neural network over a traditional computer is its ability to do many things at once.

“With traditional computers, processing is sequential. One task, then the next, then the next, and so on,” says Eric Roberts, professor emeritus of computer science at Stanford University, on his university’s website.

With traditional computers, it may appear that many things are happening at once, but this is only an appearance. Another fundamental difference between traditional computers and artificial neural networks is the way in which they function.

“While computers function logically with a set of rules and calculations, artificial neural networks can function via images, pictures and concepts,” says Roberts.

“Traditional computers have to learn by rules, while artificial neural networks learn by example, by doing something and then learning from it.”

You’re already involved

Even if you classify yourself as a doomsayer who fears the rise of machines, chances are you have already been participating in your own demise. Ever tried to log into a website and had Google’s infuriating reCAPTCHA act as gatekeeper? Well, instead of the machine knowing all the answers, you’ve actually been an unwitting participant in training it.

You’ve been at it since 2009 and have helped Google Books digitise its entire archive, helped train Google Street View to read house numbers and shop signage, and you’re also playing a part in driverless cars one day being able to read street signs.

“Reading house numbers was a real challenge for Google,” says van den Hengel, “because house numbers are surprisingly difficult due to each one having a different font and different colour.”

Algorithmic HR

Machine learning is already playing a part in our day-to-day lives, but how can it impact HR? One key area, says van den Hengel, is efficiently shortlisting candidates. Machine learning can be used to identify key traits that your most successful employees all possess. It can then use that information to shortlist the most ideal candidates.

“The current process is that you put together a team that has all the skills you need,” says van den Hengel, “but the real opportunity is the ability to put together a team that has the personality types that you need to work together as well.”

Coaching co-buddy

Coaching is another area where machine learning will have an impact, predicts van den Hengel.

Say you’re an employee at a bank and a customer asks you a difficult question. In times gone by your options would be to wing it, wait until someone more senior is available to help out, or get back to the customer at a later date. But machine learning will allow you to engage with a digital coach for advice on the spot. Think Siri or Alexa, but for employees.

“There’s a whole lot of legal questions in banking, and people who work there struggle with the unbelievable complexity of the system,” says van den Hengel.

“So machine learning coaching resources can not only help them answer the technical question, but also figure out what the opportunity is.”

The same opportunity exists for customers. Chatbots have exploded onto the scene in recent years, and machine learning will only help them improve their customer service.

Opportunities knock

So should HR run for the hills, or should it face the machines head on? Van den Hengel says that while some people will likely see their jobs reassigned due to future machine learning efficiency gains, there’s an opportunity for most professionals to add more value and provide better outcomes.

“There will be winners and losers. But as far as I’m concerned, the winners will be the ones who can figure out how to get better outcomes using machine learning.”

This article originally appeared in the November 2018 edition of HRM magazine.


Identify meaningful measures of workforce performance and learn how to develop a sustainable measurement and reporting process, in this AHRI course, ‘Workforce Metrics’ that can be delivered in-house.

Leave a reply

avatar
500
  Subscribe to receive comments  
Notify me of
More on HRM

Why HR should care about machine learning


HR should understand machine learning. Not because it will change work, but because it already has.

Depending on what sci-fi stories you grew up with, machine learning will either herald a Jetson’s-like utopia where robot butlers scoot around on wheels and predict our every whim, or mankind will be rounded-up, enslaved and exterminated by its own creations.

It’s not difficult to trace where fear of the latter stems from. Machine learning has already played a part in disrupting and revolutionising some industries, in turn feeding our pre-existing fears of becoming disempowered, obsolete and ultimately bumped off the top of the food chain. But Arnold Schwarzenegger clones aren’t exactly coming for us wielding shotguns in the dead of night.

“It’s fear of the Terminator, right?” says Anton van den Hengel, director of the Australian Institute of Machine Learning. “That fear is motivated by a lack of information. And actually those fears are not really founded.”

Describing a cat

Machine learning is a relatively simple idea, insists van den Hengel. Let’s say you want to code some software with a set of rules so it can recognise pictures of a cat. Sounds easy enough, right? Not exactly.

Ask a human to describe a cat, says van den Hengel, and they’ll likely describe it along the lines of small and furry with four legs. But those exact same rules apply to dogs.

“If I showed you an image of a cat and an image of a dog, you will instantly know what the distinction is,” says van den Hengel.

“But you can’t describe how you know that – nobody can describe how they distinguish an image of a cat from a dog.”

Which explains why it’s impossible to code a computer to make the distinction.

But there is a solution, or one that’s very close to a solution. It involves humans guiding the machine to create algorithms – or rules – for itself. This process is known as machine learning.

“If you want to solve that problem, you give the computer 5000 images listed as cats and 5000 images listed as dogs,” explains van den Hengel.

“Then you ask the computer to figure out what the distinction is, and it will create algorithms so it can make the distinction for itself in the future.”

The result will have up to 98 per cent accuracy. If that doesn’t sound great, van den Hengel says that humans only achieve about a 95 per cent success rate from the same test.

“Most of the things a human can do in less than a second, machine learning is more accurate at,” he says.

“Even if you classify yourself as a doomsayer who fears the rise of machines, chances are you have already been participating in your own demise.”

Neural networks

Neural networks are the underpinning technology for machine learning. They help deal with large sets of data and are very much inspired by the human brain.

“They are interconnected layers of algorithms, called neurons, which feed data into each other, with the output of the preceding layer being the input of the subsequent layer,” says Nick Heath in a September 2018 article for TechRepublic.

One of the biggest benefits of a neural network over a traditional computer is its ability to do many things at once.

“With traditional computers, processing is sequential. One task, then the next, then the next, and so on,” says Eric Roberts, professor emeritus of computer science at Stanford University, on his university’s website.

With traditional computers, it may appear that many things are happening at once, but this is only an appearance. Another fundamental difference between traditional computers and artificial neural networks is the way in which they function.

“While computers function logically with a set of rules and calculations, artificial neural networks can function via images, pictures and concepts,” says Roberts.

“Traditional computers have to learn by rules, while artificial neural networks learn by example, by doing something and then learning from it.”

You’re already involved

Even if you classify yourself as a doomsayer who fears the rise of machines, chances are you have already been participating in your own demise. Ever tried to log into a website and had Google’s infuriating reCAPTCHA act as gatekeeper? Well, instead of the machine knowing all the answers, you’ve actually been an unwitting participant in training it.

You’ve been at it since 2009 and have helped Google Books digitise its entire archive, helped train Google Street View to read house numbers and shop signage, and you’re also playing a part in driverless cars one day being able to read street signs.

“Reading house numbers was a real challenge for Google,” says van den Hengel, “because house numbers are surprisingly difficult due to each one having a different font and different colour.”

Algorithmic HR

Machine learning is already playing a part in our day-to-day lives, but how can it impact HR? One key area, says van den Hengel, is efficiently shortlisting candidates. Machine learning can be used to identify key traits that your most successful employees all possess. It can then use that information to shortlist the most ideal candidates.

“The current process is that you put together a team that has all the skills you need,” says van den Hengel, “but the real opportunity is the ability to put together a team that has the personality types that you need to work together as well.”

Coaching co-buddy

Coaching is another area where machine learning will have an impact, predicts van den Hengel.

Say you’re an employee at a bank and a customer asks you a difficult question. In times gone by your options would be to wing it, wait until someone more senior is available to help out, or get back to the customer at a later date. But machine learning will allow you to engage with a digital coach for advice on the spot. Think Siri or Alexa, but for employees.

“There’s a whole lot of legal questions in banking, and people who work there struggle with the unbelievable complexity of the system,” says van den Hengel.

“So machine learning coaching resources can not only help them answer the technical question, but also figure out what the opportunity is.”

The same opportunity exists for customers. Chatbots have exploded onto the scene in recent years, and machine learning will only help them improve their customer service.

Opportunities knock

So should HR run for the hills, or should it face the machines head on? Van den Hengel says that while some people will likely see their jobs reassigned due to future machine learning efficiency gains, there’s an opportunity for most professionals to add more value and provide better outcomes.

“There will be winners and losers. But as far as I’m concerned, the winners will be the ones who can figure out how to get better outcomes using machine learning.”

This article originally appeared in the November 2018 edition of HRM magazine.


Identify meaningful measures of workforce performance and learn how to develop a sustainable measurement and reporting process, in this AHRI course, ‘Workforce Metrics’ that can be delivered in-house.

Leave a reply

avatar
500
  Subscribe to receive comments  
Notify me of
More on HRM