machine-learning
Human Resources

Will machine learning spell the end of programming?

This is a contributed piece by Guillaume Bouchard, Senior Scientist at the Xerox Research Centre Europe

Could you spend your working day teaching a machine to do its job better?

In a future of increased automation, with petabytes of data to be translated into meaningful information, the roles of human and machine will become more distinct in the workplace. A new role for humans is unveiling itself: improving machine learning systems.

With the assistance of humans, machine learning is likely to spell the end of computer programming as we know it. Today computer programmes are used to automate actions, but to change programme behaviour, you need a software engineer. With machine learning, the programme behaviour improves by itself when told by a human what’s right and what’s wrong.

Learning through the data surge

From the days of Turing, scientists have explored the notion that computers should be able to improve themselves over time. This was the birth of machine learning, which introduced us to the idea of artificial intelligence. Its popularity has since fluctuated but over the last few years we’ve seen the field once again become the scientific topic du jour.

The study of neural networks, from being an almost forgotten art, is now being revisited by the likes of Andrew Ng at Google, who recently developed large scale artificial neural networks using Google's distributed computer infrastructure. These networks were ‘trained’ to recognise higher level concepts such as cats, simply through ‘watching’ YouTube videos. In addition, the study of ‘deep learning’ algorithms, such as those that helped Facebook do tasks such as automatically tagging uploaded pictures for example, is back underway in some of the world’s leading laboratories.

One advance fuelling this surge is the growth of big data. As the reams of mostly disorganised information businesses deal with increases exponentially, humans recognise the potential for feeding this to machines that can then make sense of and act upon this data at speeds we can only dream of.

This will help businesses stay ahead of changing times and customer needs. The potential to increase efficiency and improve customer experience is overwhelming, and suggests the rise of machine learning is only just beginning.

Machine learning in today’s world

The practical benefits of machine learning are already clear, particularly within the field of customer care. Virtual agent technology, developed by WDS, a Xerox company, is now able to learn from online and live customer interactions to diagnose problems and propose solutions. Today however, machines lack semantic reasoning, and struggle to detect things like emotion in language. As this is essential to any good human interaction with a system, dialogue management will rely heavily on human training to make it possible to have a real conversation with a machine.

As this ‘learning’ continues, we expect to be able to put the machine on the front line of most calls. Human agents will act as ‘trainers’ or ‘teachers’ that give continuous feedback to improve the machine’s performance and will deal directly with customers on the more complicated problems currently requiring human expertise.

In all of these instances, for the trainer it’s like working with an utterly obedient pet. Your results are instantaneous, which means increased gratification and, in all likelihood, job satisfaction.

The role of the human

What’s clear from these examples is that the production of algorithms that drive our daily computer programmes is becoming more and more automated.

Machine learning is set to revolutionise, and democratise, traditional computer programming. Instead of asking programmers to make changes and updates, machines will be perfectly capable of doing it themselves as they see fit from what they have learned.

What are the implications of this? In my profession, I consider the outcomes of such change very positive.

Look at virtual agent technology in customer care. Consumers can now be connected to sophisticated helpdesks that watch and anticipate customer needs. These virtual agents are proactive, they know the best time to talk to you if a problem crops up, and can talk to you in a manner that best suits your personality and knowledge-level.

If machines take this understanding into programming, it would lead to full automation in this field, meaning we could simply ‘speak’ about our new ideas to the machine and have them implemented automatically.

Automated programming means the technical barriers to innovation will come tumbling down. It promises to democratise the creation of new software. But this does not replace the need for humans. It gives them the potential to achieve more.

Automated programs will still rely on data analytics, implemented by humans, to understand problems. With so much information at their “fingertips,” machines can be equipped to solve almost any problem – but it’s all about the information they are fed.

Developing virtual beings to serve everyday needs is part of a new economic model. Today’s successful business models are increasingly based on outsourced services which may or may not be provided by automated virtual agents.

What’s important is the impact you make on the individual you reach. More than ever, it’s all of us, not just computer scientists, who will be needed to help create technology that will successfully reach or even surpass our expectations.

PREVIOUS ARTICLE

« The OS that knows no name is at large across Africa

NEXT ARTICLE

My life as a tech teacher, part 9: Planning for phase 2 »
author_image
IDG Connect

IDG Connect tackles the tech stories that matter to you

  • Mail