No job too small: how machine learning could change business

By on
No job too small: how machine learning could change business

We explain how artificial intelligence predicts the future and how it could change the workplace and even create jobs.

While driverless cars and robotics may hog the headlines, artificial intelligence, deep learning and similar technologies may make their biggest impact almost invisibly – by streamlining and advancing everyday life and business in countless ways.

In the words of Alex White, Nvidia’s vice president of enterprise, deep learning heralds a “new era of computing”, one that could ignite a “powerful and profound change [...] for everyone and everything.” At least if Nvidia gets its way.

Nvidia is a global leader in visual computing, wielding an army of 8,000 data software engineers. The company is developing more complex graphics processing units to improve machine learning, while at the same time supporting 1,500 startups in a bid to see where machine learning can take the human race. Not only in headline-grabbing areas such as cancer treatment, robotics and driverless cars, but also in the day-to-day running of businesses.

Sure, it’s mundane in subject matter, but that’s the whole point. Deep learning could revolutionise the way people go about their day-to-day jobs with regard to data processing, freeing them up for more fulfilling, creative and ultimately fruitful endeavours. In other words, deep learning does the dull stuff, so you don’t have to. 

AI as a job creator?

If you work in marketing and sales, for example, deep learning could be used to track customer interaction with your brands in real-time across all social media outlets. It could free you up from arduous data-collating legwork to pursue more worthwhile objectives, such as building on the strengths of what works where.

In this light, deep learning could constitute a means of bolstering employment levels; when businesses improve, they develop and expand, creating more jobs for us lowly humans to fill. At least, that's the theory.

We have every right to be sceptical, but there's some logic behind the theory. For example, with ecommerce sales reaching US$2 trillion last year – a figure that’s expected to double by 2020 – sales and marketing constitutes a vast swathe of the job market that could be refined by deep-learning infrastructure.

White was speaking at Nvidia’s Deep Learning Institute event in London recently. Another standout speaker was Dr Anthony Morse, an academic who, in his own words, “makes machines that learn like children”.

In a video, Dr Morse introduced a humanoid robot to an object – an orange star. Then he displayed the orange star alongside a red ball, which the robot had not yet been introduced to. When asked to pick up the red ball, the robot was able to infer from previous instruction which of the two objects was the orange star, and thus to deduce which object was the red ball (the only other object left on the table).

This relatively simple task veiled a world of nuance and sophistication, the point of the exercise being that this humanoid robot was teaching itself things in the way a small child might have the capacity to do. Now that this technology has taken flight, the possibilities are incredibly far-reaching.

How do these possibilities translate into real life for us mere mortals? Morse says that visual image recognition is often cited as an example of machine learning as it’s easily understood by everyone, not to mention the fact that deep learning easily outperforms humans in visual recognition.

Anyone familiar with Facebook’s auto-tagging tool will recognise the usefulness of this, both for users and advertisers, although deep learning’s capabilities extend well beyond social network photo-sorting. Morse identified deep learning’s capabilities as broad and twofold: first, to enhance existing automation; and second, to enable new production lines and markets.

For the former category, he cited the advancement of factory production lines that, he argued, could evolve to become far more complex systems churning out newer, more developed products if they were replaced with deep learning. Enabling new products and markets was where things got very glamorous.

From speech recognition and real-time translation to autonomous vehicles, from early-stage cancer detection to a curious software Morse described that could take a picture of someone’s clothes and buy them immediately online – there was something for to suit everyone’s needs and/or bizarre idiosyncrasies.

Limitless possibilities?

So far-reaching, yes. But infinite? Not quite, according to Morse.

“It can’t solve everything under the sun,” he stated, in a sentiment that jars with a lot of what one hears regarding the expansive potential of deep learning. Echoing sentiments from another speaker, Dr Adam Grzywaczewski, Morse attributed the recent explosion of deep learning to three developments.

First, the increased availability of big data. People are uploading 100 hours of video to YouTube every minute; there are 350 million images a day uploaded to Facebook.

Second, data on this scale has given rise to research and development of new deep-learning techniques, and the third development is the leap forward in the computational horsepower necessary to bring these techniques to fruition.

But with deep learning comes deep limitations. Morse lamented that labelled training data (examples for AI to learn from, sense-checked by humans) is usually expensive (tough, when you work in academia, as was aptly pointed out by one delegate at the event) or it’s non-existing.

Then again, you need to weigh up this caveat against deep learning’s potential ability. Understanding how children learn and applying it to machine learning is no longer fodder for science-fiction movies or Pixar films. It’s happening now, it’s happening quickly, and it’s happening on a business level.

Dr Grzywaczewski was keen to stress that the deep-learning ecosystem is dynamic – it’s constantly changing. And in the words of Nvidia’s Alex White, while involvement with machine learning has been largely confined to the world of academia rather than business, “that wave is coming to shore pretty quickly”.

Next: how AI predicts elections and horse races.

How AI predicts elections and horse races

The US election result came as a surprise to many – but not MogIA. The predictive artificial intelligence system out-figured most professional pollsters, calling Donald Trump’s electoral success at the beginning of November.

A rival system called UNU picked Hillary Clinton to win the popular vote, which she duly did, and it correctly predicted another notoriously difficult to call American race, the Kentucky Derby. Both systems work by listening to people.

MogIA scans public social networks, while UNU surveys respondents, taking that human opinion and “amplifying” it into a system called “swarm intelligence”.  

AI using neural networks seeks to mimic the brain’s network of cells, while “swarms” gather intelligence via a network of people explains Louis Rosenberg, founder of UNU-creator Unanimous AI.

“It’s an emergent intelligence, moderated by computer software, but leveraging the wisdom, insights, and intuition of human agents,” he said. “And it turns out that these artificially intelligent ‘human swarms’ are very smart. They outperform individuals. They outperform experts. And they outperform simple methods of tapping the intelligence of groups, like polls and markets.”

It’s more than averaging human selections, as existing polls do. With the Kentucky Derby, UNU built a swarm of 20 people. If the system had simply gone with the most popular picks, it would have picked only one of the top four finishers.

“We amplify the intelligence of individuals into a unified emergent intellect that exceeds natural capacity,” he said.

Popping filter bubbles

Dr Chris Brauer, director of innovation at Goldsmiths, University of London, explained that AI is essentially maths in action. “Weather forecasts improved dramatically with investment in weather stations, satellites, data, and processing power,” he said.

“Similarly, forecasting will improve in any complex socio-technical system such as political elections when the system continually improves and looks out for new data sources and forms – and has the power to process and analyse these inputs in meaningful ways. Next to this, polling looks like a man standing in front of a map pointing at clouds with a stick.”

Humans could perform better, but our hopes and assumptions get in the way, he noted.

“One benefit of these emerging AI prediction platforms is their ability to burst through filter bubbles,” said Brauer. “This may seem paradoxical, since many are blaming smart algorithms for creating filter bubbles in social media. But when it comes to predictions, too often our models accept and embed flawed assumptions. AI systems are better than human beings at dealing with what we’d normally term unknowns.”

Smart swarm?

Just how smart are these predictive systems? Rosenberg said his swarm intelligence outperforms between 90% and 99% of the individuals in the swarm.

“We also compare swarms to standard polls, and outperform them in almost all situations,” he said, pointing out that most polls have more participants than a swarm. But there are limitations.

“Our system performs best when asked complex problems that have many possible outcomes,” Rosenberg said. “For simple questions that have only two possible outcomes, our performance is good, but becomes more similar to a poll.”

And there are downsides to looking to machines for answers without understanding their limitations.

“These predictive AI systems aren’t foolproof or objective by design,” Brauer said. “In many ways, the frailties of algorithmic bias are more threatening than human biases since there’s a presumption of innocence.”

Trust in human authorities – such as governments and journalists – has long been falling, but that doesn’t mean machine-learning algorithms are entirely benign or lack bias.

“When we encourage learning systems to adapt autonomously in the wilds of data and human behaviour, we invite machine discretion and ethics,” Brauer said. “One day a machine is predicting the winner of an election; the next day that victor is turning to the machine for advice.”

Betting on AI

Aside from picking horse races and beating pollsters, UNU and swarm intelligence have been tapped for market research and consumer forecasting, such as reactions to movie trailers, as well as examining financial markets and medical diagnoses. Will such AI put the bookies out of business?

“I doubt it,” said Rosenberg. “If swarm intelligence becomes widespread, it will simply change the odds, since it means the gambling public has become smarter.”

This feature includes articles on machine learning in everyday business and how AI predicts elections and horse races that originally appeared at alphr.com.

Multi page
Copyright © Alphr, Dennis Publishing
Tags:

Most Read Articles

Poll

What would you like to see more of on BiT?
News
Reviews
Features
How To's
Lollies
Photo Galleries
Videos
Opinion
View poll archive

Log In

Email:
Password:
  |  Forgot your password?