´óÏó´«Ã½

15 June 2018

Artificial Intelligence (AI) is an increasingly large part of our lives and could be as world-changing as the 19th Century's Industrial Revolution. While the prospect of intelligent machines has spawned scare stories galore - in science fiction and real life - author James Scott told the Hay Festival some reasons to be optimistic, too. ALASDAIR MACRAE investigates.

It’s difficult to imagine education without human teachers, but AI could be a better substitute for the current model of learning from textbooks, scripts, and teaching plans – one that could be more easily and cheaply updated as the education needs of children change in a rapidly evolving world.

Some argue this wouldn’t just improve existing education systems, but could also allow a high quality of schooling to be administered at less expense, levelling the playing field for poverty-stricken parts of the world.

In addition, technology could change what we even think of as an education system. Education experts argue our current method of teaching a standard curriculum to all students and then testing them on it is an outdated practise. It is effectively a ‘talent sifting system’, but will be irrelevant in a world where menial work is more efficiently done by AI-driven machines and employers will place greater value on individual creativity and softer competencies like communication skills.

Instead, they argue that students could use AI-assisted technology to tailor their own education, getting better at what appeals to them rather than what the curriculum dictates they should know. This self-directed learning would change the role of teachers; rather than being a delivery system of information to students, they would leave that to technology and assume the role of coaches or mentors, closer to art, sports and design teachers who already assess students more subjectively than merely training them to pass a test.

The core promise of AI systems is that they will take care of trivial tasks and leave humans to focus on what AI is not capable of, and the field of journalism is taking advantage of this time-saver. The Washington Post has its own AI technology, , which has written hundreds of articles on the Olympics, political races, high school sports, and similar statistic-driven reports that traditionally would be tedious work for a human writer to churn out. The papers say that AI-assisted journalism allows reporters to focus on meatier stories that require more research and human decision-making.

But automatic writing systems are developing further, with various AI programs assimilating snippets from genre fiction and then generating original stories. The Princess and the Fox is the first fairy tale written by an AI, created by scanning the works of the Brothers Grimm, producing phrases that would then be pieced together into ‘the rough shape of a story’.

However, the day when machines can tell us a good story is still distant. While AI can generate content, it still requires the human eye to judge tone and personality, and sometimes even just to make sense. See the bizarre short film for what happens when actors perform an AI-generated screenplay with no human rewrites.

Scott’s book AIQ opens with the story of a cucumber farm in Japan that he feels is inspiring for the democratisation that AI tools can bring. The woman who owned this farm used to spend eight hours per day sorting cucumbers by hand, judging them by quality to sell to different markets.

Her son, an automobile engineer, wanted to relieve his mother from all these days of drudgery. He was able to download open source software from Google called that he’d never used before, and read the free documentation on how to use it. He then created cucumber-sorting software that took a picture of the cucumber and automatically sorted it into one of eight different classes, cutting out all that tedious work for his mother.

Additionally, cloud computing is becoming ever-more widespread which means that the small business owner or would-be data scientist doesn’t need to invest in the huge cost of servers required to make use of these algorithms. Instead, space can be affordably rented from anywhere in the world, allowing the productivity of an AI to be available for anyone who takes the time to learn, not just the corporate giants like Google and Apple.

AI analysis can help us overcome our unconscious biases – take Hollywood, for example. Oscar-winning actor Geena Davis (pictured, right) set up her own to get some solid data on the under-representation of women in American movies.

Using technology from Google to face-scan the highest-grossing films from 2014-2016, they collected information on how often males and females appeared and spoke on screen. , men were seen and heard nearly twice as often as women.

This raw data can expose blind spots in media. A survey of filmmakers found that 68% reconfigured their projects after seeing these findings and 41% stated it had impacted one or more of their movies. “We'll see some major improvement in both the number and quality of female characters within the next five to ten years,” Davis predicts. Of course, there is no reason this technology couldn’t be used for race, age, or any under-represented group.

Above image © 2018 Geena Davis Institute on Gender in Media.

There are many examples of computers defeating humans at various games, such as Deep Blue famously beating world champion Garry Kasparov at chess in 1997. However, the true power of these AIs was made evident by a different game.

Go (pictured) is a 3,000-year-old Chinese board game that is very simple to learn, but features incredible complexity, far more than chess. It has been described as the most elegant game ever created. In parts of Asia, children who show promise are taken out of school to attend special ‘Go schools’, where they train for 12 hours a day, seven days a week, in order to become champions.

When Google’s AlphaGo AI defeated the world champion Go players, it surprised them by playing in ways that had never been seen before. In 3,000 years of the game, no human had come up with the strategies this AI had. Now, the system is being used to train human players, opening their minds about moves they had never even considered.

Of course, there’s little value in AIs being good at games in and of itself, but the way that an algorithm can learn and test thousands of strategies in a short space of time is fascinating as it effectively speeds up our own learning. This could be applied to many different fields to find solutions we would take hundreds of years to come up with.

AIQ: How Artificial Intelligence Works and How We Can Harness Its Power for a Better World, by James Scott and Nick Polson, is published by Bantam Press.

All images courtesy of Alamy unless specified.