Written by JP Kloppers, CEO DataEQ.
The future is dystopian: a world in which we humble humans will be replaced by fleets of slick automatons – mechanical menials destined to not only solder, weld and glue us out of jobs, but account, diagnose, and translate us out, too. Or, so goes a certain line of argument.
Certainly, there have been some heavyweight concerns voiced about the rise of artificial intelligence. Among them, by no lesser figures than those of physicist Stephen Hawking, Tesla/Space X chief Elon Musk and Microsoft’s Bill Gates. Of course, there are counterarguments too. Just as the Industrial Revolution sparked fears around the supplanting of man by machine (fears which lead some as far as destroying the new mechanical marvels: hence today’s use of the word ‘Luddite’ to denote those opposed to technological progress), all new vistas are likely to provoke both optimism and hesitance.
Leaving aside more extreme visions of armies of self-replicating nanobots and a workless future, AI – and talk of the next big thing – is, seemingly, everywhere. Some of the existing tech remains impressive: consider Amazon’s bright orange fleet of order-fulfilment Kiva robots, or at-your-service virtual assistants such as Apple’s Siri, Amazon’s Alexa or Microsoft’s Cortana.
In still another AI arena lies the computing power behind robo-vacuums or driverless cars. Then there are the more Kubrick-esque machines: neural network-based technologies capable of so-called ‘deep-learning’. Examples include Google’s DeepMind (whose AlphaGo succeeded at beating world champion Lee Sedol at the ancient game of Go), along with IBM’s multi-talented Watson.
The exponential rate of development in the field of machine processing has led many to believe that the future may, in fact, be now. Indeed, for certain applications, computers do a pretty good job at replicating – and indeed, in some instances, bettering – their human counterparts’ abilities.
There are other areas, however, in which the skills of machines remain far from their apogee. Among these are the areas in which analysis of human cognition is required – namely, understanding the subtleties inherent in both how people communicate and how they feel. The fields of social listening and opinion data analytics – essentially, areas in which sentiment meets data – are chief among them.
Unlike the rules of a game – take chess for example – human communication cannot mapped by simply encoding for a limited number of possible interactions. Nor are the ways in which we communicate always predictable or direct. Being socially embedded, we express ourselves differently in different contexts; moreover, norms of interaction vary by both place and time. Add to this humour, irony, sarcasm, allusion and the modelling universe expands drastically.
In search of insight
Both language itself – and the emotions we express using it – are highly complex. As such, it is unsurprising that pure machine processing remains not fully up to the task of interpreting the contents of the online landscapes in which the two interact, such as Facebook and Twitter.
To this end, the issue of so-called ‘topic discovery’ is the first challenge – honing on what subject, theme, person or event that a given speaker is referring to. At a more granular level, the process also requires isolating relevant sub-topics. A broad theme such as ‘politics’ may, for example, have sub-topics such as ‘elections’, ‘foreign policy’ or ‘unemployment’ associated. Discussion surrounding an automotive manufacturer could be broken down by product feature – into ‘airbags’, for example, or ‘safety’ and ‘ride comfort’.
The next step requires evaluating sentiment – the feelings associated with the topics revealed. Many intelligence tools, however, fall down in this crucial respect. What is required, at core, is a level of insight that remains – for the time being – essentially human. Not only insight, but empathy too, along with familiarity with local languages and modes of expression, an appreciation of sarcasm and more.
People only, please
Not only do people remain better able than machines at recognising what is being referred to in a chunk of text, for example, but they’re also far better equipped to gauge a writer’s emotional state. By integrating actual human understanding with algorithmic processing through a crowd-sourcing platform, it is possible to unlock granular, sentiment-based insights.
Moreover, anyone – from students to retirees, to those living in remote locations – can be trained to work as part of a local crowd. In so doing, contributors add layers of accuracy to data that will ultimately be used by corporates, NGOs or governments: essentially any organisation needing reliable opinion based data insights.
Mind versus machine
Undeniably, AI continues to develop in sophistication. And certainly, for some processes – some of which are repetitive, and some of which demand high-level integrative ‘thinking’ – machines do a good, or better job than humans are able to.
Crucially, however, other aspects of intelligence are most nearly tied to what would likely be termed ‘insight’ – part of which requires the ability to inhabit another’s point of view. For the time being, and likely for long into the future, we humans will simply be far better at doing this than are our robotic counterparts. Certainly when it comes to the field of social listening and opinion data analytics, the human edge remains a necessary one.
This article was originally published on InsideBigData