Author: Sutapa Amornvivat, Ph.D.
Published in Bangkok Post newspaper / In Ponderland column 26 July 2017
On Friday, Bangkok will host TechSauce Global Summit, a Southeast Asian technology conference gathering 6,000 attendees to discuss the global technological trends. An invitation was extended to me to share my experiences regarding the use of machine learning and Artificial Intelligence (AI) in businesses.
Before we jump ahead, let's visit the most important question: what exactly is machine learning?
Today, anyone can amass a huge amount of information at an increasingly cheaper cost. Machine learning is a branch of AI that deals particularly with data by letting computers learn or find patterns that can be useful for us. The science behind the scene integrates various advanced tools from statistics and computer science, super-powered by powerful processors and parallel computing techniques that distributes work across networks.
Machine learning can help firms answer massively complex questions with greater speed and accuracy. Commercialized applications range from loan pricing models, fraud detection to chatbots that answer customer's FAQs. There are also platforms developed by companies like LawGeex and Ravn replacing work previously done by junior lawyers, mining legal documents for evidence and reviewing contracts. This comes at a time when virtual assistants like Siri on iPhones and Google Home products are gaining popularity and mass appeal.
It's no wonder why there has been much hype and buzz around the emergence of AI and machine learning. But, can this be just another empty promise to the business world? What are realistic expectations we can have for this new technology?
To make machines work, tasks given must be well specified and often routine. For example, it would take us decades to individually examine through millions of bank transactions to detect fraud cases, while it could take machines much shorter time to get similar results. If given an individual case, though, professionals might be better at identifying what constitutes as a fraud. Unfortunately, humans are capped by both physical and mental limits, while machines do not experience fatigue like we do.
At this point, machines can only follow processes within a pre-determined framework, whereas human minds can go beyond that. Machines rely on past data for predictions and process improvement, in some way, hinging on the assumption that history always repeats itself. This is why machines can only do a poor job at predicting the unexpected, where human intuition can sometimes guide us further ahead. Humans are also good at innovating new ideas. Nevertheless, it's only a matter of time before an AI can build a better version of itself—in another word, evolution.
A huge amount of data input is also required for machines to perform their job efficiently and accurately. Without high-quality data, the system ceases to work properly, and the algorithms will not produce useful results. Moreover, there is still a limit to what can be captured as data. The system does not intuitively understand why we make decisions on its own, as there are countless factors that come to play into a human mind. While technological advances are helping to close the gap with computers being able to detect facial gestures or body language, the current technology is still far from truly understanding our emotions.
Notwithstanding overhyped sentiment in the business communities, there seems to be much fear and anxiety among the public towards AI. It certainly does not help when innovators like to illustrate the technological advancement by pitching machines against humans. This was seen when Google's AlphaGo, an AI program beats a world-champion in Go, long considered the most complicated board game. Recently, another robot was put to a skill contest against a golf champion.
The human vs robot matches often fuel eye-catching headlines of how robots are stealing our jobs and possibly taking over the world. Hollywood blockbusters like in the Terminator and I, Robot contribute partly to this inner fear.
A recent article by the New York Times adds another reason to fear the rise of machines. It discusses how AI and robots are being deployed to do malicious things like buying off Broadway tickets before people can or faking support in political campaigns. This is another case in point of bad press for machines, when the false here lies in human's abuse of technology or regulatory loophole.
Admittedly, my ability to foresee whether this calamity would happen in the distant future is limited. What we know for a fact today, though, is that robots are helping us improve productivity.
It would be dismissive if we outright pan the advancement of automation and AI as our existential threat. Rather, we should view machine and its ability to learn as merely a tool to improve our ways of life. It will be long before AI can replace human interactions and understand complex relationships. It can however improve upon some aspects of such interactions. In the world of business, AI may be used to improve on customer relationship perhaps with services personalized to each customer. But in the end, it is human that provides a personal touch, not the machine.
As the technology ushers in a new era of productivity, developing a realistic expectation of what machines can do and cannot do will have strong implications for businesses and policymakers alike. This requires an open debate on what proper regulation and appropriate business environment should look like. What skills do the new generation of children need now that having a great memory and fast calculation may become irrelevant?
The rise of AI is inevitable. And this fear warrants action, which manifests into the Partnership on AI—an alliance among big tech conglomerates to establish best practices for future research in this area.
Yet, fear should not induce resistance to adopt new technology for corporate process. Understanding the technology today will help business adapt without unnecessary fear or failure from over-expectation.