How AI can come close to human problem-solving behavior

“I always find pick-ups near construction sites, where customers wait for the cab they booked through a mobile app but never showed up. [It’s] better than waiting in a long queue to pick up customers at Changi,” remarked the savvy cab driver who fetched me from a friend’s place last night.

He was right about the location, down beside a construction site for the new MRT line in Singapore.

It was fascinating what a bit of friendly banter uncovered. He was happy to reveal that he made higher net margins simply by avoiding commissions paid to taxi apps and leveraging instead his innate local understanding to find a steady stream of waiting customers near construction sites.

He’d probably just call it street smarts, but what he demonstrated was the ability to capitalize on a vulnerability within AI booking systems – an “edge case” that came naturally to him without explicit training or data analysis. He was hands down beating the “smartness” of AI algorithms used by modern taxi apps.

Perhaps, the cabbie was onto something: AI’s smartness can’t beat human street smarts.

The new frontier of tech-driven revolution

AI and machine learning (ML) are clearly in vogue. It’s hard not to fall under the spell of this techno-based fascination. I never fail to be amazed by the new developments in driverless vehicles, the great victory of Deep Mind against the world’s best chess player, or the global sensation that is Pepper Robot and its complex emotional analysis.

Is this the new stage of evolution?

The fundamental principles of AI and ML are surprisingly simple. It’s about analyzing data, identifying patterns, making informed predictions, then continually validating and iterating to increase the accuracy and speed of the next set of predictions.

With every passing day, computing speeds get faster, data sets larger, and algorithms more efficient. This exponentially enhances a system’s ability to learn and predict, ultimately mimicking human intelligence or, dare we say, even beating it.

Organizations around the world are already weaving in AI and ML, as well as other evolving technologies – cloud computing, personalization, internet of things, neural sciences, and robotics – into the very fabric of their DNA. Even in our own backyard – Southeast Asia, that is – you’ll find an ever-growing cohort of firms rapidly adapting to this new AI trend.

Time for a regional AI and ML (r)evolution

Let’s look at Indonesian fashion innovator Sale Stock, an ecommerce company based in Jakarta. It engages the latest AI technology to mine and analyze data and customer behavior beyond human capability. This allows the company to identify popular designs and tailor production to optimize costs to a level far below industry peers.

Such AI innovation isn’t limited to consumer goods. In Thailand, Bumrungrad International Hospital has deployed IBM Watson for oncology, an AI construct developed to advance cancer care.

At the state-run Rajavithi Hospital, AI is used to analyze patients’ eye screen results to assess if they are at risk of vision loss, helping inform decisions on preventative treatment. The accuracy rate of 95 percent disease detection is 11 points higher than that demonstrated by opticians or eye doctors.

AI is optimizing agriculture, too. In Vietnam, a crop intelligence startup called Sero has encouraged rice growers to take pictures of sick crops and upload them online. AI programs analyze the photos to identify crop diseases with an increasingly high degree of accuracy. The next evolutionary step is for the system to send disease diagnoses and suggested treatment options via smartphone apps to farmers directly.

AI’s Achilles’ heel

With every passing day, it seems AI-enabled systems are unlocking better reasoning that progresses them towards mimicking human problem-solving behavior.

What’s the answer for AI?