- AI techniques will be able to contribute up to US$5.8 trillion to the world economy
- 69% of current AI use cases involve improving already existing techniques.
ARTIFICIAL Intelligence techniques can contribute at least between US$3.5 trillion (RM13.9 trillion) and US$5.8 trillion (RM23 trillion) to the global economy according to Michael Gryseels (pic), McKinsey DIgital Labs APAC Leader. He was referring to a McKinsey Global Institute discussion paper titled "Notes From the AI Frontier".
The paper examined the potential impact of AI across 19 industries and nine business functions and found that this represents about 40% of the annual impact that could potentially be enabled by all analytical techniques.
"We found more than 500 use cases and they're applicable across industries so you know from travel to retail to banking to insurance through even heavy-duty industries," said Gryseels.
In fact, the report stresses that the greatest potential in AI is to improve on existing techniques to improve performance and create new insights. Applications such as credit scoring, insurance risk scoring and recommendation engines altogether account for 69% of the AI use cases in the study.
In comparison, only 16% were "greenfield" AI solutions that had no effective analytic method equivalent. Examples of these include "social listening" where a computer can detect the mood of a customer's voice at a call centre, and help advise agents on the best course of action for that mood.
The maturing of Artificial Intelligence
The hype on Artificial Intelligence is stronger than ever, and Gryseels says there are three good reasons for it. Firstly, the cost for computing power has dropped to such a point that even your phone is utilising AI to implement facial recognition.
Secondly, with more than 50 years of research in AI, techniques have improved to the point that machine-learning is now accepted as commonplace (i.e. that machines are programmed to train themselves).
Finally, the amount of data being collected is phenomenal - and if it's one thing that machine learning thrives on, it’s bountiful data.
"As a result of these advances, machines can now do things far in advance of what a human is capable," said Gryseels. "For example, for face recognition, a machine is now a hundred times more accurate."
Facial recognition is one of those tasks that is very difficult to have a human programme. But a computer can work on its own, fine-tuning its algorithm millions of times to land at one that works. What is crucial for this - or any other AI function - to work is that there is sufficient data.
"The rule of thumb is that for the amount of potential variables to consider, the amount of data needs to be at least ten or getting a hundred times (more)."
For example, if you want to create a model to assess credit scores, you will need about a hundred thousand customers that accurately reflect your customer base. If you want to build a driverless car, you need much, much more.
The democratisation of AI
All this is leading to the growing acceptance that "AI works", which in turn means that there now is a drive to offer these tools and techniques to organisations that previously may have thought they lacked technical expertise.
"The technologies are democratising fast," said Gryseels, with companies like Microsoft using practically the same phrase, and Apple also throwing their hat into this fray to offer developers the best shiny new toys with which to enhance productivity.
The demand for skills have also changed now that machine learning has become commonplace. "The focus has now shifted from humans programming the machines to humans providing the datasets and providing the labelling of the data sets," explained Gryseels.
"Before, they used to make the models. Now they prepare the data sets because the quality of the model depends on the data!"
Even though Gryseels admits that the vision of AI being simple enough to "plug and play" isn't quite there yet, but he considers technology to no longer be the stumbling block to companies using it. "It's actually applying the AI in a business process," he said.
Business improvements with AI are within reach
"I think for most companies the way to start is doing the things you do better," Gryseels stressed.
He gave pricing as an example. "Most SMEs pricing is human judgment," Gryseels pointed out. "This is typically area where our artificial intelligence can really help and have immediate impact."
Another area of potential is customer interaction. "For example, for companies having a service business," he said. "If the situations that we're getting in a call centre are predictable, 80% of the calls can be handled by an algorithm."
Gryseels was enthusiastic about who could take advantage of the power of AI. "Industries that have a lot of data, financial services, telcos, retailers - I mean all industries that have a lot of data."
"If you're not leveraging on AI I think at some point in time (you'll be) at a disadvantage," Gryseels concluded.
Brave new world – How AI will impact work and jobs
McKinsey: Deep learning AI based on neural networks can capture trillions in value
Companies embrace chatbots, but only a step at a time