AI Predictions for 2020
Driving business value out of Artificial Intelligence (AI), modeling at the edge, increased focus on data privacy and governance, and growing talent wars, among key AI trends
AI: everybody’s into it, but only a few do it well. The 2019 MIT SMR-BCG Artificial Intelligence Global Executive Study and Research Report states that nine out of ten enterprises today are invested in AI, but 70 percent have seen minimal impact so far.
2020 would be an active year on the AI front. As the phase of experimentation through 2018–19 starts becoming mature, adoptions will begin in a serious way and business leaders will start assessing the value of the investments made. Following are the relevant trends unfolding that the C-suite and business leaders should be expecting in 2020:
1. Business will get real operationalising AI and measuring the impact
In 2020, the C-suite will start assessing the value of their AI bets and prove the ROI to the business. Forrester’s AI prognostications state: “We believe 2020 will be the year when companies become laser-focused on AI value, leap out of experimentation mode, and ground themselves in reality to accelerate adoption”.
To achieve this, the first thing businesses should do is to understand the way to measure results. Data scientists are used to evaluating their “success” in terms of recall/ precision, AUC and similar “science” metrics, which are different from how businesses measure the effectiveness of a program. To bridge the gap, data scientists must sit down, a priori, with the business and work on a common framework with business to understand and measure the impact of their work on the business.
For example, at Airtel, one of the top telcos of the world, we know that for our campaign to “roamers” (people who use their local phone while traveling internationally), below are the relationship between Data Science, Business (Marketing in the case) and Finance metrics.
Once these “translation tables” are set up, it’s easy for all the relevant businesses and organizations to focus on the same goal and measure impact.
2. AI would be engineered for trust and fairness
As AI penetrates more spheres of user’s life, the systems have to be reliable, fair, and accountable. Consider the case of an AI system that calculates the creditworthiness of an unbanked customer based on “alternate” credit data — blue-collar or gig-economy paychecks, usage and repayment of micro-loans, social profiles, to name a few. Both the non-traditional lenders and the users have to trust the system for this product to be effective in the long run.
Almost all spheres of life, starting from everyday products we use in retail, eCommerce, banking, entertainment (to name a few), or the interactions we have, with increasingly smart systems or devices, will start getting incorporated with AI. The practitioners of AI will have to ensure that the public can be certain that the technology is transparent, secure and that its conclusions are not biased or subject to manipulation. In 2020, technologies that provide a measure of trustworthiness and “fairness” will start getting incorporated into AI lifecycle to help us build, test, run, monitor, and certify AI applications for trust, not just performance.
3. AI will further move to the edge
My “AI Prediction for 2019” predicted a shift from cloud-only to cloud-edge hybrid strategies to enable more effective Machine Learning (ML). While 2019 saw the development of technologies such as Federated Learning to kick start that, 2020 would see a major acceleration of that trend. Today, the ability to analyze high-fidelity, high-resolution, raw machine data in the cloud is expensive in terms of both transport and compute and therefore, does not often happen in real-time, resulting in minimum business value derived from the data being collected.
Take the case of digital Over The Top (OTT) entertainment and B2C internet apps. Being able to provide real-time recommendations to a music enthusiast based on the last song they skipped or listened to, would be highly valuable. However, today, many organizations have settled for smaller-sized sample data or time-deferred data for their efforts, which provides a semi optimized outcome.
The maturity of technologies such as Federated learning are now making it possible to put “edgified’ versions of ML models in real-time on the remote infra or even on the leading smartphones with AI processors.
4. Man and machine will make their “first contacts”
Man and the machines driven by AI are already starting to work together, consider the case of Alexa or Siri at your home, or the Google Assistant this morning which alerted me of unexpected early morning traffic to the airport, enabling me to catch the flight and write this piece on the flight today.
One of the first places we are seeing the rise of the machines is in the contact centers. The processes there are more standardized and codified, the costs of services are still relatively high, and the consumers’ expectations for “instant services” across a growing number of digital channels have challenged even the best-managed contact center management. Rather than replacing the agents outright, AI can take away the repetitive tasks (account balance, address change, the addition of new services). Most importantly, the quickly emerging ability of AI to handle complex queries are already complementing agents, enabling them to provide more contextual and informed responses across channels.
As with any new technology implementation, AI in the contact center will give rise to its own challenges — a customer journey that appears to be too automated may end up creating alienating or frustrating customers, especially baby boomers who are often are the top paying customers and expect personalized services.
5. AI pros will be in the limelight, everywhere
The “AI Winter” was officially over in 2019, when the pioneers of AI, Yoshua Bengio, Geoffrey Hinton, and Yann LeCun — sometimes called the ‘godfathers of AI’ — were recognized with the $1 million Turing prize for their work developing the AI subfield of deep learning.
AI and ML have been the hot topics with the news have largely been driven by tech companies such as Facebook, Google, Netflix etc. Many non-tech legacy enterprises, having created their ‘AI strategy,’ have started focusing on solving real-world problems that move their business metrics. After spending the past few years on digitization efforts to get their data pipelines in order and identifying opportunity areas where AI could bring rewards, legacy enterprises are moving ahead with use cases and applications of AI.
Keeping up with the sudden growth of demand, hiring for AI practitioners has grown 74 percent annually over the last four years, according to LinkedIn, with the top practitioners drawing cult-like followers and professional football player type compensations.
6. And, finally, the dark side of AI
AI is not an “all cure” for the technology and civilization, as sometimes the columns in your daily newspaper might make you believe. If not handled carefully, AI can cause harm in ways not imaginable today. We have already seen the early days of abuse. Facial recognition systems can now be fooled using AI-generated 3-D printed facial masks. AI-generated automated fake contents today are causing havoc and the next generation of the same are considered too dangerous to be released to the general public by its creators. Overuse of personalization without consideration for privacy, can harm, or offend customers, employees, and society in general.
Summary
2020 will be a watershed moment for AI, with the technology out of the experimental cycle in 2018–19, and start becoming adopted in all types of businesses, processes, products, and services. In 2020, businesses will get real operationalizing AI and measuring impacts. Consumers’ understanding of technology will start changing. However, as AI penetrates more spheres of users’ lives, the systems have to be reliable, fair, and accountable.
In 2020, AI will further move to the edge. While 2019 saw the development of edge computing technologies such as Federated Learning, 2020 would see a major acceleration of that trend. However, privacy will still be a challenge that businesses will have to address in order to ensure widespread public acceptance of the technology.
Epilogue: I write on Data Science, Machine Learning, Product Management and Career Success Stories. You can follow me to get these in your Medium feed.
Previous Story: Powering the World with “India Class” Technologies
Next Story: The Economic Impact of Poor Addresses in India: $10–14 billion a year
