Artificial Intelligence, specifically, machine learning and deep learning, have been fashionable keywords of 2018 and we don’t expect the hype to die down in the next few years. In the long run, AI will eventually become normal everyday news, being yet another technology powering our lives, just like what happened with the ‘internet’, ‘electricity’ and ‘combustible engines.’
In the next few years though, exciting technological breakthroughs will change the way we live, conduct business and run our societies. AI promises to bring about changes that are beyond anything we dreamt of during previous technological revolutions. Unlike the past, AI-driven machines will do the “thinking” work of analyzing, planning, predicting and making decisions, contributing to new roles that we have always considered reserved for human beings.
We are still unsure whether AI will eventually lead to a “Star Trek”-like-society, with humans free to spend their lives following more meaningful pursuits. Many are however convinced that AI will lead to mass unemployment and social unrest, eventually leading to a Skynet style elimination of human race. While I do not have crystal balls in the distant future, here are my predictions for 2019.
I no longer own my music CDs or movie DVDs, we subscribe to Spotify or Netflix. Among everyday products bought by people, music and videos were probably the first products that signalled the beginning of the end of ownership era. Today, AI platforms are in the midst of turning every manufactured product and services into a connected ‘smart’ product. We have already seen this in transportation and consumer electronics – cars, scooters, washing machines, coffee makers, thermostats, etc.
Driven by data captured from smart products; AI algorithms and new business models, the trend to end ownership will accelerate all industries, products or services. We are starting to subscribe to office space (WeWork), housing (Roam, Common), furniture (Fernish), clothes (Le Tote) and, even dog walking (Wag). These trends will accelerate and similar services will start propping up in all walks of life in 2019 and beyond.
While AI and ML have been the hot topics, the news has largely been driven by tech companies such as Facebook, Apple, Amazon, Netflix and Google (“FAANG”). Many non-tech enterprises, having created their ‘AI strategy,’ will now focus on solving real-world problems that moved their business metrics. After spending the previous few years on digitization efforts to get their data in order and identifying opportunity areas where AI could bring rewards, enterprises will move ahead with proven initiatives, learning from a pilot, and then soft-launching to global deployment.
In this scenario, a retailer focuses on building customer engagement model to maximise omnichannel presence and conversion to sales. Alternately, a churn prediction model will help them get early signals that a customer might be disengaged and may stop shopping altogether, necessitating business interventions to prevent this.
Reaping the benefits of digitization and AI, businesses will start using their data to generate new revenue streams. Building up large databases of transactions and customer activities and partnering with adjacent industries can essentially let any sufficiently data-and-AI-savvy business to begin to reinvent itself. For example, a telco can start building models for when a customer will likely buy a new smartphone. Armed with the model prediction, it can partner with phone manufacturers to offer customers a highly customised incentive to enable a transaction – generating revenue for itself in the process.
Consequently, we will see the focus shift from “AI strategy” to “AI-driven” results as companies look for real business impact from their technology and people investments. The technology will be less important: the business insights and results delivered will be the key. On the flip side, as AI makes inroads, enterprises will start to realise that AI is an investment for the transformation of their processes, people and culture, and not just a magic tool that can be used to instantly fix inefficiencies.
As AI gets beyond the hype and the daily headlines, and as usage of AI-based devices and services surges, our understanding of AI will shift. Initially, the daily interaction with AI will be in the form of digital assistants such as chatbots or voice bots and devices like Alexa. As interactions and usage increase, we will no longer associate AI with just autonomous cars that never crash, rather as productivity tools and predictions to help everyday tasks and make our lives better. Practical AI will be focused on making shopping delightful, patient care better, disease detection more precise, and learning more enjoyable.
On the flip side, while AI will work well in most cases, we will see occasional glitches or ridiculous failures. This because of a lack of deep understanding of the underlying statistical nature of AI by many “data scientists” and a dearth of the programmatic approach to implementing algorithms causing unintended consequences.
As more businesses use AI to power their products and services and start relying on data-driven decisions, it will take time for the entire ecosystem to develop the new processes and frameworks to work with it. For example, a marketing department, before deploying a customer-churn-prevention globally, would want checks and balances to ensure no “revenue leakage” or customer harm takes place. This is especially problematic when it involves dealing with human data and as AI is still hindered by the “black box problem”, most people outside the community of data scientists and some even within, do not seem to understand what the system is doing.
As the new AI ecosystem takes time to adapt to new processes and frameworks, bad actors will take advantage of the infancy of the systems. Across the digital ecosystem, using data platforms and sophisticated AI technologies, they will increase their efforts manifold and execute global and highly customised fraud schemes, resulting in significant losses for brands and marketers. A number of risks surrounding sensor tampering, data manipulation, priming, and sophisticated AI model-driven frauds and phishing attacks will come to light.
As enterprises drive AI into their systems, processes and everyday businesses, AI needs to be trusted to achieve its full potential. Consumers of AI would want to know what it is doing with our data, why and how it makes its decisions when it comes to issues that affect our lives. From a technological perspective, this is often difficult to convey. What makes AI useful is its ability to draw connections and make inferences that are non-obvious or even may be counterintuitive to us. Consider the case of when Google’s AlphaGo beat Lee Sedol, one of the world’s top Go players in 2016. No human could understand the moves by AlphaGo and there was a sense of disbelief as the game was being played in front of spectators and commentators. Besides reassuring public, research and business will also benefit from openness which exposes bias in data or algorithms.
In 2019 we will see an increased emphasis on technologies and processes designed to increase the transparency of AI, driven by The General Data Protection Regulation (GDPR) and similar measures that will come into effect in major economies such as India. The GDPR, put into action across Europe in 2018, gives citizens protection against decisions made by machines, which have “legal or other significant” impact. The drive for businesses, especially the leading ones, to “Googlify” their businesses through the power of AI, will lead them to share their data with third parties. Ensuring data privacy, and in turn, customer privacy, will not only be a good business practice and risk management strategy but will soon become a legal requirement.
Privacy-enabled AI technologies will provide the foundations to enable AI applications while maintaining strong privacy using cryptography in 2019. One of my favourites is the exciting emerging technology of Secure Computation. Homomorphic Encryption (HE), one of the secure computation techniques, is a specific way of encrypting data so that third parties can operate and glean valuable insights using machine learning techniques while the data continues to be encrypted, thus preserving the privacy of the users.
Based on HE, Federated Learning (another distributed machine learning technique popularised by Google that does not require centralised data) and other Secure Computation methods we will see startups focussed on democratizing AI on the edge. The bet here is that a billion+ smartphones are being equipped with AI chips and significant local compute in the next 3 years, many of the AI models will be able to run locally on these mobile devices. Distributing the computations over billions of smartphones will drastically reduce the cost and time to develop AI products such as hyper-personalised recommendation engines, AI assistants, etc. for a majority of the enterprises. Both, large companies and startups are building the distributed, secure and privacy-enabled computing framework to enable this.
2019 will be a watershed moment for AI when the technology will be gradually coming out of the hype cycle and start becoming widely adopted in all types of businesses, processes, products and services. Consumers’ understanding of the technology will start changing: however, privacy will still be a challenge that businesses will have to address in order to ensure widespread public acceptance of the technology. Privacy-enabled AI platforms, smartphones equipped with AI chips and significant local compute, will change the way AI gets distributed. By the end of the year, AI would be delivering highly personalised content and recommendations that will delight the consumers and would feel unnaturally personal.