The year 2019 will be remembered in the software world as the year when containerization, cloud native architectures, and Machine Learning broke out into the mainstream.
As we approach the end of the decade, it’s time to look forward to the year 2020 and make some predictions about where these disruptive technologies will take us in the next 12 months. Read on to see what we can expect from Artificial Intelligence (AI) and Machine Learning in terms of growth, innovation and adoption as a new decade begins.
Table of Contents
Shifting from adoption to automation
Marc Andreessen famously said that “Software is eating the world,” and these days it seems like every organization is becoming a software company at its core. The year 2020 will, of course, bring about new trends in technology, and failure to adapt means increased technology debt for enterprises. This debt will eventually have to be repaid with compound interest. Therefore, rather than growth in tech adoption this year, we may expect to see a shift in tech spending. Enterprise budgets will continue to move from IT to the business side of the house, with far more funding for initiatives that increase revenue as business value replaces velocity as the most meaningful DevOps metric.
The focus of software development and information tech spending will be on the implementation of Artificial Intelligence. One of the major themes of 2020 will be the automation of existing technologies. AI-based products like Tamr, Paxata and Informatica CLAIRE that automatically detect and fix outlier values, duplicate records and other flaws, will continue to gain acceptance as the only way to cope with cleansing Big Data and maintaining quality at scale.
Faster Computing Power
AI researchers are only at the beginning of understanding the power of artificial neural networks and how to configure them. This means that in the coming year, algorithmic breakthroughs will continue to come at an incredible pace with almost daily innovations and new problem-solving techniques. AI can address a wide range of hard problems that require finding insights and making decisions. But, without the ability to understand a machine’s recommendation, humans will find it difficult to trust that recommendation. So, expect continued progress in improving the transparency and explainability of AI algorithms.
AI computing power at the edge will definitely improve in the coming year. Established corporations like Intel and Nvidia, as well as startups like Hailo, are working to provide cheap and fast neural network processing via custom hardware chips. As the industry determines that it needs more and faster computing power to run Machine Learning algorithms in real time, more institutions will develop hardware fit for data sources along the edge.
Machine Learning will come mainstream in SME’s
Machine Learning saw tremendous growth in 2019, and we can only expect it to persist and become more accessible in 2020. Machine Learning will become widely available to medium-sized companies as Natural Language Processing (NLP) enters a golden age. Machines are now better than humans at some NLP tasks like answering questions based on information inferred from a story. BERT, the hottest NLP algorithm in 2019, will be forgotten by the end of 2020, replaced by ERNIE or some other whimsically named new algorithm.
Machine Learning will also continue to be introduced as a component of almost every software product category, from ERP to CRM to HR, making it a staple in daily business management. Additionally, Python will strengthen its hold as the Machine Learning language of choice, lowering the technical barrier to entry and allowing more individuals the chance to try out the latest Open Source AI algorithms.
Despite the availability of Machine Learning to a wider user base, the name of the game will still be data. Those who can leverage more information will reap the most benefits from their analytical models. Because its government collects such a vast amount of data, China will continue to lead the world in supervised learning accuracy. To counteract this, expect the Western world to pioneer advances in algorithms that require less training data, for example, active learning, where the algorithm asks for the next best piece of training data to maximize its learning speed. Efficiency in data training will also improve thanks to AutoML tools like Amazon’s SageMaker and Pachyderm, which automate the process of creating and deploying new machine learning models.
Consumer-centric solutions in AI and ML
As accessibility increases, the number of consumer-facing devices employing AI and Machine Learning will follow. Digital assistants and chatbots have become a staple in our daily lives, redefining customer service and in-home internet connectivity. Products that integrate Amazon’s Alexa or Google’s Assistant will proliferate, and smart speakers will continue to enjoy a sales boom as consumers remain loyal to their digital helpers.
In the retail space, an initial rollout of in-store frictionless shopping will begin to redefine the industry. Integrated AI will be able to train computers to identify a product’s location and the items the consumer put in their shopping cart. We may also see the use of augmented reality in physical spaces that will guide customers through the store. Because AI and computer vision technology can seamlessly identify and bill for a customer’s purchase while he or she shops, retail will transition to a customer experience free from friction points like checkout counters and create an undisturbed retail reality. The technology for frictionless shopping will not be ready for mass rollout in 2020, but expect to see progress in trial locations.
Finally, as hopeful as we are that each new year will bring us the perfect driverless car, automated driving will not be our reality in 2020. The Machine Learning algorithms that power automated vehicle systems still have too many fundamental flaws to be fully trusted. For example, a stop sign can be augmented with pixels that are invisible to the naked eye but cause machine learning algorithms to read it as “Speed limit 40 mph.” These types of failings are what prevent the full-scale development of driverless cars. Widespread adoption can only come to fruition once algorithmic weaknesses are addressed and systems can be trusted to keep drivers and pedestrians safe. In the meanwhile, we will see the continued rollout of AI-assisted driving, where AI provides guidance and warnings to a fully active human driver.
Overcoming AI and ML barriers
Although we can expect remarkable advancements in AI and Machine Learning in the coming year, there will be some impediments to its propagation.
The severe labor shortage of skilled Machine Learning engineers will make it difficult for second tier companies to keep up. While accessibility may grow and provide a gateway for midsize organizations, those already in possession of tremendous amounts of usable data and the employees capable of leveraging it will be the ones to thrive, and ultimately have the biggest advantage in terms of successful AI and machine learning integration.
Trust will also remain a barrier in our adoption of Machine Learning and AI next year. In addition to flaws in autonomous vehicles that put safety at risk, ethical concerns about biases in algorithms remain without solutions. Can we rely on insights derived via training data that may express historical bias against women, the elderly, or minorities? This must be addressed before humans will be able to fully embrace the autonomous decision-making of AI tools.
Finally, a bit of perspective: all the advances described here are part of “narrow” AI, where a machine performs a specific task better than a human being, based on algorithms and statistics. The Holy Grail of AI is “general” intelligence, where the machine has a base of real-world knowledge and logical capabilities that enable it to apply knowledge and skills to new tasks. Narrow AI is progressing by leaps and bounds, but general AI is still many decades away.
The coming year is set to be a challenging new age for tech with many innovations and disruptions. The benefits of ML and AI are clear, and accessibility is increasing. But significant issues will still need to be addressed before its widespread impact on businesses and consumers can be fully realized. As a new decade begins, it will be interesting to see how many of these predictions come to fruition.