For artificial intelligence and machine learning, the era of expansive hype and limited substance is swiftly coming to a close. Organisations big and small have progressed from researching AI and ML’s potential to purchasing and deploying the technologies — confident that these investments will deliver substantive benefits.
Venture capital investment in AI now tops $3 billion annually, and the number of active startups in the U.S. that are developing AI technologies has gone up by a factor of 14 since 2000. Despite this momentum, the industry is still in the early days of delivering enterprise-relevant AI- and ML-backed solutions that are manageable and provide high ROI. Early adopters, however, are already achieving significant benefits. Deloitte surveyed 250 early adopters of AI and found that 83 percent said they “have already achieved either moderate or substantial benefits from their work with these technologies” across a wide spectrum of business activities. And 76 percent of those early adopters say cognitive technologies will transform their business in three years or less while only seven percent gave a timeline for transformation beyond five years.
From fraud detection to IT optimisation, AI and ML are delivering on years of promise — and 2019 will see wider adoption as “early adoption” gives way to “table stakes.” Open source systems and communities will become integral in AI and ML development.
As AI- and ML-powered tools take centre stage, they’ll require a supporting cast of tools and communities. Larger volumes of data and projects focusing on building out AI- and ML-backed tools are essential. Open-source systems and communities create a haven for those willing to explore — and they open doors to affordable, accessible means of processing large data sets, usually via the cloud.
Forward-leaning organisations are already augmenting home-grown capabilities with open source software (OSS) systems and communities. TensorFlow and other OSS systems are useful for large-scale machine learning processes and deep insights (i.e., distinguishing spoken words from gibberish, or translating a word from one language to another). Other systems, such as Apache Spark, facilitate running repeat queries on data sets to sharpen and improve insights. OSS communities, such as GitHub, make it easier to share ideas and gain feedback across teams and organisations.
Expect this open collaboration to continue and pick up speed as organisations and researchers increasingly throw their hats into the AI ring. Working together toward widespread application of AI and ML is imperative if we want it to happen quickly and effectively. Doing the inverse will only stunt growth.
AI will create new ways of interacting with data (and machines)
Fear of AI and ML replacing workers abounds, but there’s more to the narrative. The human in the loop (HITL) won’t be eliminated; intuition, quality assurance and general training will remain vital to optimised AI and ML deployments. There are human tasks that technology can bolster but not quite replace.
HITL has emerged as a key design pattern for managing teams where people and machines collaborate. The goal is to manage the impact of AI and ML to be less jarring and more useful and accessible. It also allows AI to offload the edge cases it can’t handle to humans the same way we offload monotonous tasks to machines.
So what is AI and ML taking out of the picture? Smart technology is taking on the complexity and load of tedious and data-heavy tasks so practitioners can focus on higher-order work. In the process, seen in the aggregate, jobs won’t be eliminated, they will evolve. McKinsey suggests that by 2030, 375 million workers — 14 percent of the global workforce — will need to “switch occupational categories.” Gartner predicts that AI will create 2.3 million jobs in 2020, while eliminating 1.8 million.
In the transition, effective AI and ML will require the development of new skills and the advent of new processes as humans take on the responsibility of overseer. The value of AI and ML will increase organically as this happens, as smart tools are fed more data and trained in new scenarios. The day-to-day consumer is already experiencing these organic AI benefits: Amazon continues to recommend great books; Siri conversations are becoming more natural; ESPN football predictions are increasingly accurate.
Additionally, engaging with most of these innovations will not, and cannot, require a doctorate degree in data science. There are not enough Ph.D.s available. Look for more intuitive interface options to emerge. From natural language search to prebuilt, point-and-click models, AI will no longer be the domain of the few. Expect more substantive and tangible results across an ever-expanding set of scenarios as a result — from spot-on sales forecasting, to accurate weather and traffic projections, to precise anomaly detection and automated remediation.
AI and ML will take off in healthcare and finance
Smart technology is now being launched and trained in specific business settings, and we can expect more of it in highly regulated industries.
In financial services:
Unsupervised machine learning techniques will increasingly help banks and insurers segment their customers and offer personalised, targeted products. This technology will also improve speed and agility, helping organisations compete with specialised fintech firms through enhanced customer intelligence.
Machine learning will boost regulatory compliance using automated reports, stress testing solutions, and behavioural analysis of e-mails and phone records to identify suspicious customer or employee behaviour. It will also enhance fraud detection, improve anti-money laundering efforts and more effectively detect credit risk.
Finally, by analysing the constant data being generated by consumers with machine learning, financial services companies will be able to automate back-office operations, reduce errors and accelerate process execution in the year to come. This will allow insurers to improve and automate the handling of claims by recognising patterns in pictures or individuals involved in damages, for example.
We know that the volume, variety and velocity of data have exploded in the healthcare industry in the past few years due to electronic health record (EHR) adoption. In parallel, that wealth of data brings opportunities to better predict and manage medical conditions. Already, the first AI-powered diagnosis of images was approved by the FDA in April. Many businesses have been working on continuing this trend. Expect even more precise results and recommendations, like tailored treatments, to become more accessible as researchers and physicians increasingly come to use AI and ML in their diagnostics processes.
The natural progression of AI and ML adoption in the space means AI will affect a majority of U.S. patients — and most patients won’t even know it. For example, U.S. patients will be unknowingly affected by the discreet AI solutions that their providers use for clinical decision support, by payers to predict their risk of hospitalisation or by pharmaceutical companies using chatbots for managing patient engagement.
Image Credit: Sergey Nivens / Shutterstock