Deloitte Global recently released its annual Predictions report, which looks at transformation and opportunity in tech, media and telecom over the next one to five years. In part of our report, Deloitte Global predicts that in 2018, large and medium-sized enterprises will intensify their use of machine learning (ML)—an artificial intelligence, or cognitive, technology that enables systems to learn and improve from exposure to data without being programmed explicitly. Deloitte Global predicts that the number of implementations and pilot projects using the technology will double compared with 2017, and they will have doubled again by 2020.
Today, most enterprises using ML have only a handful of deployments and pilots under way, but, according to Deloitte Global, progress in five key areas should make it easier and faster to develop ML solutions. These vectors of progress are:
- Automating data science. Time-consuming ML tasks, such as data exploration and feature engineering, are increasingly likely to be automated. A growing number of tools and techniques for data science automation, offered by established companies as well as venture-backed start-ups, should help shrink the time required to execute an ML proof of concept.
- Reducing the need for training data. Training an ML model can require up to millions of data elements. A number of promising techniques are emerging that aim to reduce the amount of training data required for ML—for example, a team at Deloitte Consulting LLP tested a tool that was able to build an accurate model with only a fifth of the training data previously required. Another technique that could reduce the need for training data is transfer learning, in which an ML model is pre-trained on one data set as a shortcut to learning a new data set in a similar domain.
- Accelerating training. Hardware manufacturers are developing specialized hardware to slash the time required to train ML models by accelerating the calculations required and the transfer of data within the chip. These dedicated processors can help companies speed up ML training and execution manyfold, which in turn brings down the associated costs. Early adopters of these specialized AI chips include major technology vendors and research institutions in data science and ML, but adoption is spreading to sectors such as retail, financial services and telecom.
- Explaining results. ML achievements get more impressive by the day, but many are “black boxes,” meaning it is not possible to explain with confidence how they make their decisions. This presents challenges, for reasons ranging from trust in the answers generated by a model—as when customers are offered incentives—to regulatory compliance. A number of techniques have been created that help shine light into the black box of certain ML models, making them more interpretable and accurate. As it becomes possible to build interpretable ML models, companies in highly regulated industries such as financial services, life sciences and health care can be expected to intensify their use of ML via pilots and deployments over coming years.
- Deploying locally. ML use will grow along with the ability to deploy it where it is needed. ML is increasingly coming to mobile devices and smart sensors, expanding the technology’s applications to smart homes and cities, autonomous vehicles, wearable technology, and industrial IoT.
In response, technology vendors are creating compact ML software models to undertake tasks such as image recognition and language translation on portable devices. Semiconductor vendors are developing their own power-efficient AI chips to bring ML to mobile devices. With smartphones an increasingly viable deployment option for ML, the number of potential applications is growing.
The bottom line
Collectively, the five vectors of ML progress should double the intensity with which enterprises are using this technology by the end of 2018. In the long term, these vectors should help make ML a mainstream technology. Enterprises interested in ML may wish to consider the following:
- Look for opportunities to automate some of the work of oversubscribed data scientists, and ask consultants how they can use data science automation.
- Monitor emerging techniques, such as data synthesis and transfer learning, that could ease the bottleneck often created by the challenge of acquiring training data.
- Find out what computing resources optimized for ML are offered by their cloud providers. If they are running workloads in their own data centers, they may want to investigate adding specialized hardware to the mix.
- Explore state-of-the-art techniques for improving interpretability that may not yet be in the commercial mainstream.
- Track the performance benchmarks being reported by makers of next-generation chips, to help predict when on-device deployment is likely to become feasible.
Interested in learning more? Our full Predictions report is available at www.deloitte.com/predictions, and a recent US cognitive technology survey is online at www.deloitte.com/us/cognitivesurvey. As always, I welcome your thoughts and feedback.Today, most enterprises using ML have only a handful of deployments and pilots under way, but, according to Deloitte Global, progress in five key areas should make it easier and faster to develop ML solutions. These vectors of progress are: