Exclusives

Analyst: Major Progress to Be Made in Machine Learning This Year

Artificial intelligence (AI) continues to evolve and grow rapidly across the media and entertainment sector, and we can expect to see major enhancements in the technology, from methods to tools and ethics in 2018, according to Ben Lorica, chief data scientist at O’Reilly Media.

“Substantial progress” will be made in machine learning “methods, understanding, and pedagogy,” he said in an O’Reilly report, citing that as one of five AI trends to watch this year.
“As in recent years, new deep learning architectures and (distributed) training algorithms will lead to impressive results and applications in a range of domains, including computer vision, speech, and text,” he predicted, adding we can expect companies to “make progress on efficient algorithms for training, inference, and data processing on edge devices.”

We can also expect to see “interesting breakthroughs” as a result of “collaboration between machine learning experts,” he predicted.

But, despite deep learning’s success to date, “our level of understanding of why it works so well is still lacking,” he said, adding: “Researchers and practitioners are already hard at work addressing this challenge. We anticipate that in 2018 we’ll see even more people engage in improving theoretical understanding and pedagogy.”

Meanwhile, another trend we can expect to see in 2018 are new developments and declining hardware costs that Lorica predicted will “enable better data collection and faster deep learning.” Because deep learning is “computationally intensive,” a lot of the hardware innovation around it involves training and inference, he said. We can expect to see new processors, accompanying software frameworks and interconnects, as well as optimized systems assembled to enable companies to “speed up their deep learning experiments to emerge from established“ hardware makers, cloud providers and startups in China and Western countries, he predicted.

However, “the data behind deep learning has to be collected somewhere,” he pointed out. “Many industrial AI systems – lidar for instance – rely on specialized sensors,” he said, adding: “Costs will continue to decline as startups produce alternative sensors and new methods for gathering and using data, such as high-volume, low-resolution data from edge devices and sensor fusion.”
He also predicted that AI and deep learning developer tools “will continue to evolve.” Although TensorFlow is still the most popular deep learning library by a wide margin, other frameworks – such as BigDL, Caffe and PyTorch – “will continue to garner users and use cases,” he predicted, adding: “We also anticipate new deep learning tools to simplify architecture and hyperparameter tuning, distributed training, and model deployment and management.”

Also expected are “many more use cases for automation, particularly in the enterprise” sector, he said, explaining: “As more companies enter the AI space, they’ll continue to find tasks that can be (semi) automated using existing tools and methods. A natural starting point will be low-skilled tasks that consume the time of high-skilled workers. Other applications include automation products using speech and natural language technologies, industrial automation and robotics, and use cases in health and medicine, such as drug discovery, medical assistants, and genomics.”

He also predicted people will “increasingly use automation for creative pursuits,” including AI-generated music, images and visual arts that will start appearing in commercial products.
The innovation isn’t just in the U.S., he noted, saying: “With government support for AI technologies, access to large datasets and users, and a highly competitive market that rewards early movers, Chinese companies and startups are poised to stake a claim in automation.”

The fifth trend Lorica cited was the AI community continuing to address concerns regarding privacy, ethics and “responsible” AI. “Fairness, transparency, and explainability are essential for most commercial AI systems,” he said, predicting we’ll see “continued discussion about how to create tools to create tools to ensure responsible AI,” such as those used to fight against “fake news.”

He also predicted we’ll see “developments aimed at safety.” With Europe’s General Data Protection Regulation set to become enforceable May 25, “we’ll see companies (like Apple) develop or improve privacy-preserving machine learning products,” he predicted.