M+E Connections

Recent advancements in machine learning, thanks in part to growth in computing power, have significantly increased what is possible to achieve using artificial intelligence (AI) across multiple enterprise sectors, according to Michał Iwanowski, product director at deepsense.io, a deep learning solution provider.

AI has already been yielding good results for many years in fields that include finance, he said during an Aug. 23 webinar. AI algorithms, for example, have been employed in stock trading to take over certain functions from humans and they “greatly surpass” the capabilities of human traders when it comes to short response times, he said.

AI algorithms also assist humans in making decisions in the areas of expert systems and mechanical engineering, where it is often difficult to find out what caused a failure, he said, adding AI algorithms can figure out the cause much faster than humans. AI algorithms have also been used effectively in games including chess, he said.

However, AI algorithms haven’t been so successful when it comes to natural language processing and image recognition, he said. As an example of the latter, it’s still a “very difficult problem” for AI to figure out what image contains a donut and which one contains a Chihuahua, he said.

But there’s been “huge progress” made in image recognition during the past couple of years through deep learning, he went on to say. Google’s DeepDream computer vision algorithm, for example, uses deep neural networks to “amplify the features that it is trying to detect,” he said.

AI can be used across many enterprise sectors to create “tangible savings” and “make completely new things possible,” he said. They include advancements in traffic safety, he said, pointing to data that showed about 20% of U.S. car accidents are caused by distracted driving each year. An insurance company asked whether something can be done to prevent this by using a dashboard camera, he said. Research has shown that deep neural networks can be used in such usage scenarios to tell if a driver is distracted or not — whether from texting while driving or from doing something else — with 96% accuracy, he said.

“There are multiple areas where machine learning,” especially deep learning, “can do things that were not possible a few years before,” he said. But “the biggest problem” with machine learning today is “the awareness of industry experts in various domains who may not be aware that there are solutions that are able to solve their problems,” he said, calling it an “awareness chasm” between AI experts and industry experts in many industries.

The AI market “promises to be one of the fastest growing business and technology markets over the next decade,” Eliot Weinman, CEO of AI World conference producer Trends Equity, said in an Aug. 17 news release promoting the webinar. “Over the past few years we’ve seen dozens of AI companies quickly acquired by the likes of” Google, Apple, Facebook, Amazon, IBM and others, he said, citing Venture Scanner data saying there are 930 global companies across 13 categories in AI that have raised $4.4 billion.