olly - Fotolia
- Ed Burns, Executive Editor
Take a look at the range applications we call artificial intelligence today, and you're likely to notice that the technology is not very useful.
That's not to say engineers and programmers haven't made huge strides in recent years, developing deep learning algorithms capable outperforming humans in diverse tasks from image classification to strategic game playing. But problems occur when developers try to connect AI functionality to business processes. Developers are increasingly finding that, in the real world, AI functions differently than it does in training environments. Though industry people are very excited about AI, and the hype surrounding the technology has grown significantly, enterprises might have to wait a while before they see truly game-changing tools.
Take, for example, chatbots. These natural language processing-powered algorithms are popping up everywhere, from online ordering systems to smart assistant platforms. They represent some the more advanced and usable tools in the AI world right now. But when it comes to actually chatting with them, these bots don't seem to be so intelligent.
"We've been doing natural language processing for a long time, but it still has a long way to go," said Michael Facemire, analyst at Forrester Research, in a presentation at December's AI World Conference in Boston. He said today's chatbots are generally effective when they function as order-takers. If you tell a chatbot what you want in clear and concise terms, and your query corresponds to something the bot is familiar with, then it can return a sensible response. But the more you try to converse with a chatbot, the more likely the conversation stops making sense. The conversational abilities chatbots only go so deep.
Better AI looms on the horizon
So if today's AI functionality is so limited in its utility, why are we seeing so much hype around the technology? Because tomorrow's applications could be truly revolutionary.
The AI hype train may have left the station a little early. We've been talking about it as an influential technology for a couple years now, when in fact AI has done very little substance for many enterprises. But that doesn't mean AI lacks promise.
"The AI revolution started a few years back with machine learning," said Deep Varma, vice president engineering at online real estate listing site Trulia. "We're all collecting data at a much faster pace, and now businesses are trying to figure out what to do with this data. That's why we're hearing more about AI."
For Varma, the reasons AI is so trendy are twofold. On one side is the demand. AI is increasingly being seen as a means to make sense of all the data piled up over the last few years as more enterprises pursued big data. But the tool side is where things really get interesting. Varma said there's been a steady progression from the machine learning-based recommendation engines of the last decade to more impressive functionality like visual search and natural language generation -- two AI technologies Trulia currently uses to improve its site usability and fill out content.
Business AI demands will overshadow consumer AI
Artificial intelligence is poised to explode over the next decade, and most of that growth will be seen in business applications, according to Tractica, a research firm that focuses on human interaction with technology. Tractica estimates that the enterprise AI market, expected to be at more than $50 billion annually by 2025, "will be 10 times bigger than the consumer AI market."
Although "AI is highly fragmented," Tractica research director Aditya Kaul said during December's AI World Conference in Boston, it's not concentrated in hyperscalars or consumer web, internet and hardware companies. Citing Tractica research, Kaul characterized enterprise AI as companies building their own apps, including specialized product companies, platform providers and infrastructure players. Eventually, there will be consolidation in the platform space, according to Kaul, "while specialized AI products will converge around specific verticals. Internal AI development will continue specifically in highly specialized verticals like automotive, manufacturing, finance and investment."
The main market drivers for AI, he said, will be big data sets, internet of things analytics and significant improvements in machine learning algorithms, while the main barriers will be defining analytics AI, too many technology choices, access to clean and labeled data, and overhyped AI capabilities. -- Ron Karjian, Managing Editor
In the year ahead, Varma said he expects his team to dedicate substantial time to using AI to personalize Trulia's services for individual users. In an industry like real estate, where customers have unique needs and desires, depending on their lifestyle, a service needs to be specially tailored to individuals, Varma explained. So in the year ahead, he plans to use machine learning to develop a better picture of users, learn their preferences and anticipate what types of content and listings suit their needs. These learning systems will then be able to change the site and related services like email to give each user a customized experience.
"If we are going to help our customers make the right decision, it has to be personalized," Varma said. "That's where AI comes in. We've been investing a lot in our recommender systems and personalization platform."
Beyond personalization, Varma said Trulia is looking at ways to use data from real estate listings to build augmented reality views of homes, enabling buyers to see what a home would look like with some cosmetic changes or with their own furniture placed in a room. All of this will be driven by deep learning algorithms that interpret objects in images and, paired with listing data and images, create true-to-life images.
Can hundreds of enterprises be wrong?
There's good reason for companies like Trulia to be excited by what AI functionality could look like, said Heath Terry, managing director of Goldman Sachs' research division. In a keynote address at AI World, he said enterprises in industries from agriculture to retail are making investments in AI. Whether they're implementing fairly run-of-the-mill recommendation engines or more advanced chatbots, AI has become a primary focus for enterprises.
Terry cited his own research showing that agriculture, finance, healthcare and other industries could each see billions of dollars in value created from AI. Venture capital funding of AI companies doubled in 2017 compared to 2016. During that time, public companies were four times more likely to talk about their AI capabilities on earnings calls, and that's not limited to tech companies.
Heath Terrymanaging director, Goldman Sachs
All these companies are responding to the same developments, Terry said. They have more data than ever before, which can be useful for training models. Compute power and storage are cheap. And deep learning algorithms have made substantial advancements in recent years. All of that adds up to a strong foundation for AI.
"We've been through periods of AI excitement before, and we've all seen the AI winters," Terry said. "What makes this different as opposed to some of these periods we've seen in the past? It's the data that's available. We have faster hardware, and [data] can be analyzed in real time. It's open source and better algorithms, frameworks like Caffe, TensorFlow and Torch."
Thanks to all the hype that has built up around AI functionality in the last couple years, some enterprises may expect quick and substantial gains from the technology. But that's not likely to be the case. While we have the foundational elements for AI success, building effective tools and using them in ways that move the dial on real business problems can be a long process.
AI finds strength in niche markets
Artificial intelligence is still in the early stages of wide-scale penetration into consumer, business and government markets, but there are plenty of opportunities for AI applications in niche markets. Amid all the hoopla that AI is poised for exponential growth, many companies have narrowed their focus to specific and lucrative applications.
"We have some visionaries like Stephen Hawking, Elon Musk and Bill Gates [with] tremendous credibility usually terrified to some degree about computers taking over the world, likely when we get to the expected promised land of generalized artificial intelligence," said Boyd Davis, CEO of machine learning platform maker Kogentix, during December's AI World Conference in Boston. "The reality though is that AI technologies that are practical and implementable today are narrow technologies -- tools that can solve some very targeted problems and create meaningful opportunities."
AI is expected to bring added value to a growing number of niche problem areas, according to Tractica, a research firm that focuses on human interaction with technology, allowing small and medium-sized businesses to compete with larger companies in providing services like algorithm development, training and support, cloud services and maintenance. -- Ron Karjian, Managing Editor
At AI World, Marc Hammons, principal software architect at Dell, talked about how his team helped roll out an automated chatbot for customer service. The Ava smart assistant can deal with customer inquiries and help them through the purchasing process. Hammons said this didn't come about overnight. It required a focus on the business problem and faith that investing in AI will eventually pay off.
"Start practically, small, build trust," Hammons said. "These things are coming and they will benefit people."
How chatbots and AI are changing the insurance industry
- Optimising content management workflows with AI –ComputerWeekly.com
- Using Cloud-based AI Technology for Remote Language Testing –Emmersion Learning
- NVIDIA Guide to Deep Learning and Artificial Intelligence –Hewlett Packard Enterprise
- A Guide to Enterprise AI - 451 Research Pathfinder –Hewlett Packard Enterprise