AI technologies have come far, but the road is long
Machine learning and other artificial intelligence technologies are poised to offer businesses big benefits, but companies have to walk before they run with AI.
If you believe what you read, AI technologies have officially permeated the tech market, and you need to get on board now or be left in the annals of history with Blockbuster and Myspace.
Problem is, it isn't that easy -- at least, not yet. We've only begun to see what AI technologies such as machine learning can do, and only a small percentage of companies actually have the right data, people and money to invest in the cutting edge.
To complicate matters further, a number of companies that claim to offer AI products really only provide one aspect of it, like speech for building chatbots. It makes me wonder if marketers realize that when they overhype products, their customers are underwhelmed and it denigrates all AI. And do they realize that AI washing makes jaded tech journalists even more skeptical than they already are? (I'm asking for a friend.)
As writer Ed Burns says in his feature, "The AI hype train may have left the station a little early."
That's not to undercut the value machine learning and other tools that come under the umbrella artificial intelligence; those are the seeds what will grow into strong AI. Indeed, machine learning, natural language processing (NLP) and image are all revolutionary in their own right, says Nick Patience, co-founder and research vice president at 451 Research, where he covers digital transformation as well as AI and machine learning.
Playing devil's advocate, I can see the case for holding back. IT pros know that only fools rush in and adopt 1.0 technology.
"The ability to have computers what an image is, using training data unaided or human speech or being able to see using cameras -- that's huge," Patience says. "If you look at cloud, mobility, PCs -- [AI] is as big as any these, if not bigger."
Analysts say it's important for companies to learn about AI technologies now -- and to implement them where they make sense -- because this stuff is complicated, and if you snooze on this one, you really will lose.
Playing devil's advocate, I can see the case for holding back. IT pros know that only fools rush in and adopt 1.0 technology. It's safer to wait a hot minute for vendors to work out the kinks and make improvements before you invest in something. That approach can be easily applied to machine learning, which, as any data scientist will tell you, requires a lot time, data and money -- not to mention an advanced education.
AI tools afoot in 2018
For AI to be democratized so that anyone can use it, there need to be user-friendly tools, said Samuel Madden, faculty co-director at MIT's Computer Science and Artificial Intelligence Laboratory, or CSAIL, during his presentation at December's AI World in Boston. Machine learning is a "black art" that requires multiple doctorates to make it work, and it's difficult to maintain and evolve, Madden explained. "It's also not fast; it takes a long time to train and develop," he added. "That doesn't fit in the rapid test-dev-deploy cycle software engineers are accustomed to."
MIT is working to address those challenges by building tools that make machine learning easier to use, integrable and tied to production. The commercial market is working on it, too, and 451's Patience expects we'll soon see tools that people who don't call themselves data scientists can use.
AI's sky-high trajectory
AI technology adoption is only beginning in enterprises, but it won’t be long before artificial intelligence pervades business applications, products and devices:
12% of IT professionals say AI is a broad initiative this year, according to the "TechTarget 2018 IT Priorities Survey."
76% of early adopters of AI technologies say artificial intelligence will substantially transform their business in three years, as reported in Deloitte's 2017 survey "The State of Cognitive in the Enterprise."
85% of CIOs will pilot AI through a combination of buy, build and outsource projects by 2020, the "Gartner Executive Summary Report" estimated.
Emotion AI systems are becoming so sophisticated that by 2022, our personal devices will know more about our emotional states than our own family, according to "Gartner Predicts."
The amount of revenue generated from the direct and indirect application of AI software by 2025 will reach $60 billion, as reported in Tractica's "Artificial Intelligence Market Forecasts."
44% of AI revenue will come from deep learning technologies, with machine learning at an additional 23%, totaling a combined two-thirds of all AI revenue in 2025, Tractica's survey, "Artificial Intelligence: 10 Trends to Watch in 2017 and Beyond" predicted.
"It's easy to set up on IBM Watson or [Amazon Web Services] and have a [machine learning] data set, but the next steps aren't easy -- the training models," Patience says. "We will start to see more tools to bridge that gap." As machine learning becomes operationalized, he notes, software vendors will begin to sell tools across the coding spectrum -- including low- and no-code platforms -- that can be used by software developers.
Get into AI early but carefully
Even before those tools become available, Patience believes it's important for company leaders to understand machine learning and its impact, because it has already started to affect many of the technologies consumers and businesses use.
Research shows companies are heeding that advice. In TechTarget's 2018 IT Priorities Survey, 12% of 940 respondents within North America said AI is a broad initiative this year. That’s up from 9.9% in the 2017 survey. These stats jibe with what Gartner sees as well. In a November 2017 report, the research firm said AI investments are in the early stages in many domains. In just two years, however, that number will be exponentially higher; the firm predicts 85% of CIOs will be piloting AI through a combination of buy, build and outsource projects by 2020.
As more companies dabble in low-level, supervised machine learning using labeled data sets, researchers are moving the ball forward. Unsupervised machine learning algorithms, which label unstructured data on their own, will mature, and hardware providers will amp up CPUs and graphical processing units to do it all faster. Gartner predicts emotion AI systems are becoming so sophisticated that by 2022, our personal devices will know more about our emotional states than our own family.
So, yes, AI technologies are advancing quickly, and they're poised to change the world, but there's a ways to go before we see the full potential. In the meantime, it's difficult to parse fact from fiction about what truly offers the benefits of AI, what's required and when to use it.
I'll leave you with this bit of somewhat self-serving advice: Research machine learning, NLP and the like using unbiased sources; learn about the resources and the requirements you need to use these technologies; and keep a close eye on the products in this market to understand your options. As Robert Bogucki, chief science officer at Deepsense.ai, told AI World attendees, if you're thinking about adding AI to your business, "start small but with the big picture in mind."
