Raise your hand if you've heard this complaint before: There's too much hype about what AI can accomplish, and...
businesses and consumers alike should temper their expectations.
Now, raise your other hand if you've heard this: AI technology is already useful, evolving quickly, and boasts endless potential.
As 2019 nears, there's a lot of mixed information out about what AI technology actually is, where it's going and the possible role the future of AI technology will play in day-to-day business.
Some say current forays into AI have yielded results that don't live up to the publicity. Others argue that those same advances appear to indicate a global shift toward AI in many industries and professions, including healthcare and journalism, for example.
But some critics argue that AI advances are happening too rapidly, that humanity might be at risk of spawning something it will wish it hadn't created. Advocates for AI, meanwhile, decry such messages as doomsday prophecies.
The experts disagree, and they should. The technology has advanced rapidly with little oversight or standards. The future of AI technology is unknown, of course, but a picture of that future is starting to take shape.
And it starts with startups.
More startups, more consolidation
Looking around the showrooms of the Javits Center in New York, crowded with hundreds of attendees of the Dec. 5 to 6 AI Summit, Todd Lohr, principal at KPMG, said he had trouble keeping up with the ever-blooming number of AI startups.
"Every show that I go to, there are a lot of new names," he said.
The conference, in its third year in the city, featured dozens of vendors, developers and potential buyers of AI-powered technology and products. Many presenting vendors had "founded in" dates going back only to 2016 or 2017, and many touted their unique machine learning and natural language processing (NLP) technology and the millions of dollars they've raised recently from venture capitalists.
Yet, despite all the funding and self-described innovation at the fast-growing startups, consolidation and the continued dominance of the AI space by megacorporations like Google and Amazon appear likely in the future of AI technology.
"When you're looking at the long-term AI players, it's going to be about the large cloud providers," Lohr said.
It's an assessment that the Future Today Institute, a consulting firm founded in 2006 by futurist Amy Webb, concurs with.
"Just nine big companies dominate the AI landscape," the firm noted in its "2019 Trend Report for Journalism, Media & Technology." In the U.S., there's Alphabet, Amazon, Microsoft, IBM, Facebook and Apple; in China, Tencent, Baidu and Alibaba rule.
"As with any technology, when just a few companies dominate the field, they tend to monopolize both talent and intellectual property," the report noted.
In the future of AI technology, the report continued, it may be important to debate the benefits and drawbacks of this type of consolidation and whether more competition might be needed.
Easier to use, continued democratization
Even as smaller organizations rise and fall at rapid rates, enterprise AI trends in 2019 are likely to see the continued democratization of AI tools as the technology becomes cheaper and easier to use.
"Increasingly, we're moving towards citizen data scientists," Tom Davenport, professor of information technology at Babson College and author of numerous publications on analytics in the enterprise, said at an event at AI Summit.
Arising along with the push for self-service analytics and AI tools, citizen data scientists are persons who can carry out data analytics tasks, yet may not be formally trained in data science. A few years ago, those tasks would have required that formal training, but AI and machine learning tools in augmented analytics have enabled people with IT, engineering or even business training to competently work with data.
"Machine learning tools become easier to use every year," said Seth DeLand, a trained engineer and product marketing manager at MathWorks, a Natick, Mass., vendor of mathematical computing software.
More and more, DeLand said, data science tools MathWorks sells have focused on point-and-click interfaces and integrating machine learning technology to improve ease of use.
According to DeLand, engineers, who are a large segment of MathWorks' customers, have said that MathWorks applications that help users build machine learning models, for example, are fairly easy to use and can even be used as learning tools in themselves.
The engineers, while not trained in data science, have solid backgrounds in mathematics and science, DeLand acknowledged. Still, he said, creating such models might have been left entirely to trained data scientists not too long ago.
This AI-based software is only expected to become easier to use as organizations get more access to big data, more computational power and myriad AI vendors to choose from.
Meanwhile, in a late 2018 survey on the future of AI technology, London-based consulting firm RELX Group found that many senior business leaders -- 88% of leaders polled in the survey -- think AI and machine learning technologies could help their businesses be more competitive.
Only 56% of polled organizations are using such technology now, however, the survey found, concluding that there is a desire and an opportunity for the expansion of AI tools in the business world.
AI in healthcare
While organizations in most industries are expected to start using AI-powered data tools, experts are expecting to see high adoption in healthcare and fintech in particular.
In healthcare, health systems are expected to continue to take on AI-powered tools, even as limited funding will slow the process.
Advances in deep learning, NLP and computer vision technologies will likely make electronic health records easier to create and manage and will contribute to wider use of machine-assisted, image-based anomaly detection.
Already, many healthcare providers are pursuing a machine learning, computer vision approach to automatically detect anomalies in radiology images. Providers have been reporting high levels of accuracy for years.
"As more and more data is available … and with deep learning techniques, these technologies will be beneficial" to health systems and more widely adopted, Zunaid Kazi, co-founder and CTO at AI vendor Infolytx, said.
At least in the near future, these AI tools will not replace medical staff or make key patient care decisions. That, Kazi noted, would be "a problem." Instead, they will instead "help the physicians making the decisions," he said.
And in fintech
The fintech sector, according to Lohr, has also been abuzz recently with the promise of AI technology.
In customer service, chatbots have freed up phone lines and cut costs, even while annoying some consumers. Financial planning tools, both on a consumer-level and an enterprise-level, have gotten smarter and more personalized with machine learning algorithms. Biometrics, predictive analytics and NLP tools have helped organizations crack down on financial fraud.
Lined with proven use cases, fintech vendors have started to develop and hawk AI-powered tools quickly over the last few years. This trend is expected to continue in 2019.
Regulation and ethics
With the future of AI technology likely to see AI-powered tools integrated into more and more applications and vertical markets, the AI industry will need some type of regulation, experts say. That prospect is likely.
"If this is going to be such a powerful concept … what happens if that power is centered with big companies, government?" Lohr asked.
Oversight, at least in the coming years, will revolve around fights over the automated collection and use of private data, as well as the role AI developers should, or should not, play in helping to advance military technology.
Referencing the 2018 U.S. trade war with China, Kevin Gidney, co-founder and CTO of AI-powered contract analytics vendor Seal Software, said the dispute with China is not only about goods, but also data and AI technology.
China has few data privacy regulations and has been able to use data generated from its massive population to rapidly advance AI technology. Gidney said that, to keep up with China, the U.S. will have to work with private tech companies to form better federal regulations around data and AI development -- regulations that will allow the government access to enterprise-held data.
Gidney, who said his U.K.-based Seal Software has filed patents for intelligent data management contracts to give consumers more control of their personal data, predicted that there will soon be a battle between the U.S. government and data sharing and privacy advocates.
"The one who wins?" he said, "I'm not too sure."
Meanwhile, corporate AI ethics, Davenport said, "emerges as a big issue" in 2019.
Inherent bias is likely to be addressed by AI-powered tools more as well. As machine learning algorithms advance, experts have begun to see that many machine learning models are inherently biased, as they rely on data that, whether knowingly or not, was subjectively collected and processed.
"The data being used is not representative of reality" in some cases, Gidney said, adding that, in many cases, "we have no way of knowing" if bias was willfully introduced into a model or the result of such data.
In the years ahead, developers will likely focus on creating more open models to try to adjust for these biases and become more aware of how even massive amounts of data can paint a false or distorted picture.
At the same time, the future of AI technology will likely see humans involved less in the collection of data to limit any unintentional bias in the actual data.
HR companies have been using AI for this for years now -- setting up blind algorithms to help in candidate interviewing processes to try to create a more fair hiring process, for example. That practice is expected to become more widespread.