How AI can help humans, not replace them

If you’re using artificial intelligence just to do something faster, you’re not harnessing its true potential.

“The real power is unlocked when used hand-in-hand with human skills and uniquely human talents,” said Michelle Sipics, senior editor at Accenture Labs and the panel moderator of an AI-and-machine-learning-focused session at this year’s Introduced by Technical.ly conference during Philly Tech Week 2019 presented by Comcast.

While you may have heard of Accenture, many aren’t familiar with its labs. Located in seven research hubs around the world, Accenture Labs is able to incubate and prototype rapidly with the latest tech, while drawing from the massive resources of Accenture as a whole.

In Accenture Labs’ yearly trend forecasting report, Technology Vision, AI featured heavily as one of the four pillars of the post digital era: distributed ledgers, artificial intelligence, extended reality and quantum computing — or DARQ. For many, these technologies are a complete mystery, and the assumption is that they are here to take people’s jobs.

“AI creates features, not solutions,” rebutted panelist Kris Skrinak, global machine learning segment lead at Amazon Web Services. As apparent from the ever-growing services list from AWS, there are possibilities now that simply weren’t an option before.

Brian Orrell, CTO of Pariveda Solutions, discussed a case where a hospital saw a spike in deadly CLABSI infections, carrying a 20% mortality rate for children. The hospital had gathered data and over six months of analysis, discovered the underlying cause. But its team wanted to see if help from AI could have sped up the process. With the same data, Pariveda Solutions arrived at the conclusion in four weeks. The next step is developing processes to gather data in real time, create predictive models and prevent infections in the first place.

Steve Zemanick, director of consulting services at CGI, discussed the vast amounts of unstructured data that companies are responsible for managing. Imagine an acquisition in which a company has inherited terabytes of data; AI can process that in accordance with the company’s retention policy in hours or days. Humans would likely expire of boredom one millionth of the way through.

But as AI penetrates more and more sensitive areas of our lives, there are real pitfalls to avoid. AI can seem like an impartial decision maker, but humans are the ones who choose the data that it will learn from. Biased data leads to biased AI.

There have already been plenty of AI mishaps, from chatbots turning racist after being on Twitter a few hours, to learning sexism from images. If we’re to trust AI, we need to know how it learns and reaches conclusions.

“Explainability goes a long way toward building the kind of trust you need to show people why this can be valuable,” said Sipics.

For example, she said, a financial institution might want to build an AI system to recommend whether a mortgage application should be approved. If the institution receives 10,000 applications a day, that can save an awful lot of time, but if they can’t explain why an application was rejected, the customer isn’t likely to accept it. What’s more, they’re missing an opportunity to not only explain why they were rejected, but help the potential customer understand what they need to do to be approved in the future.

As Sipics has learned with her work at Accenture Labs, the exciting applications of AI aren’t optimizing what’s done today. However, using AI to solve these short-term goals is an important part of the process.

“It gives people an opportunity to experiment with the technology, get comfortable with what it can do and establish the kind of trust that you need,” she said, “so that down the road, you’re ready for those even larger, transformative opportunities.”

Accenture is a sponsor of Philly Tech Week 2019 presented by Comcast and reviewed this article before publication.