Siri creator: Accountants need to bring AI in-house

Register now

There are major opportunities in artificial intelligence for accountants, according to the co-creator of the Siri digital assistant.

Speaking in a keynote at and the American Institute of CPAs’ Digital CPA conference this week just outside of Washington, D.C., Dag Kittlaus laid out the advantages the assembled accountants have when it comes to AI: “Wherever you have data -- and accountants have lots and lots of data – you can add value by becoming adept at figuring out how to apply machine learning to it,” he said. “Ask, ‘What do we have a lot of information about, and how can we apply AI to this?’”

“The basic definition of AI is using a computer to do something that human intellect used to do,” the AI entrepreneur explained. “They are computer systems that can do things that previously required human-level intelligence. Machine learning is a discipline within AI that needs a lot of data -- often in very narrow problem sets,” such as the specialized industry data that accountants often have access to.

The capabilities of AI go “way beyond just digital assistants,” Kittlaus said. As an example, he noted that while it was possible to use raw computing power to calculate all the possibilities in chess and develop a software program that could beat grand masters of the game, that wouldn’t work for the Japanese strategy game Go, which is significantly more complex.

Instead, researchers applied AI and machine learning techniques to teach a computer to play: “The computer beat the best Go player in the world 10 years before it was expected to,” Kittlaus said.

Next, they created a duplicate version of that AI, and matched the two programs against each other. The version that only played the other computer won 100 games to none -- it played better with no human intervention.

On a more practical level, he described using AI to identify cancers more quickly and accurately than human doctors, based on the records of millions of previous patients, or to correctly predict when the wheels of freight trains need to be replaced, without having to inspect them – again, based on previous maintenance records.

“Machine learning is taking a lot of data and gleaning insights from it,” he said. “So where there’s a bunch of data that’s consistent, using AI and machine learning you can figure out, with a lot greater granularity, what’s going to happen, and spot things early.”

Kittlaus encouraged accountants to dive in, using their domain knowledge to identify opportunities to apply AI to their clients’ problems and growth strategies – and not to pay too much attention to the doomsayers.

“This is not an area where people are going to worry about jobs,” he said. “AI isn’t about job loss – it’s about ‘task loss,’” with repetitive work going away, freeing accountants and advisors to focus on high-value tasks, client relationships, and change management.

In fact, Kittlaus suggested creating jobs around AI. “Get these competencies in house – find people who can recognize in your customer base how these tools can be applied,” he urged. “We’ll see a thousand billion-dollar companies applying machine learning and deep learning to new things.”

The China Syndrome

While Kittlaus was very optimistic about AI, another speaker at the Digital CPA conference sounded a few warning bells.

In a keynote on major geopolitical trends, Anja Manuel, a partner at international strategic consultancy RiceHadleyGates, noted that artificial intelligence is an area of significant international competition.

“Almost half of all AI funding last year went to China, and only 25 percent to the U.S.,” she said. “If there’s an AI race between the U.S. and China, it’s not at all clear who’s going to come out on top.”

And while she agreed with vice president of strategic alliances and business development Michael Cerami when he pointed out that all of the top 100 pioneers in AI are in the U.S. and Canada, and that those two countries account for 70 percent of the top 1,000, she said that there are areas that we in North America may be neglecting. “Now in the U.S., all the research and development is being done by private companies,” she warned, without the sort of government and academic work that underpinned, say, the development of the Internet. “Our basic research in AI is low, and going down.”

What’s more, there are very few rules in the field. “We are not coming up with comprehensive, enforceable standards on how to develop AI,” she said. “Ideally, you would have a sort of Bretton Woods system for AI – getting all the players in one room, so that everyone is playing by the same rules,” and not just looking out for their own national or corporate interests.

Finally, she warned that we need to be alert to flaws and biases in the data we use in AI systems, noting as an example that most cancer research in the U.S. is done on men. “So when you use that data to inform AI, it’s skewed,” she said.

For reprint and licensing requests for this article, click here.
Artificial intelligence Machine learning AICPA