Savor this moment — these coming weeks, the next few months, the year or two that will follow. These are the last moments before artificial intelligence (AI) really starts to have a massive impact on one’s daily life, before the world really turns to AI.
That is not meant to be mournful, or an expression of grief for a way of life — a non-AI way of life — that is on its way out. Rather, it’s to tell how big the AI revolution will be, and how much it promises to change as machines become better at learning and solving all types of problems, including those in the realm of payments and commerce.
The future of artificial intelligence was the subject of the latest edition of the PYMNTS Masterclass series. Karen Webster and Sudhir Jha, Mastercard senior vice president and head of Brighterion, spoke about where AI stands as a new decade starts, where it is going, and what the challenges and opportunities will be along the way.
As Jha explained, infusions of funding and increases in computing power are leading to significant AI expansion. Not only that, but the thought has settled among business leaders and others that artificial intelligence is finally here, and is not some fleeting technology.
“This is not the same thing that happened 30, 40 years ago when AI was a big hype and it never kind of delivered the promise,” he told Webster during the Masterclass discussion. “And not only that, it is also going to be the fundamental, essential technology to differentiate going forward. … And I think that, in the last few years, that has been crystal clear — when I talk to executives, it is very, very imperative.”
ROI Becoming Clear
The last decade has shown the promise of AI, Jha noted.
“If you think about Netflix and Google, [as well as] Facebook and Amazon, without AI, there was no way they could scale in that short amount of time,” he said, moving on to offer a specific example to which most consumers can relate. “Netflix would not be able to serve millions and millions of titles, and allow me to browse through that. If I had to browse that one by one, I would never be able to watch it, and never be able to find the shows that I want to look at.”
Such efforts have demonstrated the return on investment (ROI) for artificial intelligence, and have helped to move AI from a Sci-Fi concept to a mainstream consumer tool.
In addition, deep learning — basically, machines analyzing data to find their own patterns and come to their own conclusions — has advanced at a rapid pace, he noted. Proof of that comes from facial recognition and other technologies, where accuracy rates have quickly grown higher.
Even so, the challenge of bias in AI remains a big potential problem — a machine’s intelligence is only as good as the data that fuels it, and humans still have a vital involvement in that process. Addressing that issue in the conversation with Webster, Jha was optimistic that the bias problem would be solved, or at least dealt with over time, assuming the humans behind the machines understand what they need to do.
“Any training has to be done in a way that it [uses] both unbiased data and unbiased techniques to train the people or the model,” he said. “And once you do that, AI models would not be biased. And so, while it is true that AI models can get biased, I don’t think you can blame the models [themselves] for that. It is the techniques and all that stuff.”
However, getting past bias is never going to be easy — which means AI-powered systems, perhaps, will require close monitoring.
“The issue is that the AI models are so complex, and so hard to read, that you don’t know if [they are] biased or not,” he said. “So, it is not easy to just look at the result and say, ‘Hey, this is a bias result.’ So, there is a lot of work that is going on.”
He offered an example of what kind of work that might be.
“If the model can explain and say, ‘This [is the] decision I took because [of] the other 17 things I looked at,’ and you can see clearly that three of those are sort of gender and location, demographic or whatever, which are not [things] that you want the decision to be based on,” Jha said, “you can sort of go and fix that.”
There is also always the potential for misuse of artificial intelligence, including when it comes to privacy issues, and in such areas as healthcare, where people store their most sensitive personal data.
“Unless you’re ahead of that game, and ensure that the misuse doesn’t happen by ensuring the regulation of that,” he said, “it is going to definitely [be troublesome] in [the] future for all of us.”
“And that would be really unfortunate because the gains are amazing,” Webster said.
Still, the future is looking more likely to be a place where artificial intelligence plays a big role in daily life, especially in commerce and payments. That holds true under the rules of current computing — that is, before the advent of quantum computers, he noted.
“I think that there’s enough computing power available in the classic computing — there’s enough data available right now, and there is enough technology on the AI side,” Jha said. “I think that AI itself will [evolve over] many decades into many, many different areas.”