Artificial Intelligence, According to These 5 Experts

By Scott Simone, Contributor

Artificial intelligence (AI) manifests itself in a multitude of ways, from online chatbots to your inbox’s spam folder. (Yes, the tool you use to send unsolicited newsletters to the trash is, in fact, an everyday form of AI.)

Yet regardless of where your baseline lies, we all have a bit of learning to do when it comes to understanding AI. To take a deeper look at the ever-evolving technology, we spoke to five top experts in the field to get their input on what makes AI, well, AI.

This is AI according to those who study, tinker with, and implement artificial intelligence into our daily lives.

It All Adds Up

According to Jana Eggers, the CEO of Nara Logics, an artificial intelligence company focused on turning big data into smart actions, “AI is just maths.” The plural on “maths” is important to note, Eggers said, because the multiple computations that are the basis of AI are also where the technology faces limitations.

“While we have many problems that can be solved with current forms of AI, there is still a huge realm of problems that can’t be calculated or even represented with current maths,” Eggers remarked. For her, we’re just getting started when it comes to understanding the relationship between mathematical knowledge and AI’s capabilities.

“We are in the cave-painting days of communication and collaboration between us and the AIs that are developing.”

It’s a Software Game

Despite all the buzz, AI is just software. “There’s no bright line separating AI software from any other kind of computer software,” Michael Littman, a computer science professor at Brown University, explained. “It is very common for problems that are considered AI problems to cease to be considered AI once they are well solved and widely deployed.”

This result, as Littman described it, is often regarded as the “AI Effect.” To exemplify just how common the AI Effect is, Littman cited another AI specialist, Kevin Kelly:

In the past, we would have said only a super-intelligent AI could drive a car, or beat a human at Jeopardy! or chess. But once AI did each of those things, we considered that achievement obviously mechanical and hardly worth the label of true intelligence. Every success in AI redefines it.

“We are in the cave-painting days of communication and collaboration between us and the AIs that are developing.”
— Jane Eggers, CEO of Nara Logics

China Aims to Lead

For Toby Walsh, a professor of artificial intelligence at the University of New South Wales, China had its coming-of-age AI moment with AlphaGo. The AI system, which plays the board game Go, beat the Korean champion Lee Sedol in 2016 and the Chinese prodigy Ke Jie earlier this year.

And today, China is investing a great deal in furthering this type of technology. “Few in businesses appreciate China,” Walsh said. “But the Chinese government has put AI at the center of its economic plan, with technology giants like Alibaba and Tencent investing billions in AI.”

In 2017, Chinese scientists dominate the leading AI conferences, and China aims to become number one in the world in AI in the next decade. “Given its natural advantages, like the largest smartphone market on the planet, a rapid adoption of electronic money, and a very relaxed attitude to privacy,” Walsh predicted, “they are likely to succeed.”

It’s Constantly Learning

The potential of AI lies in its ability to learn, and it’s learning from humans.

“Most people assume that historical training data [past data used to program an AI system’s functions] is all you need,” Mikhail Naumov, co-founder, president, and CSO of DigitalGenius—a venture-backed artificial intelligence company based in London and San Francisco—said. “In reality, practical applications of deep learning and AI benefit just as much from ongoing human-use data as the historical data.”

For Naumov, you need both—historical data to build the baseline, and ongoing user interaction data—to drive continuous learning for the AI. In other words, the more human usage a particular AI system receives, the stronger and more useful it will become.

“Simply put, humans should be focused on teaching machines, so that machines can focus on executing against jobs that are too big for humans to process.”
— J.J. Kardwell, CEO/co-founder of EverString

It Needs a Human Touch

“Humans still have—and will for a long time—a massive advantage over AI when it comes to being right nearly all the time about small decisions,” J.J. Kardwell, CEO and co-founder of EverString, a predictive sales and marketing analytics platform, said. “The full power of AI is best realized when paired with around one to two percent of human effort.”

In the right partnership between humans and AI, Kardwell explained, humans can accelerate machine learning by curating and labeling training data. This collaboration helps machines to ingest, learn from, and scale decision making.

For Kardwell, a human-in-the-loop (HITL) approach to machine learning currently yields the most powerful results in the most efficient manner. It’s the story of the human and the robot working in harmony.

“Simply put, humans should be focused on teaching machines, so that machines can focus on executing against jobs that are too big for humans to process.”