Tomorrow’s AI Will Reason Like Humans, IBM Watson Developer Predicts

Image of David Nahamoo
DAVID NAHAMOO

By: Joanna Goodrich

When David Nahamoo was a high school student in Iran, he wanted to pursue a career in mathematics or physics. But after talking over career options with his friends, he says, he was "pointed in the direction of a good career in Iran" and instead decided to become an electrical engineer. Today the IEEE Life Fellow is CTO of Pryon, a startup in Raleigh, N.C., that is developing a natural-language-processing AI system for businesses. The company's programs aim to make companies more productive, reducing costs and eliminating inefficiencies.

Nahamoo is an expert in speech and language technologies. He spent almost 35 years at IBM Research in Yorktown Heights, N.Y., developing innovative AI technologies such as Watson. The supercomputer won a game on Jeopardy! in 2011 against two of the U.S. TV show's most successful contestants. Watson answers questions using advanced natural language processing, information retrieval, knowledge representation, automated reasoning, and machine learning technologies.

"I love taking a problem, sitting down, and figuring out how I can solve it," Nahamoo says. "And when I do [solve the problem], I get a rush of joy."

He received this year's IEEE James L. Flanagan Speech and Audio Processing Award "for contributions to and leadership in research and deployment of spoken-language technologies." The speech and audio-processing award is sponsored by the IEEE Signal Processing Society and Mitsubishi Electric Research Labs.

Nahamoo worked with Flanagan throughout his career. Flanagan, who died in 2015, worked at Bell Labs in Holmdel, N.J., for 33 years before he joined Rutgers University in New Brunswick, N.J., as vice president for research. He was a pioneer in the field of acoustics and provided the technical foundation for speech recognition, teleconferencing, MP3 music files, and the efficient digital transmission of human conversation.

"I was very happy and thankful that I got an award that honors a person I knew very well," Nahamoo says.

CHOOSING A DIFFERENT PATH

Nahamoo grew up in Hamadan, and when he was 12 years old, he moved to Tehran. After graduating from high school at the top of his class, he earned a bachelor's degree in electrical engineering in 1975 from the University of Tehran, then a master's degree in EE from Imperial College London in 1976. He left England to pursue a doctorate in electrical, electronics, and communications engineering at Purdue University, in West Lafayette, Ind. He conducted his doctoral thesis research at Purdue in ultrasound diffraction imaging. He developed a technique that allowed for tomographic medical imaging using diffracted projections—images created when waves bend around objects.

After graduating in 1982, he switched his interests and joined the continuous speech recognition team at IBM Research.

"I just fell in love [with the group] because I thought that they were solving problems that appeared to be impossible to solve," he says. "And I love that challenge of impossibility."

Looking back, he says, he can divide his career at IBM into three parts. For the first 10 years, he was a "pure technologist" who solely worked to improve the technology being developed. During his second decade, he became more interested in the delivery of speech recognition products to the market. For the last 10 years, as CTO of speech technology, his interests centered more on the strategic impact that spoken language technologies have on businesses.

WHAT IS: IBM WATSON

Nahamoo was dedicated to expanding the capabilities of speech recognition technologies. He wanted to develop a machine that could interact like humans with other humans.

That idea led to a one-year study by IBM that gave birth to the Watson Group, Nahamoo says. The study examined market needs for cognitive computing technologies and business opportunities surrounding automating human sensory, language, learning, and reasoning capabilities.

The supercomputer's data analytics processor analyzed human speech for meaning and syntax so that it could answer questions posed to the machine, similar to how Amazon's Alexa and Apple's Siri now work.

Nahamoo says the Watson project was all about making progress on programming machines with the cognitive abilities of humans. The supercomputer was the first step to creating an AI machine that people could interact with as if they were speaking to another person, he says.

One day, he says, AI machines will be able to understand physical cues such as head nodding and posture changes, as well as mimic human emotions. That would enable machines to interact more closely and could let them form connections with humans, he says.

Building such a machine, Nahamoo says, is "my interest, my love."

SWITCHING GEARS

Nahamoo now has an opportunity to take his work forward. In 2018 Igor Jablokov, founder and CEO of Pryon, offered Nahamoo the CTO position. The two had worked together at IBM in the early 2000s, when Jablokov led the development of IBM Watson Assistant—a forerunner to IBM Watson. The two collaborated again in the late 2000s, when Jablokov, founder and CEO of tech company Yap, licensed IBM's Attila speech-recognition engine.

Nahamoo says he took the offer because he wanted to go back to working at the "cross section of deep technology innovation and business impact."

Pryon has about 30 employees and aims to help companies use AI to improve their operations. One of its products, Answer, is a question-answering platform that companies can add to their AI assistants, chatbots, and help desks.

An employee can ask via text or voice how many vacation days she has available, for example. The platform organizes, reads, and searches a company's applications and documents to find the answer. Unlike other AI assistants and chatbots that can respond only to a limited set of requests, Pryon attempts to retrieve responses to any company-related questions. The system also collects user feedback and monitors usage in order to improve the quality of the answers.

"We are trying to build the best question-answering system out there," Nahamoo says. "One that can filter content from different kinds of files—from PDFs, HTML, and Microsoft Word or Google documents to PowerPoint presentations."

The company has one customer using Answers and others that are interested in implementing the system, he says.

IEEE APPRECIATION

Nahamoo joined IEEE in 1978 as a student. He wanted to be able to read the organization's publications, take part in its activities, and connect with other engineers, he says.

"I appreciate the insight and greatness of those who started IEEE and of those who have helped the [engineering] community," Nahamoo says. "I believe our community would have been a lot less effective and capable of making the impact that it has made without IEEE."

This article originally appeared in IEEE Spectrum on 3 September 2021.