Printer Friendly

Higher Learning: AI, ML, and Speech Tech in Academia: Academics are at the forefront of the biggest AI changes in the industry.

Google, Amazon, Facebook, and other big tech players in the industry continue to expand the boundaries of artificial intelligence (AI) and speech technology as well as reap the profits earned from related products and services. But while these private sector giants--and the scientists, researchers, and innovators who work for them--continue to grab headlines and market share, arguably the most cutting-edge and important discoveries in the field are coming from the hallowed halls of academia.

That's because many universities, colleges, and other institutions of higher learning have ramped up their curricula related to AI, machine learning (ML), and speech technology in recent years and bolstered their programs and departments with impressive faculty and resources. The reason is simple: growing demand from students seeking to earn a degree in these tech fields and fill the strong demand for high-paying high-tech jobs. Armed with bright minds, smart tools, and a determination to innovate, these educational establishments are pushing technology forward faster thanks to groundbreaking research, testing, and inventions.

But even as the learning opportunities at schools have increased, so have the challenges--including recruitment and retention of skilled teachers, lack of funding, inability to meet enrollment demand, and keeping up with the pace of progress.

A Growing Presence on Campus

AI first came to the classroom in 1956, when Dartmouth offered a Summer Research Project on Artificial Intelligence workshop. The following decades spawned many related discoveries by researchers at universities, labs, and private companies--research often funded by government agencies like DARPA.

However, due to limited computing capacity and other difficulties, the hopes and promises of AI proved elusive to academics until relatively recently. Thanks to 21st-century advancements in theory, algorithms, and computing power that have fueled interest in AI and ML among faculty and students alike, the dreams of scientists past are coming to life.

"Today, AI is being taught to more than just engineering and computer science students. Students in the liberal arts and business are getting exposure to AI," says Joseph Wilck, faculty director of business analytics in the Mason School of Business at William & Mary. "Basically, it has morphed from a strictly theoretical domain and curriculum to a more applied and applications-driven curriculum."

Additionally, institutions now have many more courses that deal with "how to" rather than just "what is," according to Nava Shaked, head of the department of multidisciplinary studies at Israel's Holon Institute of Technology and adjunct professor at the Graduate School and University Center of the City University of New York.

"Advancements in algorithms such as deep neural networks," she says, "as well as conversational systems like bots and conversational virtual agents have created new interests and new ecosystems to explore." New courses are also part of the equation. For example, AI classes offered at HIT include "Autonomous Cars and IOT," "ML Algorithms and Deep NN Analytics," "Wearable Computing," and "Social Robots."

"Excellent progress is being made in universities in the area of new learning algorithms, robustness of such algorithms--especially for mission-critical applications--and AI hardware architecture and circuits," says Kaushik Roy, professor and director of Purdue University's Center for Brain-Inspired Computing.

According to research by Diffbot, the top 10 universities producing graduates skilled in machine learning (more than 45,000) are Stanford; the University of California, Berkeley; Carnegie Mellon; Georgia Tech; MIT; University of Illinois; University of Southern California; University of Washington; University of Michigan; and Columbia University.

But AI classes aren't solely for graduate students any longer, and they're no longer strictly offered at schools specializing in high-tech, notes Houwei Cao, assistant professor of computer science at New York Institute of Technology.

"Nowadays, more and more schools start offering introductory-level AI and machine learning course for undergraduate students," says Cao.

Consider that last year Carnegie Mellon University and the Milwaukee School of Engineering became the first two schools in America to offer an undergraduate degree in AI. There are also plenty of online AI courses accessible to all, including classes offered by Coursera and MIT.

Even schools and departments that don't necessarily specialize in technology are now offering AI/ML learning opportunities. Case in point: Fordham Law School's Center on Law and Information Policy launched the Intellectual Property Artificial Intelligence & Blockchain Project, directed by Shlomit Yanisky Ravid, which conducts research and offers courses, seminars, and conferences that address the discourse regarding AI and future policymaking from the legal regime perspective; it also produces works of art, like jazz songs entirely created by AI, to further study intellectual property law.

Speech Tech Advancements via Academia

Still, it's the technical-engineering schools--where AI and ML systems are being developed in labs--that have fostered the most innovation, particularly when it comes to speech tech. "Universities like these have always been leaders in developing cutting-edge technology. Great minds and the search for innovations, promotion, and fame as well as other factors have played an important role in the development of AI, similar to the role academic institutions play in developing other areas such as biomedical and computer science," says Ravid, a visiting professor of law at Fordham University Law School in addition to directing the school's AI-blockchain project.

For proof, consider the following speech technology breakthroughs made at institutions with more established and renowned AI/ML faculty and programs:

* Columbia University neuroengineers converting brain waves into recognizable verbal speech using AI and speech synthesizers (see sidebar);

* an AI-based system at the University of Vermont that can recognize signs of depression and anxiety in young children via their speech patterns (see sidebar);

* turning brain signals into natural speech with the help of a brain-machine interface created by neuroscientists at the University of California, San Francisco;

* the development of Kaspar, a talking robot that helps autistic children hone crucial social skills, by the University of Hertfordshire, in the United Kingdom;

* the creation of synthesized voices that sound more natural, aiding patients who use communication aids, thanks to the University of Edinburgh's Centre for Speech Technology Research, also in the U.K.;

* research from Washington State University which shows, with the help of automatic speech recognition technology, that young children, not their moms, initiate conversation;

* safeguarding computer speech recognition from malicious messages concealed in speech, courtesy of research at the University of Illinois, Urbana-Champaign; and

* Rochester Institute of Technology deep learning research that has created an automatic speech recognition system to help preserve the language of the Seneca Indian Nation.

Anima Anandkumar, Bren professor at Caltech's CMS department and director of machine learning research at NVIDIA, says AI breakthroughs like these are being accomplished at universities, colleges, and institutions of higher learning through a combination of academic exploration funded by grants and industry collaborations.

"For example, NVIDIA has AI labs in Seattle and Toronto that partner with many universities like Stanford, University of California [at] Berkeley, Caltech, University of Washington, and the University of Toronto. These are great examples of the collaboration that happens between academia and industry to accelerate breakthroughs with state-of-the-art GPUs from NVIDIA," says Anandkumar.

These alliances between companies and colleges create a win-win synergy.

"Academia develops the foundations that allow industry to adapt and apply at scale," Anandkumar adds. "Collaborations between academia and industry help speed up innovations and close gaps that allows both to make new discoveries at a faster pace while at the same time adapting the learnings to industries such as healthcare, retail, finance, automotive, and more."

Meeting the Challenges

It isn't easy to serve as ground zero for tomorrow's hot technology, however. Institutions large and small face plenty of obstacles on the path to providing proper instruction. "The prevalence of AI and machine learning has increased dramatically in industry in the last few years, which has created a shortage of qualified talent in the field," says Cao. "In order to address this problem, more and more universities are trying to develop new curricula and new courses in artificial intelligence and machine learning."

Kristian Simsarian, owner of Collective Creativity and professor at California College of the Arts in San Francisco, says it's also hard for universities to continue cutting-edge research in ML "because companies are offering incredible salaries and bounties for professors and leaders to join industry."

In addition to recruitment and retention of highly valued human resources, "you need faculty focused on multiple areas of AI--such as vision, natural language processing, applications, robotics, and learning theory," says Gabriel Bianconi, founder of Scalar Research and a former researcher at Stanford's AI Lab. "It's also difficult meeting student demand. Some classes at Stanford have around 1,000 students enrolled."

Introducing inexperienced students to AI/ML can be particularly tricky for universities aiming to offer undergraduate programs.

"Most of the students start from zero and have no background, which is required to reach a high level of academic research in AI. That's why top graduate students often benefit the most," Shaked notes. "It's also challenging to apply AI teaching in non-science and engineering programs and classes, such as design and industrial management."

Another inherent problem? Staying relevant and not letting your AI/ML courses become outdated, as the technologies and theories can evolve quickly.

"Even if schools have offered a miniscule number of courses about AI and ML for years, these past courses were different due to significant progress achieved in research in programming and its results," Ravid says.

Wilck sees the investment costs for starting a program as the highest hurdle to clear. "Faculty and staff, buildings and labs, and equipment and software are expensive. At some institutions, money spent on a new program means that other programs are left with smaller budgets," says Wilck.

Simsarian agrees. "Academics need a minimum of $2 million a year to run a lab with graduate students and equipment," he says.

Teaching Will Change with the Technology

Looking ahead, many are optimistic and excited about the potential for AI/ML's growth in academia. "Universities will play a pivotal role in advancing AI primarily from the perspective of developing the theoretical underpinnings and explaining the decisions of the current AI systems, and improving their robustness against adversarial attacks," Roy says. "These institutions will continue to offer specialized AI programs to cope with the increasing demand for such courses both from the students and industries alike."

Bianconi seconds that assessment. "Given the growing demand from industry and students, universities are going to continue to increase their AI offerings," Bianconi says. "We'll see more people coming to school for advanced degrees as well as people retraining on the job."

But, Bianconi predicts, most of the academic interest in the near term "will be on vision and natural language processing for text, not as much on speech."

While an increasing number of schools may offer introductory and general classes on AI/ML, Wilck believes more universities "are going to try to find their niche and specialize. And I think we may also see degree programs that are created that merge AI into an already existing application area. Many students will find the most value is not only becoming an expert in AI, but also having a foundation in another field, such as business, engineering, or public policy."

In addition, the debate over ethics and AI will increasingly be played out on college campuses. Shaked says, "Academia is a good place to raise, discuss, and educate on the importance of ethical questions, which must be addressed as more AI technologies become part of our lives."

Erik J. Martin is a Chicago area-based freelance writer and public relations expert whose articles have been featured in EContent Magazine, Reader's Digest, The Chicago Tribune, Los Angeles Times, and other publications. He often writes on topics related to technology, real estate, business and retailing, healthcare, insurance, and entertainment.

COPYRIGHT 2019 Information Today, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Martin, Erik J.
Publication:Speech Technology Magazine
Date:Jun 22, 2019
Previous Article:Chatbot Development: YOUR GUIDE TO GETTING IT RIGHT: The race is on to include chatbots in marketing and CRM efforts, but many companies still aren't...
Next Article:Graduating with Speech Tech Honors.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters