Dr Ehsan's research program aims to use techniques from artificial intelligence to amplify human ability. He models and captures the dynamics of human behavior and their relationships using machine learning, computer vision, network sciences, and design interactive systems to promote equality and access in health care and education.
His research in 2012 resulted in a patent introducing the idea of a computer being a conversation coach. Building on this patent, Microsoft released 'Presenter Coach' in 2019 to be integrated into PowerPoint. In 2009, as part of a team at Disney Imagineering, Dr Ehsan introduced Autonomatronics, debuting "Otto", the first interactive figure that can see, hear and engage with a live audience.
TBS: Tell us a bit about your personal journey and how you came to the field of AI. What are your current fields of research?
Dr Ehsan: I was fortunate to travel quite extensively when I was a kid. At that time, I didn't speak English well, nor I had any familiarity with the local language. As a result, I grew in the habit of paying close attention to intonation and their facial expressions to expedite communication. I was fascinated by the similarities and contrasts in communication across languages, cultures, and geographical boundaries.
As an undergraduate student in Computer Science in 2004 in the USA, I had convinced a professor to supervise a "self-designed" course to explore my curiosity for nonverbal communication. In particular, I wondered whether we could train AI to recognize some of the prescriptive nature of communication (e.g., nodding of the head, smiling with lips, etc.). The experience ended up being so exciting that I ended up pursuing a master's degree exploring the possibility of whether AI can recognize our intentions from spoken language.
At the end of my master's degree, I had very attractive job opportunities from the major tech companies, as well as Wall Street. However, I was very obsessed with the idea that AI one day would be able to understand and recognize the full spectrum of communication. So I decided to pursue a Ph.D.
During my Ph.D., I recognized that in addition to accelerating AI, there is tremendous value in applying it meaningfully and ethically to solve some of the most pressing problems of the society. Particularly, we noticed when modeled appropriately, technology could empower individuals with special needs such as those with autism. My younger brother, who is 20 years younger than me, has autism and down-syndrome. Therefore, the narrative of using human-centered AI to positively impact the lives of individuals to promote equality and access became a mission for me.
For my Ph.D., I worked on the provocative idea of whether we can design AI, which in turn can help sharpen some of our basic human skills (e.g., doing well in an interview). Our success added more substance to the research narrative – let's not just make AI smart just because we can. Let's make AI so smart that it can, in return, make humans even smarter. The work had received a significant commercial appeal with Microsoft building on our patent to release a new software called "Presenter Coach" to be integrated into PowerPoint.
Since 2013, I have been an Assistant Professor of Computer Science at the University of Rochester where my group has focused on developing AI to improve the lives of disadvantaged, ill, disabled and other individuals who struggle with neurodegenerative disease (e.g., Parkinson's), and terminal illness (e.g., cancer). Additionally, we have developed AI that can help individuals with public speaking, job interviews, music training, negotiations, collaboration, or even to measure credibility.
TBS: What are some common misconceptions about AI?
Dr Ehsan: We seem to anthropomorphize AI. Perhaps due to the influence of Hollywood movies, we often equate smart AI with evil robots that can potentially take over humanity. We are very far away from such a future. More importantly, individuals and the relevant stakeholders in the field are united to continue to develop trustworthy AI that benefits people and the planet. Examples include, according to the G20 AI principles, augmenting human capabilities and enhancing creativity, advancing the inclusion of underrepresented populations, reducing economic, social, gender and other inequalities and protecting natural environments, thus invigorating inclusive growth, sustainable development, and well-being.
TBS: How much of a threat does AI really pose to human employment?
Dr Ehsan: Automation/AI doesn't kill jobs or create unemployment. It kills professions and creates more jobs with newer skills producing further employment opportunities.
While jobs that require manual repetitive labor may be depleted due to the proliferation of AI, there will be demand for individuals with creativity – a fundamental human skill that will be difficult for AI to automate. Our aptitude in arts, leadership qualities, and other soft skills will become relevant.
Now, let's analyze this in the context of Bangladesh. Foreign remittance from expatriates is significant revenue for Bangladesh. When the middle eastern countries start embracing AI, our supply of workforce to them will decrease dramatically.
International Labor Organization predicts that, in 2020, our neighbor, India, will have 160M people in the age group 20-24, China will have 90M and USA will have a shortage of 17M. Bangladesh, having similar demography such as India with half of its population below 25 and 65% of the population under 35, will be on the same boat.
Given the lack of opportunity and quality in higher education, a significant part of the youth in Bangladesh relies on vocational training/manual labor. For example, 4.4M people in Bangladesh work in garment factories, 80 percent of whom are women who migrated from rural areas with middle school education. This population will immediately lose their jobs as soon as automatic sewing machines and LOWRY – a lightweight robot used in fabric handling, pick and place operations, and direct sewing – are introduced by 2026.
Imagine the consequence of half of the population of a country being underemployed, frustrated, and undereducated due to AI/automation. A possible demographic dividend becomes a demographic disaster.
So what's the solution? We need to revamp our education system and introduce scalable skills-building training so that the youth is better prepared. This isn't just about a social or economic issue; it is about national security for Bangladesh.
TBS: Given Bangladesh's economic circumstances how feasible is it for the country to adopt AI in various sectors?
Dr Ehsan: Bangladesh has shown a very impressive trajectory in digitization. The question is whether to digitize first and then AI or digitize with embedded AI. Identifying the opportunities for digitization with embedded AI, when appropriate, is the high-risk, high-payoff path forward. There are guidelines in place to appropriately identify those applications by asking the following questions: Does the solution promote: 1) equality and well-being; 2) human-centered values and fairness; 3) transparency and explainability; 4) security and safety for its entire life cycle and avoids any misuse; 5) hold all private and public stakeholders accountable.
For Bangladesh, investing more resources on facial recognition technology or robotics (let's not invite the Sofia robot again!) would not be beneficial, and it opens the door for immediate misuse. However, mining aggregated data through telehealth about the health and wellness of the population would be a great example of embedded AI as we move towards digital medicine.
TBS: How should young people wanting to study and work in the field of AI go about it?
Dr Ehsan: Focus on the problem, not the technique. Bangladesh has a unique set of challenges (e.g., traffic, corruption, lack of accountability, pollution, inequality). Those challenges, when appropriately highlighted, will have significant intellectual merit and a considerable impact. As you study AI, please do not import problems/dataset from the western world just because its there. Focus on working with our local data. If the data does not exist, create it. If the data is messy, fix it. Given the large population, Bangladesh will have a huge advantage of contributing to research with its unique data, which many views as the new oil. You could be the first person to make it happen.