19:08 Run Time | November 28, 2023
Jinjun Xiong was a young computer scientist at IBM when the company’s Watson computer famously beat the top human players on “Jeopardy!”. While the rest of the world oohed and aahed, Xiong wondered if we should be using AI for a higher purpose—not to defeat humans, but to help them. Now a SUNY Empire Innovation Professor at the University at Buffalo and director of UB’s Institute for Artificial Intelligence and Data Science, Xiong is committed to using AI for the betterment of society. In this episode of Driven to Discover, Xiong talks to host Cory Nealon about the epiphany that drove his career choices; the AI projects he’s most excited about at UB, including a $20 million National Science Foundation-funded institute to use AI to transform speech-language services in schools; and why he thinks people should stop freaking out about the downsides of AI and be more energized by its possibilities.
Cory Nealon: Flashback to 2010. Jinjun Xiong is four years into his career working on artificial intelligence programs at IBM. At the time, three disasters gripped the world. A catastrophic earthquake in Haiti, the BP oil spill, and devastating volcanic eruptions in Iceland. Seeking solace, Xiong signed up for volunteer work in an impoverished area of India. Shortly after returning home, his employer, IBM, made headlines worldwide.
Speaker 2: This is Jeopardy. The IBM challenge.
Cory Nealon: Yes, Big Blue's groundbreaking Watson program achieved a historic milestone in artificial intelligence by besting the game show's greatest champions. Watson's victory signaled a new era of natural language processing, information retrieval, and other critical underpinnings of AI. While incredibly impressive, the demonstration left Xiong feeling somewhat hollow.
Jinjun Xiong: We spent all this time, all this money, all this resources, but who did we really help?
Cory Nealon: Xiong made a decision then that guides his career to this day. Whatever he is working on, his goal is to design smart, efficient AI solutions for the betterment of society. Welcome to Driven to Discover, a University at Buffalo podcast that explores what inspires today's innovators. My name is Cory Nealon, and today I will be talking with Jinjun Xiong, director of the University at Buffalo Institute for Artificial Intelligence and Data Science. The subject: AI for social good. Jinjun, welcome to the show. So if you could take us back a little over 10 years ago when Watson was wowing the world, you had a slightly different reaction.
Jinjun Xiong: That's correct. That's correct. And that's how I started to coin the terms I call it the flashy AI versus the earthly AI. On the one hand, I see the world needs so many solutions, need so many people's help, but on the other hand, in the U.S., we are developing advanced technology at that time wasn't right? And it's artificial intelligence and able to understand not very unstructured information, and able to compete with a human in this kind of Jeopardy show, and AI is able to win. And of course, when IBM designed the Watson game (it) has its own purpose to demonstrate what technology can do. So that is totally good intention, good objective. But on the other hand, yes, I always believe technology should not really just to demonstrate how technology can beat a human, but actually how technology can really help a human extend the human's capabilities.
Cory Nealon: And that's sort of the idea behind that earthly AI. What can we do in a practical, common sense way to help people in their various needs?
Jinjun Xiong: That’s correct. That’s correct, yes.
Cory Nealon: Let’s dig a little deeper into that. You were at IBM and you were sort of making this pivot into that type of work already, right? What kind of projects were you doing there?
Jinjun Xiong: IBM had this big Smarter Planet initiative that was a little bit before this IBM’s Watson Days. And since my background is in the electric engineering and the computer engineering, so I know the electrons very well and I thought, okay, how can I leverage my background and to solve the sustainability issues? We established a consortium called the IBM Smart Energy Research Institute, where we work with a number of largest utility companies and jointly identify what are the present needs in the smart energy industry or in the power grid industry where technology or bigger data can help.
Cory Nealon: You’re talking about the more efficient use of energy as it travels through the grid. You’re talking about being able to better predict storms, outages, the demands that come and go with spikes in energy. And then you went to work with IBM into higher education, correct?
Jinjun Xiong: Yes. That’s kind of the transition period. When IBM won the Jeopardy in the Watson in 2011 and in a couple of years, IBM decided to kind of take advantage of that technology and to focus on a number of industries. And one of the industries IBM felt the technology can help in fact is education. And that opened up my eyes exactly that’s what I was looking for and that’s how I started to work with the University of Illinois and try to explore how can we establish a program so we can address these educational problems systematically. We established this IBM-Illinois Center for Quantum Computing system research. The essence is to use AI technology to provide an individualized, customized education for individual students. In that particular use case, it’s about the experiential learning. And that’s how I started really, really working into this kind of educational space. And also in academic-industry partnership kind of setting.
Cory Nealon: It sort of gave you a sense of what’s possible in higher ed and in education in general.
Jinjun Xiong: Oh, yes, indeed. Even though I have been working with a lot of faculty members in the past as well, but these were as I experienced with my collaborators at Illinois, and it really opened up my eyes to see what is possible, right? First, the talents in terms of students, and the second, there's so many opportunities to explore different ideas. And even though I worked at IBM research as a researcher, we always try new ideas, but in the company setting, you have to be kind of aligned with some of the business needs. So that's a constraint some of your exploration opportunities. But in the universities, students being young, they're still learning and they're not afraid of trying things, they're not afraid to fail. So just give a lot of more opportunities to explore.
Cory Nealon: Well, that actually takes us right into your work here at UB.
Jinjun Xiong: Oh yes. I really feel like I'm lucky enough, fortunate enough to be able to join UB at the right time. It takes a similar my experience in this space and the startup a few explorations. The very first one is this AI for [Exceptional] Education and happened to be the National Science Foundation have the third rung of this call for proposals and they have two threads, which is how to use AI to help adult learning. Another one is how to use AI to help children with disabilities. So I decided to spend time to understand this domain and the issues and to see how AI can help. And that led to the UB's winning this award from National Science Foundation and to establish this National AI Institute For Exceptional Education, basically how to use AI technology to help children with disabilities.
Cory Nealon: And we're talking specifically about children with speech, language processing challenges. A little bit of background on that is that there's not enough teachers out there in that specific field. And what you're talking about, correct me if I'm wrong, but you're talking about putting AI systems in the classroom, which would sort of act as early identification that a student might be having a problem in this area, correct?
Jinjun Xiong: That's correct. That's correct. And so by talking to many the special education teachers and also a lot of parents who have the special education kids at home and early on I identified two major challenges in this area. The first one is a lot of parents complained their children being identified late for their needs for the special needs. The second one is the number of caseload for the special education teachers just too much. So how can we use AI to develop some solutions to let's say, put into the childcare centers so we can identify children who are in need of services early? The second solution are some solutions in the public school systems for any children who are in need of the services. Can we help with the special education teachers? So to help them to monitoring the progress, provide customized dashboard for them, and then also to recommending the right intervention at the right time for them and also help them to develop more creative or more evidence-based new interventions.
Cory Nealon: So you have that customized learning platform that comes in after the children are screened. The AI program, the chatbot, or whatever form it takes, is interacting with the child learning and providing personalized lessons. And that also enables the traditional instructor to do that one-on-one work with students, which they aren't able to do right now because they're typically working with children in groups.
Jinjun Xiong: That's correct. That's correct. Yeah. That's because of the caseload is so big. Just give you some numbers. There's about 7 million children with the special education needs in the public school systems. Let's just assume half of the need is speech and language services that's about 3.5 million children. But in the public school systems right now, we only have less than 61,000 speech and language pathologists. So that's what I mean you can see the number of caseload just divided by its simple back-of-the-envelope compilations. You can see it's impossible for every SLPs to help with the child individually.
Cory Nealon: And the technology underlying this for this personalized learning, I mean theoretically it could be applied to different disabilities to not even children with disabilities, correct?
Jinjun Xiong: Oh, that's correct. That's correct. The same technologies can be applied to other use cases for helping children with other type of disabilities and even children without disabilities too, in terms of how do you monitor in children's progression in terms of learning and how do you make the right recommendations, and how to come with better ways to learn.
Cory Nealon: So it's not just education though that you're working on here at UB. I've noticed that you've working one project includes indoor farming, for lack of a better term. I mean you're going to have a greenhouse set up in your lab. You're going to have sophisticated sensors monitoring the growth of plants.
Jinjun Xiong: This is kind of related back to my early work of the sustainability. I realized there's another challenges facing this indoor growing business, and that is how do you effectively identify the plant of disease and also how do you use the technologies to enable the growing more, let's say more productively, which means that you need to monitoring the health status of the indoor plants and how do you adjust the lighting conditions so you can stimulate the plants to grow better and healthy.
Cory Nealon: I could see it this type of technology being more important with climate change, and sort of the unpredictability of the weather and the reliability of the seasons. So I mean that's really incredible. I'm looking forward to see how that works out.
Jinjun Xiong: Right. Right. I definitely hope so too.
Cory Nealon: So as director of the University at Buffalo's Institute for Artificial Intelligence and Data Science, you also are sort of charged with finding other fields to apply artificial intelligence to.
Jinjun Xiong: Well, I see tremendous opportunities. My goal is to really try to tap into the deep expertise in UB across different domains and work with the colleagues here to see how AI and data science or technology in general computing in general can help, can enable them to do their work more effectively and better. I started a few initiatives. So one of the initiatives, I call this kind of open door office hours. So I invited the colleagues across the campus. It doesn't matter which areas you are coming from, if you see any challenges in your domain, come talk to me to bring out the issues to me. And then I will organize brainstorming sessions and then I inviting people who are in the related space in terms of technologies and coming together to brainstorming together to see how technologists can help our domain experts to solve that challenge.
Cory Nealon: Yeah, I do see this sort of holistic approach to things here. You want to have your humanities experts and scholars involved because this sort of gets into what else I want to talk about is just technology doesn't exist within this defined world. Once you create it and give it to the world, it has unintended uses, it goes places you didn't anticipate it. People are concerned, how do we do this in a responsible way? How do we make sure that AI is trustworthy, that it is helpful? It seems like a monumental task.
Jinjun Xiong: It is. It is. I mean, that's a great question. We do need to have this debate between the private and the public and even everyone to in this conversation to think about these issues. But I also want to caution and do not because of the fear of technology, and then we prevent technology from advancing. And the AI has tremendous potential, but once you put into the use case context, you can see a lot of the potential. It's very harder to realize there's still tremendous gap to realize the potential to solve the real use cases and that's what I mean. I still advocate. We still need to have continuous investment into the AI research and to really make us realize AI's potentials and to solve those societal challenges. And of course, at the same time, we're still keeping an eye on the side effects of AI technology and make sure that we know how to regulate and how to mitigate the risk.
Cory Nealon: One of those side effects, or potential side effects, that I hear a lot about is the potential of loss of jobs with things are automated.
Jinjun Xiong: How do you adapt to technology change? So my message is yes, even though the concerns are legitimate, but technology is inevitable. The advancement is inevitable. Like many of these industrial revolutions too, you always can [be] concerned about the loss of job, but I would say it's not really loss of a job. It's really open up new opportunities for new jobs and how you can adapt to this change. And back to my earthly AI things, I just felt there are so many tremendous societal needs. There would be no worry about lack of jobs. There's so many jobs just because of we do not have enough resources. We do not have the right technologies. We cannot provide those solutions. So I mean, for me, I'm still more or less on the positive side and let technology to advance. Of course, we need more people to know how to utilize the technology to solve the earthly AI needs.
Cory Nealon: So there's other concerns, and I don't want to spend a lot of time on them, but privacy, misinformation, just the ability of using AI programs to cheat or plagiarize. But the last thing I did want to touch about with some level of depth was this idea, right or wrong, that can machines get to the point where they sort of take over the world. I know it sounds sci-fi, but it's a concern again that people have.
Jinjun Xiong: So this I would typically blame Hollywood, they cherry-pick, and also some of the media too, right? The cherry pickers, some of the things, yes. And if you cherry-pick some of the things to talk about, yes, technology it seem like it would be so advanced. The moment you apply to solve real use cases, and you'll realize technology is not there yet. So I'm not so concerned, I think as a researcher and a scientist here, and I still want the people to spend more time to really try to adjust the technology gaps.
Cory Nealon: So you're involved with the UB AI chat series, which is an attempt to sort of bridge the gap between public knowledge about AI and the work that's being done in the labs here at the university. Could you tell us just a little bit more about what the goals of that series is and what you're hoping to accomplish?
Jinjun Xiong: The president's office started this very good initiative. So why don't we bring UB's expertise in AI and data sciences to the communities, and have this direct dialogue between technologists and the communities, answer the community's questions, address their concerns directly. Let them know, yes, AI is here, right? AI has a lot of potential, but there's still so many earthly AI needs that technology needs to help and technology can help. I think having this great public channel to communicate, have this direct communication is a greater initiative. And I think I really have to sing a song, send a praise for our president's office [being] able to support this initiative so we can go out into the communities, have this direct dialogue.
Cory Nealon: Well, Jinjun, thank you for being here. It's been wonderful speaking with you.
Jinjun Xiong: Oh, thank you so much, Cory. Thank you for having me.