(UPBEAT ACOUSTIC GUITAR COUNTRY MUSIC PLAYS) MATT DOZIER: Hello everyone, and welcome to another episode of Direct Current. I'm your host, Matt Dozier, and I'm here joined today by a couple of co-hosts from Oak Ridge National Laboratory, Jenny Woodbery and Morgan McCorkle. Hello! JENNY WOODBERY: Hello, Matt. MORGAN MCCORKLE: Hey, Matt. MCCORKLE: Yeah, so we're at Oak Ridge National Lab, which is located in beautiful east Tennessee. So Oak Ridge started as part of the Manhattan Project, and today we are a Department of Energy Office of Science National Lab with over 4,500 folks who are doing cutting-edge science and technology research here in Tennessee. DOZIER: And I wanted to bring you folks on to talk about something that you are hard at work on, which is the "Sound of Science." So tell me about the Sound of Science. WOODBERY: So we're launching a new podcast called the Sound of Science, which highlights the voices behind the breakthroughs at Oak Ridge National Laboratory. So we're going to be talking to the scientists and researchers who have played a role in, say, the discovery of a new element, or they're working on producing plutonium-238 for NASA that will be used in deep space travel. You get to talk to these people on a regular basis, and it's just really fascinating.  (MUSIC ENDS) DOZIER: So what's it like working at a place that's just so absolutely soaked in science? That's probably not a good word. MCCORKLE: We may change our podcast title. (LAUGHTER) DOZIER: You are surrounded by some of the smartest people, probably, fair to say? MCCORKLE: Yep, I would say so. DOZIER: Yeah, so what's it like? MCCORKLE: One of the things Jenny and I love about our jobs here is that we have the opportunity to talk to some of the smartest people who are working in crazy fields like 3D printing and supercomputing, and they're funny, and they've got great stories to tell, and I always think to myself, "I wish other people could hear these interviews." And that's one of our goals with the podcast, is to help bring those interviews to life in a way that we haven't done before. DOZIER: Tell me a little about the work you do here. WOODBERY: So I manage ORNL's social media channels, and I love storytelling, and we have so many really amazing stories to tell here at the lab. And while we do that through social media, and through video, and feature stories, and things like that, we thought why not explore the realm of podcasting and actually let people hear the stories behind these breakthroughs from the scientists and researchers who are doing them in the field? DOZIER: And it's that sort of connection that people can make with some of these voices, right, that can help them understand better some of this really incredible, groundbreaking science, right? MCCORKLE: Exactly. One thing you'll run into because of our roots in the Manhattan Project, a lot of people associate us with making the bomb. And that's certainly part of our history, in where we came out of, but today we're a cutting-edge research facility. We've got the top supercomputer in the world. We've got neutron sources that can't be beat. We've got 3D printing research. We do research in biology. We do research in materials science. It really runs the gamut, and I think so many people struggle to understand that because it's such a big place. And we want to kind of tackle it one thing at a time, and let people hear from researchers who are doing this, and who are excited and love their jobs. (MELODIC GUITAR MUSIC PLAYS) DOZIER: Do either of you have a favorite topic area of the science that happens here at Oak Ridge? I know you're not supposed to pick favorites. WOODBERY: Honestly, it's all so fascinating. I do tend to gravitate toward what we do for space exploration, to further that — especially the plutonium-238 program. We also make the little insulator cups that holds the plutonium fuel. DOZIER: What about you, Morgan? MCCORKLE: I can't pick a favorite child. It's not fair, Matt. Don't make me do it. (LAUGHTER) DOZIER: All right, I won't. Jenny, is it ever overwhelming with the sheer volume of science happening around you, to keep up with it and help share it all out with the public? WOODBERY: Definitely. There is such a wide range of research that we do here, and so it's fun for me to get to share elements of each thing we do at the lab. But there's such a wide range that you can't really do a deep dive on some of those. And social media, sometimes the medium is just a little more quick. So, I think with the podcast, that's what's going to be really great, that we can go deeper into these topics that we've only — at least in my role — that I've only been able to share a tidbit with the audience. (MUSIC ENDS) DOZIER: So you also were behind something that ORNL posted recently on Twitter, with the "soothing science," right? I think this is a great example... how did you come up with that idea? WOODBERY: In my role, I also do video. And I was playing around with this technology called superhydrophobic technology, and basically what I was trying to capture someone with a dropper dropping water onto a piece of glass. And with the superhydrophobic technology — it's a coating on the glass — the water just bounces right off. So to capture that, I had to capture in a higher frame rate and slow it down into slow motion, so you could actually see the water. And when I was watching that, it just made me think of this genre of posts that I'd seen on social media called "oddly satisfying." You'd see a video of someone cutting kinetic sand with a knife, or shaving off a bar of soap, and you're thinking, "This is really weird, but I can't stop watching it." So I started thinking oh, this is kind of nice, maybe we could create a series out of this, and soothing science popped into my brain. And what seemed to be a very silly idea actually resonated with our audience, and people just didn't expect that coming from a national lab.  (SLOW, BLUESY GUITAR MUSIC PLAYS) DOZIER: So next up, the new frontier for Oak Ridge: Podcasts. So tell me more about the Sound of Science, and if you could give us a teaser of kind of what to expect from the show. WOODBERY: So in the next couple episodes, we're going to be talking to people who have discovered new elements, including element 117, Tennessine. And we're also talking to folks about how they're integrating AI into their work at the lab.  DOZIER: Which is a really good segue, because artificial intelligence is huge right now in terms of the DOE's priorities and in the news. Tell me a little bit about what's going on here at Oak Ridge for artificial intelligence.  MCCORKLE: Yeah, I think it's something that everybody hears about artificial intelligence, and robots, and phone assistants, and things like that. But it's actually kind of blowing up in the scientific field as well, and that's why Oak Ridge is hosting one in a series of town halls that's bringing a bunch of experts in artificial intelligence together to talk about the future of artificial intelligence as it relates to science, and how they're going to use AI. What are those challenges, things like that, so they're kind of mapping out the future of how AI is going to change science. DOZIER: And that, as I understand, is going to be topic of your pilot episode, coming soon? WOODBERY: It is. We're going to be looking at the various ways AI is being used in scientific research across the lab. And one of the ways scientists are using AI is DOE — and its user facilities, which we have several here at the lab — produces a tremendous amount of data. And so AI is making sense of all this data. DOZIER: Do you think you can get some artificial intelligence to help you make the podcast? Is that— because I'm in the market, if that's the option. MCCORKLE: If it would cut down on editing time, yeah. WOODBERY: I was going to say, it sure would make things a little easier. DOZIER: OK, so we've got coming up, your AI episode, pilot launch of "Sound of Science" from Oak Ridge National Laboratory. I think we're going to have a brief clip coming up from your director, Thomas Zacharia, and we'll dive into the work happening here to figure out what the future of AI research is across the Department of Energy and its National Labs. So look for "Sound of Science" coming soon, and thank you both for joining me. MCCORKLE: Thanks, Matt. WOODBERY: Thanks for having us. (MUSIC FADES OUT) (DIRECT CURRENT THEME PLAYS) (UPBEAT, BLUESY GUITAR MUSIC PLAYS) DOZIER: So, as we mentioned earlier in the show, the rest of this episode is going to be about AI — that's artificial intelligence.  DOZIER: We're going to hear from a lot of people across the Department of Energy and our system of 17 National Labs — not just for this episode, but in our upcoming series of episodes exploring all the ways AI is going to turn science, our energy system — and *our lives* — upside-down. CORT KREER: So, what is Artificial Intelligence, really? How does it work? What can we do with it? And why is the Department of Energy involved?  KREER: Many of these questions are still being answered -- in fact, hundreds of scientists and engineers met at Oak Ridge National Laboratory this summer to try to do just that. DOZIER: The event was one in a series of "AI for Science Town Halls" being held around the country, where physicists, chemists, biologists and others sit down with computer scientists and data experts to hammer out the details of how AI is going to revolutionize all of these different scientific fields.  KREER: The person who kicked off this particular brainstorming session was lab director Thomas Zacharia. THOMAS ZACHARIA: I'm Thomas Zacharia, most recently the director of Oak Ridge National Laboratory, but I've been part of Oak Ridge National Laboratory for 32 years. DOZIER: One of the most common questions he hears about artificial intelligence these days is also the most basic one: What is it? ZACHARIA: I would define AI as extracting knowledge or information from large amounts of data, and using that insight that is gained from the data using advanced computing, to enhance our understanding of a particular phenomenon or a particular activity. And then to use that knowledge to provide better services or gain better insights into the scientific phenomenon. KREER: From a scientific standpoint, that's really what AI is all about. Feeding massive amounts of data into a computer, and training it to recognize patterns — without needing a human to watch over it. ZACHARIA: What has given the current level of excitement in AI is actually a seminal work that was done about a decade ago. People call it "deep learning," which is the ability to train neural networks or extract insights from unstructured, disparate data without supervision. So that has really allowed us to essentially throw a lot of data to a computer that has the ability to process it, and it develops correlation between these data that we did not ask the computer to do. (MUSIC FADES OUT) DOZIER: Think about how the scientific method works. Someone comes up with a question, proposes a possible answer, or "hypothesis," then does an experiment to test that hypothesis.  KREER: It's worked really well over the centuries — but it also means that scientists get really focused on one particular answer, and might miss other stuff hidden in the data. ZACHARIA: To some extent we are singleminded in our pursuit of finding that information, and oftentimes you can sort of miss some important things. You can imagine taking a road trip from point A to point B, and you're so focused on getting to point B that you completely miss the scenery. So there's a real opportunity for us to go back into literature and relearn and reinvestigate, and there's quite a bit of excitement about what we might find in those treasure trove of already published data.  DOZIER: You might be thinking to yourself, wait, THAT'S what AI is? What about the computers with emotions? The killer cyborgs? ZACHARIA: Well, I think that most of us have seen a lot of sci-fi movies, and for the people who have watched the Matrix movie, that's what comes to mind. But as I said before, I think AI is really about extracting insights from data.  KREER: Like so many things, the reality of AI is way less sensational than how it's depicted in pop culture. Deep learning and evolutionary algorithms might not be as thrilling as zero-gravity kung fu on a simulated rooftop .... unless of course you're a scientist. DOZIER: Not only can AI help scientists sort through data, answer scientific questions, and pull insights from old studies — it can help them ask better questions in the first place, and design even more effective experiments. (JAZZY, UPBEAT ACOUSTIC GUITAR MUSIC PLAYS) DAVID WOMBLE: I think it can do a lot to change what and how the scientist goes about their work. If you can imagine no longer being responsible for the drudgery of the lab, just day after day of literature survey, but something like that can help assimilate that information and lead toward the hypotheses. KREER: That's David Womble, director of artificial intelligence programs at Oak Ridge National Lab — where the ways scientists are already using AI are already growing too numerous to count. WOMBLE: Sometimes I think it would be easier to answer, how are we not using it? It's a smaller set. (LAUGHS) I would say that almost every scientist here at Oak Ridge has incorporated some form of advanced data analytics, which you could call machine learning, into their research.  DOZIER: Health and medicine, transportation, energy, manufacturing... researchers in all these different fields and many, many more are working to explore the seemingly boundless possibilities of AI for science. And this exploration has uncovered even more big questions. WOMBLE: How can we take data and machine learning and put that into a loop that designs a new material? How can we take machine learning for the genome-to-phenome problem and turn that into personalized medicine?  WOMBLE: I would like to show that, using the Summit supercomputer or Frontier coming up, that we can do something that has a big impact both on the industry and on scientific research in artificial intelligence. (MUSIC ENDS) KREER: OK, so we know that AI has massive potential to shake up science. But there's a catch. Right now, the scientists who want to use AI in their research, and the computing experts who know how to create effective AI tools, aren't always on the same page.  (BLUESY ELECTRIC GUITAR MUSIC PLAYS) CATHERINE SCHUMAN: And so this meeting is to bring together the domain scientists, who have those problems, have those needs, know what their challenges are, with the computer scientists and the engineers who are actually going to be developing the new AI techniques, and how we are going bring those technologies to them and address their challenges moving forward. DOZIER: Catherine Schuman is a research scientist in the Computational Data Analytics group at Oak Ridge National Lab. She was a moderator for one of the many breakout sessions at the big AI town hall meeting. SCHUMAN: That's right. Yeah. So we're the hardware-architecture breakout session, and so we're specifically thinking about, what are the gaps in the hardware we have today that are not addressing the challenges we're going to need that will enable both the science and the AI moving forward. So we want to be able to figure out what are the challenges? What are the new hardware characteristics we're going to need? Do we need new hardware? How do we evaluate the hardware for these applications? These are all new questions for us moving forward, and it requires many different types of people to be able to answer those questions, including the domain scientists, and the engineers and the existing hardware experts to push those questions into the future and figure out how are we going to even start addressing them. KREER: Catherine's research is investigating an entirely new way of designing computers for AI applications. SCHUMAN: I do neuromorphic computing, so I'm specifically looking at new types of hardware implementations to do artificial intelligence. So neuromorphic is literally "brain-shaped" computing, so a fundamentally new type of computer, specifically to do neural network-style computation. It's well-suited to doing certain machine learning workloads. (MUSIC ENDS) DOZIER: Around 20 people attended her breakout session, which was all about the computing hardware and architecture we'll need build in order to use AI effectively. It was.... way over my head. (OBSCURE MEETING CHATTER ABOUT COMPUTING) DOZIER: One big issue kept coming up: Applications like machine learning rely on processing enormous amounts of data, so ideally you want a supercomputer like Summit to crunch the numbers for you.  KREER: But say you're a scientist who wants to use AI in your research... Only, your lab isn't close to a high-performance computing center like Oak Ridge. You're out at what computer scientists refer to as the "edge..." and you probably shouldn't expect to be able to send all your data all the way to Tennessee for a quick spin through Summit. SCHUMAN: Because we can't. We literally cannot take all that data and push it back to a supercomputer. We simply do not have the network bandwidth to support that sort of thing. So that's an entirely new challenge for us to think about. There's some of that compute that's going on now at the edge, but relatively little compared to what we are thinking about for the future — specifically within the realm of artificial intelligence at the edge.  DOZIER: So how do we make the enormous power of AI accessible to scientists everywhere? These are the kinds of problems people like Catherine, and all the other folks who attended the Town Hall, are trying to work out.  KREER: One thing is certain: as far as AI technologies have come in the past decade, this is still just the beginning. SCHUMAN: That's right, I think this morning our Oak Ridge Lab director said something like, "These town halls spin up 10 years worth of meetings, and understanding what we need to do." So we are literally preparing for what we're going to be doing 10 years from now with a meeting like today's, which is exciting. KREER: Beyond the technical challenges that lay ahead, other questions remain. For example, there are legitimate concerns about the societal and economic impacts of rapidly advancing A.I.  (OPTIMISTIC, UPBEAT BLUEGRASS MUSIC PLAYS) ZACHARIA: Certainly these kinds of shifts happen, new technologies always disrupt old technologies. I think AI will do the same thing. Is AI going to disrupt jobs? Absolutely. Is it going to create catastrophe? No. I think will enhance the human condition and it will create new jobs. Some I can clearly imagine, some that I have no idea. But I do know that it will ultimately take hold because it would be beneficial for humanity at large.  DOZIER: We'll be exploring more of this topic in future episodes, but it's worth touching on one final point before we sign off. In addition to the AI town hall meetings, the Department just launched its Artificial Intelligence and Technology Office, which aims to connect the AI research happening across DOE and all 17 National Labs within a single hub. So... why is the Department of Energy putting so much focus on artificial intelligence?  KREER: It's a good question. And with more than 30 years of experience in the Department's National Labs, Director Zacharia is well-equipped to answer it. ZACHARIA: DOE generates a lot of amazing data at its user facilities. To be a leader you not only must have unmatched computing capability, but also you need unmatched data. Ultimately, it comes down to talented people. I don’t think there’s any other organization in the world that has access to as many talented people as DOE has across its 17 national laboratories. If DOE does not lead and drive this innovation, who will? I truly believe we are in the process of inventing the future, influencing how the future evolves. It’s going to have profound impact on how we live, how we work, how people are cared for, how you consume information, it’s an exciting time.  (MUSIC FADES OUT) (SOFT ACOUSTIC GUITAR MUSIC PLAYS) DOZIER: That's it for this episode of Direct Current. We want to thank all of the folks at Oak Ridge National Lab for hosting us and talking to us about artificial intelligence, especially Jenny Woodbery and Morgan McCorkle. Their new podcast, "The Sound of Science," is dropping this fall. Follow Oak Ridge National Lab on social media to get a heads up when the first episode on AI research arrives! KREER: There's lots more AI content to be found on our website, Energy.gov/podcast. And stay tuned for future episodes, where we'll dive deeper into the many ways this rapidly advancing technology is going to revolutionize science, energy, and much more.  DOZIER: As always, if you've got a question or want to leave us some feedback, email us at directcurrent@hq.doe.gov, or tweet @energy. And if you're enjoying the show, share it with a friend and leave us a review on Apple Podcasts. KREER: Direct Current is produced by Matt Dozier and me, Cort Kreer. I also create original artwork for every episode, which you can find on our website. KREER: Normally, we'd sign off here, but I have a different message this time. This is my last episode as your cohost. I want to thank you, Matt, for your tireless work finding these amazing stories, wrangling our experts, writing the scripts, cutting it all together, and about a million other labor-intensive, sometimes maddening tasks that are necessary to create this amazing and formative, sometimes super-gross (LAUGHS), and always amazing award-winning podcast. You make it look easy, and it's been an honor. And I want to personally thank all of you folks out there for listening, and for sticking with us. None of this would be possible without you, and I know that sounds totally cliche, but it's true. This podcast, these stories, this science, this department, it all exists for you. And I wish you the best. Godspeed, and smell you later (LAUGHS). Smell you later forever (LAUGHTER). DOZIER: We're going to miss you, Cort. Thanks everyone, for listening. KREER: Now and always, thanks for listening. (MUSIC FADES OUT)