Not Alone: Leaders in Conversation
Not Alone is a service. It is designed to disseminate the ideas of established, extraordinary academic leaders to inspire, generate conversation and showcase the best in academia and related businesses and institutions. The contributions are short and deal with a single subject, by design.
Not Alone: Leaders in Conversation
In conversation with... John Hennessy
Navigating Leadership in Academia: An In-Depth Conversation with John Hennessy
In the inaugural episode of 'Not Alone, Leaders in Conversation,' host Rafael Bras engages in a profound discussion with John Hennessy, esteemed academic leader and Chairman of Alphabet. Hennessy shares his insights on shaping Higher Education, his journey from Stanford's President to an influential entrepreneur, and the pivotal experiences that shaped his career.
The conversation touches on the importance of mentorship, the evolving landscape of academia, technological advancements like AI, and leadership traits outlined in Hennessy's book 'Leading Matters.' An inspiring dialogue for academic leaders, professionals in the field, and anyone interested in Higher Education's future.
00:00:00 - Introduction to Not Alone Podcast
00:00:45 - Meet John Hennessy: The Godfather of Silicon Valley
00:03:13 - Early Influences and Upbringing
00:04:06 - The Role of Luck in Success
00:06:00 - Mentorship and Leadership Lessons
00:07:18 - Traits of Effective Leaders
00:10:21 - The Knight Hennessy Scholars Program
00:12:58 - Challenges in Academic Leadership
00:15:03 - The Future of Higher Education
00:23:05 - Technology and Education
00:25:05 - Universities and Entrepreneurship
00:33:12 - Navigating AI in Academia
00:34:19 - The Rapid Rise of AI
00:35:29 - Breakthroughs and Transformations in AI
00:38:55 - The Role of Universities in AI Development
00:42:26 - Challenges and Future of AI
00:49:27 - Ethics and Regulation of AI
00:51:09 - Preparing for the AI Revolution
00:55:43 - Personal Reflections and Legacy
01:02:35 - Conclusion and Final Thoughts
(0:00 - 0:38)
Welcome to Not Alone: Leaders in Conversation, a bi-monthly podcast where we delve into the minds and experiences of academic leaders who are shaping the future of higher education. Your host is Rafael Bras, and in each episode he will explore the complex challenges, decisions and opportunities facing academic institutions today. Whether you’re an academic leader, work in academia or simply want to learn about the latest topics that are top of mind for academic leaders, this series is designed to provide valuable insights and inspiration.
Please join us as we embark on this journey of conversation, discovery and leadership.
[Rafael Bras] (0:45 - 2:45)
It is an honor to have John Hennessy as the first guest in Not Alone’s podcast series, Leaders in Conversation. John has been called the Godfather of Silicon Valley. He is a scholar, a teacher, an administrator, an entrepreneur and an admired leader.
For 16 years, starting in year 2000, right at the birth of the technology bubble, he served as the very successful 10th president of Stanford. He has written some of the most successful textbooks in computer science. His research literally transformed computer architecture and software techniques, leading to his founding of several successful companies.
He is presently the chairman of Alphabet, parent company of Google. John has been honored with recognitions like the Turing Award and Draper Prize, National Academy of Engineering, the Royal Society and many others. He is the visionary founder of the Knight-Hennessey Scholars, a multidisciplinary, multicultural program spanning all seven schools at Stanford University and the largest university-wide fully endowed graduate fellowship in the world that prepares graduates to be visionary, courageous, and collaborative leaders who address complex challenges.
Few are more prepared to talk about leadership than John Hennessey. Thank you, John, for participating in this effort that aims to bring the wisdom of academic and research leaders to all around the world.
[John Henessy] (2:35 – 2:37)
Delighted to be here, Rafael.
[Rafael Bras] (2:37 – 2:45)
Thank you so much again. I’m going to start with the most important question: Tell me about your golf.
[John Hennessy] (2:46 - 2:51)
Oh, my golf. My golf is a work in progress, and it will be for many years to come. (laughs)
[Rafael Bras] (2:54 - 3:25)
I'm amazed because I think you are very much — and we'll get back to the theme — a man on the move, and certainly you have to get going all the time. And golf … it’s a game of patience.
But anyway, let me get to the serious matter. Most of us are shaped by experiences, particularly early experiences. Can you share how your early upbringing, your family, impacted who you are?
[John Hennessy] (3:26 - 4:05)
Surely. Well, my father was an engineer who put a love of science and engineering and problem-solving in me. And my mother was a voracious reader and a school teacher who really instilled in me a love of reading that I carry through to this day, I think.
So both of them shaped me early. My father was the first in his family to go to college. My mother was the first woman in her family to go to college.
So the college, they had the benefit of that, and they believed in education and the importance of education. So that was a key part of my upbringing as well.
[Rafael Bras] (4:06 - 4:31)
Very good. In some of your writings and talks, you make the statement that a lot of our lives are shaped by elements of luck. It is what I have referred to myself as accidental life — not truly accidents but opportunities that arise by design or truly unexpectedly.
Can you speak to a couple of those occurrences in your life?
[John Hennessy] (4:32 - 5:59)
Surely. I think the one that really transformed my opportunity was when I was just starting graduate school, a colleague from Brookhaven National Lab walked in and handed me a thesis topic. And the topic at the time was: Could you build a software system that would make it easier to do real-time control for various applications they were interested in at Brookhaven?
At the time, the field was relatively new. I jumped into it, and it turned out, just as I was graduating, several of the most important people in the field published papers in the same area. So all of a sudden, this area took off like a spaceship, and I was lucky enough to be at the very beginning.
And that was pure luck. It was pure luck that I got that topic. Certainly I did good work on it, but it was fortunate timing.
In most startups, people can have great technology. There is also an element of luck in any successful startup. And the startups I’ve been in, sometimes your product is too early, sometimes it’s too late, sometimes you stretch it, sometimes you pick the right people — and there’s a lot of luck in that process.
So I think you want to position yourself to take advantage of an opportunity, but that opportunity may or may not come along depending on timing, and that’s just a hard thing to measure precisely.
[Rafael Bras] (6:00 - 6:07)
Have you had any mentors that were particularly effective in leading you to the right place?
[John Hennessy] (6:07 - 7:05)
I think people helped me along the way. My colleague Forrest Baskett, who really instilled in me how to do good computer systems research and really do it well. Jim Clark, who was the founder of Silicon Graphics, who I got to work with and learn how to build a startup company from that.
Then Jim Gibbons, who was my Dean of Engineering for many years, who taught me a lot of important lessons about leadership. He gave me some of the most important advice I had. I was being offered the job of Dean to succeed him, and he said to me: ‘Take this job because you want to serve the faculty and students of the School of Engineering. Don’t take it because you get a higher salary, because you get a bigger office, because you have a bigger staff. Take it because you want to help make them successful.’
That was advice I used throughout my time as Dean, Provost and President.
[Rafael Bras] (7:05 - 8:05)
It is amazing how some of those little tidbits of advice can really and truly change your life. Not many people sit back and think of them, but it’s, in fact, unbelievable. In your book Leading Matters, you speak about 10 traits or characteristics of leaders.
I really like your approach because the focus is on experiential leadership development without the common attempt to codify a method. The 10 points you give are humility, authenticity, leadership as service, empathy, courage, collaboration and teamwork, innovation, intellectual curiosity, storytelling, and legacy. You never say there the one that appears everywhere, which is honesty and integrity.
And I know you have an answer for that, but can you expand on it somehow?
[John Hennessy] (8:05 - 10:20)
Yeah, it’s a good point. I put it under the topic of authenticity because I think you’ve to be authentic, you have to be honest, and you have to have integrity. That’s a foundational role, but I think it goes beyond that.
Authenticity also says, not only are you honest and have high integrity, but you are upfront with people. You are straightforward with them so that they will buy in. It’s connected very much to trust because in any leadership role, you can only lead as long as you have the trust of the team you’re trying to lead.
And if you lose that, either because you violate integrity or you’re dishonest or you’re not authentic, you’re deceptive. So there’s a kind of line there that I think is really important. And I always wanted, when I was in a leadership position, I wanted my colleagues to know that if I said something was going to happen, I was going to do my darndest to make sure it happened and they could depend on that relationship.
[Rafael Bras] (9:14 – 9:15)
Tell me about storytelling.
[John Hennessy] (9:16 – 10:20)
Yeah, storytelling is an interesting one. I think we put it in there, and Walter Isaacson, who wrote the forum for the book, observes that this is one he didn’t expect to see.
But of course, storytelling is about weaving for a leader, often weaving a vision — a vision of where you want a team to go or dealing with a crisis. You have to be able to weave a story in a crisis as well about how you’re going to succeed even though you’re in the middle of a crisis. So storytelling is a key element. It’s, in fact, for our scholars in our program, their first year, they take a quarter-long course that moves to once a week focused on storytelling. And really getting … At the end of that, I want them to get up and talk about — with passion and excitement — about their research area or their journey or their trajectory going forward. And I think that’s a really powerful capability.
[Rafael Bras] (10:20 - 10:38)
Yes, let me pick up up on that and particularly the Knight-Hennessy program that, as you know, I’m very interested in. It is, I think, a fantastic concept. Can you talk a little bit about the elements of what they go through over, I believe, is a 3-year engagement, correct?
[John Hennessy] (10:40 – 12:22)
Three years, yes. So the program varies first year, second year, third year. A key element of the program: every year there’s a retreat where we go away and we both develop some skills but also build community and connection. Because one of the remarkable things about this group is that there are scholars from around the world studying all different subjects.
So you’ve got a really interesting cross-disciplinary mix and discussion that can go on. So, in the first retreat that we do for the first-year students, we really try to build that community and interaction and get people to appreciate other people’s journeys. And they see amazing stories of what people have overcome or what they’ve done on their way there.
We also organize a global trip for them to somewhere in the world, once during their three years, and they go and visit. I went with a group to South Africa that looked at post-Apartheid struggles in South Africa. But we’ve been to Patagonia, we’ve been to India, we’ve been to Turkey, we’ve been around the world. And they get an opportunity to meet people who are trying to solve the challenges and problems that exist in some portion of the world.
So these are the key elements. A lot of experiential focus. So we do, for example, a workshop on negotiation. How do you negotiate? How do you think about a win-win in a negotiation? People, sometimes when they hear about negotiation, they think winner and loser. That’s not a good negotiation. A good negotiation is where both sides think they came out OK because that builds a lasting partnership and relationship.
[Rafael Bras] (12:23 – 12:37)
I cannot imagine. I’m envious of those students. I'm sure that the program does change their life, and I’m willing to predict that when they leave Stanford, that’s the one thing they’re going to remember.
[John Hennessy] (12:38 – 12:39)
Yes, I think so.
[Rafael Bras] (12:40 – 13:07)
You talk about experience. Experiential learning in the Knight-Hennessy covers that a lot.
I agree that you have to develop the instincts to be a good leader, particularly in increasingly complex situations. I would argue — and that’s my opinion — that experiencing academic leadership is suffering. Would you agree with that?
[John Hennessy] (13:08 - 14:19)
Yes, I think so, Rafael. I think we, as academic institutions in most cases, have not done enough to develop the next generation of leadership. Now, I think in some parts of the university, they’ve realized that shortcoming.
When I was president, for example, I was sitting on the board of Cisco at the time, and I saw what they really did to develop their leadership. They would move people around to different parts of the organization and give them different experiences, leading in engineering or leading in marketing. I came to understand that we needed to do a better job in the university so that when we were looking for somebody who might be a department chair, a dean, a provost, we’d have a pool of candidates that had a set of experiences that really prepared them.
So I think universities are beginning to grapple with the fact that they haven’t done enough to prepare people, and we’re going to have to. And the job is, the other aspect is, the job has gotten a lot tougher. (With) the pandemic and the disputes over the war in Gaza, it’s been a really rough five years, I think, for university presidents.
[Rafael Bras] (14:20 - 14:29)
I think, indeed, the job of all academic leaders is tougher, but speaking for myself, it was fun.
[John Hennessy] (14:29 - 14:43)
Yeah, I loved it. I loved every minute of it. I really enjoyed it.
Of course, there’s some tough things you have to go through, there always are, but there were many more joys than there were tough situations.
[Rafael Bras] (14:43 - 15:49)
Yeah, I get disappointed when I hear colleagues that say: Well, you must be feeling great, you don’t have that responsibility any longer. And my reaction is: No, that’s not the point. I really enjoyed my job, and maybe you should think about it also.
Talking about academia and our failure in some elements, for example, developing of leadership, but we’re going through a period of change that I think is faster than it has ever been. There are issues: changes in demographics, the erosion of the public trust that you know about and understand, technology changing the way we deliver information or how we interact and — very important, in particular, the one I really would like you to expand on — the erosion in the affordability and accessibility of good education.
[John Hennessy] (15:50 - 17:22)
Yeah, so I think we’ve got a major problem on our hands, and you just look at student debt loading, and it’s gone up the last roughly 10, 15 years, it’s gone up significantly. Part of it, of course, is states have cut back on their investments in the great public institutions that educate most of our students, and they’re putting more of the burden on families. Families are in tighter financial circumstances, and we haven’t done a great job of taming costs.
We have not done a terrific job.
It’s interesting right now that tuition is going up by relatively modest amounts, but living costs for students are going up dramatically, and while a residential educational experience is the crown jewel — it is the thing where students learn constantly — it’s becoming less and less affordable, and we’re going to have to find a way to close that gap, which probably means the universities also have to find a way to constrain their cost increases over time and be more generous with financial aid. So financial aid is clearly ... I actually like the system we have in this country where we give financial aid to families that need it, and we ask families that can afford it to really invest in their children’s education. And that’s a good balance, I think; it’s a socially progressive approach. But we’ve got to ensure that that accessibility doesn’t get removed as prices go up.
[Rafael Bras] (17:24 - 17:55)
Yeah, that is crucial, and the key to a large extent is finding the resources to provide that financial aid. In public universities, unless the state provides it, you have to raise it privately, and unfortunately the difference between privates and publics in that sense are large. Do you see that gap between privates and publics expanding, or do you think they will get closer in the way of thinking?
[John Hennessy] (17:55 - 19:02)
Yeah, I think we’ve seen a deeper understanding in the publics that they’ve got to be as aggressive about fundraising as privates have to be because they’re not going to be able to depend on the states the way they have in the past. So they’re going to have to ... and that’s really a process where you begin educating.
I said we begin educating the alumni about the importance of giving back when they arrive as freshmen because we remind them that they’re there, and the experience they’re going to have over the next four years has been subsidized by previous alumni who had the benefit of the university. So we’re going to have to do a better job of articulating that so we can continue to maintain affordability for families. That’s a crucial element.
And if you look at the economics in the US, the benefit of a college education over a lifetime is gigantic in terms of economic benefits. So we have to make sure that that remains accessible to students who really are determined to complete a college education.
[Rafael Bras] (19:03 - 19:47)
Well, one of the things that is making the job of academic leaders very difficult today is the disruptions and protests around the war in Gaza. It is not the first time this happens. You and I were in school at the time of the Vietnam War, and as I say, those were unbelievable events and protests of incredible magnitude.
This has triggered a debate about freedom of speech and what is appropriate or not appropriate. Are we handling it correctly? What is new?
[John Hennessy] (19:47 - 21:57)
I think there is an element that’s new, which has made this a much harder situation. In the past, we had demonstrations about living wage, sweatshop labor, fossil fuels, and then, of course, before that, the Vietnam War. The difference is that in most of those, the students and faculty were all on the same side.
In this case, there really is a division, right? There really is a division because what happened in Israel was horrible, and what’s happening in Gaza is horrible. So there’s a division with students against students. Oh, and that’s much harder than having … Students against administration — that's our job, to deal with that. But students against students is really hard.
I think we’ve probably erred a little too much on speaking out about public events that were not directly relevant to the university, and we got ourselves in a little bit of trouble because one of the things that students want — the students on both sides want us to say something about the war, but they want us to say something different about the war, and that’s a difficult position to be in. I think navigating this is going to be challenging, I think, for some time to come, and maybe the fall will be a little better because I think we’re better prepared for it than we were last year.
We were simply not prepared for an encampment growing up in the middle of campus, and we’re better prepared. I think you touched on this challenge. The demographics of our universities have changed dramatically over the last 20 or 30 years — dramatically — and now how do we both maintain openness and free speech and the marketplace of ideas and, at the same time, make people feel included and part of the university community?
And that’s a challenge we will have to struggle with, just as the country has to struggle with it to some extent.
[Rafael Bras] (21:58 - 22:34)
But you touched on something that I think is terribly important and is truly being missed by academic leaders, and that is that these disruptions are different, and they’re different because they are reflecting that division within the community itself, and it’s student against student, it’s faculty against faculty, and that had not quite happened the same way before. That is a very astute observation and one that should shape our responses.
[John Hennessy] (22:35 - 23:04)
Agreed. And it makes it harder. It absolutely makes it harder, and the university is trying to. It’s not that you necessarily have to be completely neutral, but you have to see merit in both sides of the equation, in both parts: Both communities have legitimate grievances, and figuring out how you deal with that in a way that recognizes them without appearing to take sides is going to be the challenge.
[Rafael Bras] (23:04 - 23:38)
Very good. Let me talk a little bit about technology. And we’re going to spend a lot of time later in AI, so you can bring it up now or not, but technology is a big player in education now. In fact, Stanford played a significant role in the whole growth of MOOCs and large-scale educational systems and platforms to deliver content.
What is in the future? How do you see that playing out in our universities?
[John Hennessy] (23:39 - 24:52)
Well, I think the good news is that content is now widely available, and there’s lots of content out of there. Whether it comes from our universities or it comes from Khan Academy, for example — there’s lots of content out there that can be used to reinforce student learning. I think what we need to do is figure out how do we ... The thing we should be doing with technology is improving and enhancing the ability of people who are teaching to really be effective in those settings.
I don’t think we can completely remove the human from the element, the human from the education process, because that’s what inspires students. It’s hard to get inspired just looking at a talking head on a screen or set of videos. It’s easy to get inspired by somebody who really explains an excitement about a field.
Now, I think we can use those things together to do a good job, and as my friend Bill Bowen once said, we’ve got to bend the cost curve in higher education, and technology is perhaps the one way we can do it that doesn’t reduce quality of student experience.
[Rafael Bras] (24:52 - 26:07)
Yeah, I think it’s doable, and I believe it’s changing faster than most realize in many ways.
Let me talk a little bit about universities and entrepreneurship. I think we both agree that the university of the present and certainly the university of the future is part economic development agent as much as it is an agent to deliver education, to prepare people or do service. Economic development is now part and parcel of what we do.
You are a successful entrepreneur. You’re chairman of one of the largest companies in the world. Yet you are a consummate academician who can never be too far from students and research. That’s been very clear to anybody that knows you. Is this duality for everybody? Should entrepreneurship be a metric for academicians?
What are the important metrics that a faculty in a modern university should be judged by?
[John Hennessy] (26:07 - 27:35)
Yeah, it’s a really good question. I started as a reluctant entrepreneur. I was convinced by a number of colleagues in the industry that if we didn’t take our research and spin it out, our work was not going to be picked up by the existing players because of not-invented-here syndrome and similar kinds of things.
But in that process, I learned a tremendous amount. I think one of the things we’ve seen at Stanford is, faculty who’ve been involved in entrepreneurship activities come back as much better teachers because most of our students are going out into industry — they’re not going to become future academics — and getting that experience, having somebody who’s been in that and understands how to articulate a technical vision, for example, to a management team is an incredible teacher.
So I think you’re right that universities are generators of economic growth and opportunity, and they’ll be expected to do that. And that’s a key role for research.
It’s one of the reasons — not the only reason but one of the reasons — a government justifies a research investment. And I think we can figure out how to manage that.
There are, of course, challenges, conflict of interest, conflict of commitment challenges that have to be managed and dealt with. But I think overall, the benefits far exceed the downsides.
[Rafael Bras] (27:37 - 29:01)
To follow up on that, the question of where the university ends and business begins has always been a very difficult one, both in private and certainly in publics. Most research universities have seen an uptick in academic dishonesty. You could argue that that is due to all sorts of things: pressure to publish, increasing competition, difficulty differentiating between the university and personal business, too much money at play, and particularly in technology.
In public universities, in particular, that has become a real issue because we advocate that coming together of the business and the university for the benefit of creating knowledge and educating, but at the same time, the public wants to make sure that the public is not paying for the enrichment of individuals. And that creates a potential conflict that we are still struggling how to deal with.
[John Hennessy] (29:02 - 31:08)
I agree with you. I think particularly the crux of the problem is when there isn’t a clean separation between the entrepreneurial activity and the ongoing work at the university. In information technologies, that separation tends to happen pretty quickly because a young company will staff up and quickly have 20 engineers, which is a lot more than most research activities at the university can ever put together. So there’s a rapid separation.
Sometimes it’s harder in the biomedical fields and the basic biosciences, where typically there’s still a lot of ongoing technical development work that needs to be done. And the way I’ve seen this over time is, if you look at most IT spinouts, information technology spinouts, most of them don’t fail because the technology doesn’t work. They fail because they miss the market, they don’t get enough takers, they’re too late.
In the biomedical area, it’s a little different. If you find a solution for a disease, cost is not the issue, but you don’t know whether or not you actually have a solution at the time you may spin it out of the university. And it may take another five to 10 years to determine whether or not your discovery at the university is really going to be an effective therapy. That just creates potential for conflicts that you have to manage somehow. And that’s something I think we have to accept.
Given the importance of university research in the biomedical area, you have to hope for startups. But if you look at the National Institutes of Health regulations, they’re struggling with how do we balance conflict of interest and yet ensure that our research developments are leading to therapies that improve people’s lives. That’s a really difficult thing to do.
[Rafael Bras] (31:09 - 31:46)
Yeah, I think many times agencies, funders, universities — we are providing a sort of a model message. And the faculty and the students in general have to deal with it. And it’s sometimes not very clear.
But that begs the question: Is the way that universities operate, is the way we hire, let’s say, faculty, the right way for the future? Is there a different approach?
[John Hennessy] (31:48 - 32:24)
There could well be. I think we’ve tried to hire people who are both great teachers and great researchers, particularly at our research universities like Georgia Tech and Stanford. We try to get both, right? And maybe even sometimes we try to get great researcher, great teacher — and entrepreneur as well, or potentially university leader. And then we’re asking a lot from individuals. And maybe we need to rethink those roles and how to do them better.
[Rafael Bras] (32:25 - 33:10)
But what about the professor, like I saw this many times, that chooses appropriately to take a leave and go and work for six months in Google, and then come back, and then a year from then goes back another six months, and then comes back. At the end of the game, I wonder whether, in this particular example, Google gets the most of the individual and we get the most of the individual. Is there any way where we can bridge that gap and potential conflict so that the individual is really working for both?
[John Hennessy] (33:11 - 34:18)
Yeah, yeah. I think we’re going to have to deal with this issue. Of course, you can put in place rules about how often people can be on leave and things like that. And most universities have these so that faculty are not gone for very long, because they do have tenure after all. And so you can’t afford to have them for a very long time.
But I think particularly with this rise of AI technology and the importance of information technology, and the fact that the resources available — the computational resources available in Google or OpenAI or Microsoft or Facebook — far, far exceed what universities can provide. Therefore, they have access to a research tool that we do not have access to inside the university anymore. So figuring out how we build a constructive relationship that lets both companies thrive and the universities thrive is going to be key. And I think we’re going to have to do some deep thinking about that.
[Rafael Bras] (34:19 - 35:28)
Let me move to what I think is going to be a significant amount of discussion on AI that you just brought up. I was fascinated because I was listening to an interview you did in 2016 when you stepped down or you had announced at that point that you were stepping down as president. You gave a wonderful interview.
You discussed many things: the future of technology, things that were in your mind important in the future or will be important in the future. AI was never mentioned. Yet here we are, 2024, and that’s about everything that people talk about is AI. An incredible change when you think of it in a matter of eight years, particularly for somebody like you who is very much embedded in these issues.
Oh, how did it happen? How did it move so fast?
[John Hennessy] (35:28 - 36:52)
It did move incredibly fast. So I think a couple of things happened. First of all, the first real breakthrough paper, I think, was the AlexNet paper, and that’s in 2012. But that’s purely image recognition, but it was a big jump forward in image recognition.
The next key turning point was the AlphaGo victory, where basically an AI-based system from DeepMind beat the World’s Go Championship. That occurred probably 20 years earlier than people thought we would get there.
So all of a sudden, we knew we were on a curve that really was a discontinuity, I think. And from then on, it piled on rather quickly: the Transformer paper, generative AI. And what empowered this were two things.
First of all, availability of data. So we had massive amounts of data to train. It began ImageNet, first of all, to train image recognition. But then the World Wide Web as training for natural language, and then massive amounts of computing power, whether GPUs and TPUs. We amassed massive amount of computing power to train these very large models. And that, all of a sudden, created a discontinuity in terms of what these models could do. And now you’ve just seen it accelerate at an amazing rate.
[Rafael Bras] (36:53 - 37:04)
It is mind-boggling. I’ve been hearing about the promise of AI for 50 years, and always led to disappointment.
[John Hennessy] (37:04 - 37:11)
What we call AI winters. We call them AI winters. In years, oh my God! (laughter)
[Rafael Bras] (37:12 - 38:22)
And then all of a sudden, as you articulated, things change, both the tools, the data, but also the framing of the AI concept in some of those breakthrough papers. It’s just mind-boggling. Now, one of the interesting things to me about this is that all this development, all these breakthroughs we’re living through right now, are really coming, for the most part, from the very large companies with very large resources.
Governments, despite billions in investments, were really not the drivers, and I would argue are not the drivers at the moment of that. OpenAI, Microsoft, when ChatGPT came about, it was like from one day to another, and people turned around, including me, and said, ‘What happened? What went on?’
Will the future of AI remain in the hands of the very large technology group?
[John Hennessy] (38:23 - 40:05)
Well, they certainly will probably be the builders of these very large models. So a couple of things we understand already. You don’t need these very large models for more limited domain kinds of things. And so, we’re seeing that you can reduce the size of the models by factors of five to 10 and get results for a particular domain that are just as good as they were from that large model. And that obviously improves access to the model because it reduces the computation time.
I think universities have a vital role to play on a number of issues related to AI that the companies will be slower at pursuing.
So, issue of explicability. Suppose I build an AI system that does medical diagnostics or reads x-rays or does some other kind of medical procedure. Not only do I want to get an answer from that, but I want to get an explanation for why a particular answer: why this particular patient has some disease or should have therapy. I want to understand why that’s the case. And currently, that’s not what these models do. They give you an answer, and good luck trying to figure out why that answer is the right answer.
So that’s an area where university research can play a vital role in this. I think similarly on issues of ethical use of AI, of bias, of various problems of this form, the universities will be important players in trying to chart a course for how do we get more responsible use of AI.
[Rafael Bras] (40:06 - 40:55)
Let me follow up on that because we’re describing, or you’re describing, a world where, as you said, the large models are in the hands of the companies with, (for) most of us, are infinite resources. The universities are going to be there to try to explain, to get behind the walls and understand better what’s going on, but still limited in what they can do because of the resources. Then there are all the related issues of, well, I guess I can put it this way:
Are you concerned that the companies will have just too much power?
[John Hennessy] (40:56 - 42:25)
I’m more concerned that … it’s never been the case in the past that universities have not been able to chart a research course at the frontier of a field. They’ve always been able to get the resources, sometimes through national efforts. Obviously, you need a collider of some sort if you want to do basic research in atomic structure, right? So, we built SLAC, we built other colliders. Similarly, telescopes: We build large telescopes as a country in order to probe the structure of the universe and understand that.
I think something similar has to happen and there has been some movement towards a national computational resource that would allow universities to get access to very large amounts of computational power.
So I think we’re going to have to think about how do we chart that course going forward. I also think universities are just going to have to accept that computing is as important as a resource for computational and research as the large developments that we’ve put in the life sciences or biology or chemistry, where we spend a lot of money creating research environments that allow the faculty to thrive. Similar kind of thing now in computing.
Computing has been cheap as a research field in the past, but now participating in this AI revolution is a lot more expensive.
[Rafael Bras] (42:26 - 43:01)
So, we’ve been talking about the expense that AI requires: The computational, the curating and dealing with this massive amount of data is almost hard to comprehend. The energy consumption that goes on behind all this and the time … Is there a business model for AI products, particularly the large models? Can you actually make money?
[John Hennessy] (43:00 - 44:35)
Sure. I think you have to separate out two parts, right? You have to separate out training, which is very costly but happens rarely, once a year or something like that, and inference, right, where you’re actually using the model. Now, the inference cost is, that’s going to happen millions and billions of times a day as people use AI tools to do all kinds of things, right?
So what we need to do is drive down the inference cost significantly. Some of that will come from tuning and adjustment of models. Some of it will come from algorithm enhancements.
I think my good friend Don Knuth once said, the fastest way to get better performance is with a better algorithm. And so, if we can find algorithms which can do inference more effectively because they’re structured differently, the model is structured differently, then I think we can drive down the cost of that. Currently, the costs are considerably higher than many other functions we do on computers.
So we’re going to have to work hard to kind of reduce that cost and make the technology more applicable and more affordable. And we’ve got to deal with energy use as well. Some of that will come from making the hardware more efficient. Some of it will probably also come by trying to get more renewable sources of energy. You know, if you want to build a data center, my suggestion is you build it someplace that’s cold and sunny. So, that’s a great combination. (laughter)
[Rafael Bras] (44:36 - 45:27)
Yes, indeed. Yes, indeed. We had a cooling system crisis here a few weeks ago, which created havoc on all the computers, interestingly enough.
So, I’ve been told that every time I do a query on AI or on any of the many available platforms, that in fact is costing a lot more than I'm paying in many ways. So you wonder how, as you said, the cost has to come down. Otherwise, how it happens is quite a challenge.
Let me ask you about the future of AI. Will the genAI tools become indistinguishable from human intelligence at one point or another?
[John Hennessy] (45:28 - 46:22)
I don't think we’re there yet. So, the observation I’d make is, the reason these systems, the generative AI systems, are so impressive is they have mastered natural language. And language is fundamentally a very deep human thing. It’s what makes us different from all the other creatures on the Earth, right? We can communicate with language, and we can communicate sophisticated ideas. So they’ve mastered language.
What they haven’t mastered yet is reasoning. They’re not particularly good at mathematics or reasoning through a math problem or things like that. It’s a particular problem that they haven’t seen before, right? So they haven't been able to train on something that gives them some key insight to it.
So that’s the gap that I think we really need to bridge if you want to get to so-called AGI — if you want to get to artificial general intelligence as opposed to limited domain intelligence.
[Rafael Bras] (46:23 - 47:01)
Yeah, but what about emotions? This is something that I believe in my playing with it, particularly generating images, trying to reproduce art. And when I do it, at least some time back, the art I was getting had some of the traits of the individual I was trying to reproduce, but it lacked the emotional component of it. It just was not there. That muse, that emotion was not there.
[John Hennessy] (47:02 - 47:57)
Yeah, I think there’s an interesting … and it’s one thing to tell an AI system, you know, ‘Paint me something that looks like a Van Gogh.’ It’s another thing to paint the first painting that looks exactly like a Van Gogh, right? Or a Picasso or a Monet or something like that. That’s a whole different, a whole different … And I always point out to people there’s a, the Museum of Modern Art in New York, there’s this incredible piece by Picasso basically is the seat of a bicycle and the handlebars, and he’s mounted them together. And you look at it and you immediately say, ‘That’s the head of a bull.’ Now, there’s no way an AI system is going to dream that up without having you told it what you wanted, right? And that’s the kind of … but that’s human creativity.
[Rafael Bras] (47:58 - 48:00)
Exactly. But that’s what defines us, right?
[John Hennessy] (48:01 - 48:24)
It is what defines us. And I think that’s a hard thing to get over. I mean, these AI systems are very good at cloning the style of artists that already exist, whether it’s in music or in visual arts, right? Very good at doing that. Much harder to create the next thing, right?
[Rafael Bras] (48:25 - 48:53)
That’s the key. And I think that it was, to me, that’s the exciting and scary part. If it ever gets to that, I don’t know what we will do.
You know, and how does an AI system accept a mistake? That’s another human trait. Does it know … It never knows that it makes a mistake because the algorithm is always right.
[John Hennessy] (48:52 - 49:25)
Yeah. So then human intervention comes in. That’s where these, you know, the systems that are being fielded now, they’ve used human training with reinforcement learning to kind of correct themselves and try to eliminate negative properties. So if it makes a mistake and it’s told it's made a mistake, it can use reinforcement learning techniques to avoid that mistake the next time. But you have to tell it it’s wrong and adjust the training appropriately.
[Rafael Bras] (49:27 - 49:53)
Clearly, the technology can be used for nefarious things. It’s been done already, misrepresenting individuals, et cetera, et cetera, et cetera. Should the technology and those that produce it be regulated in some fashion to control that? And can the industry regulate itself?
[John Hennessy] (49:54 - 51:08)
Well, I think ... So, the way I like to think about this problem is an analogy. Think about automobiles. Automobiles kill 1 million people around the world every year. But we don’t eliminate automobiles because they’re useful. They’re useful tools.
Instead, what do we do? Well, we require that you get a driver’s license. We put in stop signs and stop lights, and we have rules of the road. And we say you can’t drink and drive. So we regulate the use to prevent the negative occurrences. The challenge with AI will be to regulate the use of it.
Software technology is flexible; it can be used for many things. And so, the same piece of technology can be used to create deep fakes that can be used for something really useful. And what we’re going to have to do is separate out the bad uses from the good uses. Now, the companies can also do some things which try to inhibit or prevent bad uses. But in the end, we’re going to have to go after the people who misuse the technology to do something illegal or unethical, rather than just try to stop the entire technology from progressing.
[Rafael Bras] (51:09 - 52:01)
Are we ready for AI? And when I say we, the public at large as well as universities. In a recent survey that Elsevier did, that is called From Above (View from the Top), university leaders stated that AI was, in their mind, one of the big things for the future that concerned them. But they also stated that they were the least prepared to deal with it at that point.
In another report called Insights 2024: Attitudes toward AI, the overall 44% of researchers are unaware of any institutional plans for preparing for AI usage. What can we do?
[John Hennessy] (52:02 - 53:06)
Well, I think AI just, in my mind, it represents the next step in this. I mean, when you think about Wikipedia and you think about people who’ve taken material from Wikipedia and included it in a research paper as if it was their own crafting of a topic, AI just enables that to go a step further because it’s going to be much harder to distinguish where that material came from. So I think we need to set clear limits for students and talk about what is permissible and what’s not.
I have no objection to students using technology to help them do their job better, but I want them to tell me what they used to do it so that I understand what role they’ve done and what role the tool has done for them. And I think we’re going to have to think about, and this is going to be partly up to the leadership of the institution as well as to individual faculty: What are you going to permit and where are you going to draw the line between them?
[Rafael Bras] (53:07 - 53:18)
But that assumes that the individual faculty also has the know-how and the whereabouts to actually understand the technology, which is not obvious to me.
[John Hennessy] (53:15 - 54:36)
Yeah, no, I agree with that. I agree with that. And I think, you know, we can build tools that determine whether or not something, or with some probability, determine whether or not something was AI generated and use those kinds of tools at least to ... But we may also have to change the way we interact with students. So we may see a return to oral exams, for example, as a way of avoiding the problem of an AI response to an exam.
So I think we’ll have to navigate our way through this. I mean, in some sense, you can’t say, don’t use the technology. That’s like saying don't use calculators, right? I mean, remember when calculators first came out, they said whether or not you could bring a calculator to an exam. Come on, guys. (laughter) And if you think about these AI systems, they’re going to be incorporated in your word processor.
So the word processor you’re going to use is going to have an AI tool built into it that’ll help you improve your writing. So you can say, you know, structure this paragraph better. I mean, is that legitimate? I think it probably is a legitimate use and certainly will benefit students who perhaps whose English language skills are not as good as other students’ in the same way that a calculator may help students to different amounts depending on how good they are at math.
[Rafael Bras] (54:37 - 54:55)
So what advice would you give a university leader? You have a new president at Stanford. What advice would you give a leader in terms of getting the university ready to keep up if not move ahead of this AI revolution?
[John Hennessy] (54:57 - 55:40)
Well, I think the advice I’d give them is: Think about what role you want this technology to play. Be absolutely clear with students, applicants to the university, faculty, what role you want. In some recent work we were doing with some colleagues to look at the role of AI in science, one of the things we said is that people need to talk about — in their paper — what AI tools they used in the pursuit of some scientific result so that it’s crystal clear what it is, where the model came from, how it was trained, and the dependability of that model and the insights that the model got.
[Rafael Bras] (55:42 - 56:34)
Wonderful. John, I’m going to move a little bit to more personal questions here, which I think you’ll find interesting. I certainly, in thinking about you, these are to me always important.
You’ve had an extraordinary career with global consequences and impact, yet as I said earlier, you are the consummate professor. You just exude the love for the students, right? You are there, your satisfaction is there, yet you managed to do all this while being a very good family man. You’ve been married, I believe, for 50 years now. You have, I think, two children and grandchildren. How are you doing?
[John Hennessy] (56:35 - 57:38)
I don’t sleep that much. (laughs) I can survive on six and a half hours sleep. My view has always been to live life to the fullest and to try to make the most of things and realize that over a lifetime, there’s an arc of how things develop.
There’s a time when my children were young where basically my work schedule would end at 5 or 6, and we’d break for dinner, bedtime, story time, things like that, and I’d go back to work at 8 or 9 at night for a few hours, and that worked fine. And so you can find ways to adapt.
I think it’s also a question of choices. What are your priorities — and pay attention to those. Learn to say no. Don’t let yourself be overwhelmed by things so that you can’t do a really fine job on the things that really matter.
[Rafael Bras] (57:39 - 58:08)
That’s tremendous advice. I really, really like that. Let me go back to that element of luck and ask you a question that I always ask a lot of people. Imagine that somehow you get transported back when you were a teenager and you were in high school. Could you do it again?
[John Hennessy] (58:08 - 58:53)
(Laughs) I think there are so many little twists and turns in life that it’s so unpredictable. I think a lot of the places where I made choices, I made the choice I really wanted to do. I dreamed of, since I was an undergraduate and got involved in both some teaching experience as well as some research, decided I wanted to be a faculty member. And that's a job I've always relished. And while I enjoyed helping a startup get going, I came back to the university because I missed engaging with students. And I still find that to be one of the high points in my life.
[Rafael Bras] (59:03 - 59:13)
I already mentioned your love for golf. I’m not going to say which one of your students also questions your golf.
[John Hennessy] (58:08 – 59:06)
(Laughs) All of them.
[Rafael Bras] (59:07 - 59:12)
Let me address another passion of yours. I understand you’re a gardener.
[John Hennessy] (59:13 - 59:29)
Oh, I am, I am, I am. We probably have, sitting in our kitchen, probably 60 tomatoes from our plants. (laughs) We’ve got about two more weeks in the California season you can get tomatoes, and then that’ll be the end of it.
[Rafael Bras] (59:31 - 1:00:00)
I live with a gardener. My wife is an avid gardener, and she loves everything that has to do with dirt and plants. And one thing I’ve learned about gardeners is it requires infinite patience, and I have yet to meet a satisfied gardener. There’s always pests or bugs or something happens, and you’re always unhappy. That’s not you.
[John Hennessy] (59:59 - 1:00:35)
No, no. Actually, we’re lucky we live in California where there are fewer pests. So that probably helps. And I focus on the things that really are worth the effort. So tomatoes off the vine are worth the effort because they’re so magical, right? And I grow a lot of tomatoes as a result. So I end up freezing particularly little cherry tomatoes.
You can just freeze them directly because we have so many. So I’ve got about 2,000 cherry tomatoes that I’ve frozen this year.
[Rafael Bras] (1:00:37 - 1:01:18)
I have to say I find it fascinating because as I know you and everything you’ve done and your time commitments and your zest for life and excitement and challenges, yet you are involved in sports that require patience and hobbies that are the ultimate of patience demanding. So it’s really, really interesting.
All said and done, what would you see as your legacy? What would you like to be remembered as?
[John Hennessy] (1:01:21 - 1:02:11)
I’ve got a career, as you said, that spans many things. So, successful innovator. I think as a university president. I’d like to be remembered as somebody who tried to do the best job he could for the students and the faculty at the university.
I’m an eternal optimist about the ability of young people to really make a difference in the world. And that was really the motivation behind creating Knight-Hennessy. And I think we’re investing in a set of people who I think over the next decades will pay back many times over the investment we made in them.
And so that’s what I’m most proud of, I think — having made a change.
[Rafael Bras] (1:02:12 - 1:02:28)
Wonderful, wonderful. Well, John, let me say I thank you very much for not saying no to my invitation. I have enjoyed this tremendously.
Thank you so much. I’m sure many people will benefit from this conversation.
[John Hennessy] (1:02:28 - 1:02:31)
Thank you, Rafael. It’s been a joy to talk to you.
(1:02:35 - 1:03:08)
Thank you for tuning into this episode of Not Alone: Leaders in Conversation. We hope this discussion today sparked new ideas and left you with plenty to think about as you continue to lead in your own institution. If you found this conversation interesting, insightful or thought-provoking, please share this episode with your colleagues, peers and friends.
Leadership in academia can feel isolating, but through shared experiences and insights, we’re never truly alone. And don't forget to subscribe to the podcast on your preferred platform so you can stay up to date with our latest episodes.