Dr Irving Wladawsky-Berger is a Research Affiliate at MIT’s Sloan School of Management, a Fellow of MIT’s Initiative on the Digital Economy and of MIT Connection Science and Adjunct Professor at the Imperial College Business School. He was with IBM for 37 years to their internet strategy. He has been an advisor to Citibank, HBO and MasterCard. In this wide ranging conversation, Irvin talks about life in year 2050, predictions Vs decision making in Ai, how human limitations make up for technology adoption, why not to conceptualise Ai in human terms, intangible investments in research and Imagination as a critical leadership attribute. Read on his blog.
Podcast Transcript
0:00:05.6 Intro: Welcome to Prepared with your host, Hari Abburi. Join the conversations that will prepare you for a future yet to be discovered.
0:00:24.2 Hari Abburi: Hello everyone. An absolute pleasure to introduce Irving Wladawsky-Berger. He is a research affiliate at MIT Sloan School of Management, fellow of Initiative on Digital Economy and of MIT Connection Science initiative. He spent 37 years at IBM defining strategy for the internet era and has been involved in several strategic initiatives, including shaping the internet division of IBM in those days. These initiatives included internet supercomputing, Linux, which as all of you know, was introduced in 1991.
0:01:07.6 HA: After IBM, he’s been an advisor on Digital Strategy Innovation at Citigroup, at HBO, and MasterCard. He’s a frequent blogger. One of the interesting things I learned about him is that he started blogging in the year 2005, and has never failed to miss a blog till date. He blogs almost every week or every 10 days. And he is a frequent contributor to Wall Street Journals, CIO Journal. Irving, welcome to the show. Thank you so much for joining me.
0:01:40.3 Irving Wladawsky-Berger: Hari, my pleasure to be with you.
0:01:42.9 HA: I wanna take you way out into the future in 2050. From everything that you know today and possibly everything that we do not know today, what does human life look like in 2050?
0:01:55.9 IWB: That’s a very good question. And when looking at a non-predictable future, whether it’s unpredictable, because as you just asked me, it’s out in time several decades or as we are now living through the coronavirus pandemic, it’s unpredictable because there are so many issues surrounding what may happen in the relatively near future. But when looking at an unpredictable future, it’s important to sort of look at a number of scenarios rather than giving a definitive answer because there is really no definitive answer.
0:02:44.8 IWB: So at one extreme of what life might be like in 2050 is the world that Ray Kurzweil wrote about maybe 15 years ago of the singularity, where he said that given the exponential advances in technology, computers will become so intelligent. And I believe when he first wrote about the singularity, he said, by 2045 where computers will far surpass human, computers will then keep developing their own more advanced forms of computing, and we will be into a whole different world and that’s one extreme. I think that at this time, very, very few people whose opinions I trust, believe that there is any kind of reasonable probability of such a future. I haven’t even heard Ray Kurzweil talk about it, so I don’t know if he himself believes that.
0:04:02.0 IWB: So let me go to the, I don’t know if it’s the other extreme, but it’s what I think of as a more pragmatic view of the future. And the way I look at it, is that humans have been developing tools for a few hundred thousands of years, starting with the stone axes. And every time we develop a new tool, things keep advancing and… We developed the stone axe and that allowed us to hunt and by hunting, we were able to increase the size of our brains because we were able to capture bigger animals, eat meat, on and on and on and on.
0:04:50.8 IWB: Then we learned how to develop levers and various stones then came the industrial revolution and came different kinds of machines, steam vapor, electricity, and on and on and on and on. And in many ways, I think of AI as another tool, a very, very, very advanced, sophisticated tool, but let’s remember that in its day, electricity and the internal combustion engine were very, very advanced tools, let alone computers later, and the internet in the 1990s.
0:05:48.6 IWB: Now, by calling the internet and other advanced technologies a very advanced tool, I’m not taking anything away from how much they will transform life, but it helps me put it in the context that the transformation is likely to be on the order of transformations we have been living through at least since the industrial revolution. Some things will happen that we didn’t anticipate, and some things that we have predicted will likely not happen. One of the major predictions was a hydrogen-based energy, that because of hydrogen power, we would have so much energy at incredibly low prices and fission energy like hydrogen… I’m sorry, fusion energy like hydrogen is totally clean as opposed to fission energy from a nuclear power plant, so that would have been fantastic.
0:07:13.8 IWB: Well, we don’t have any hydrogen nuclear energy 55 years later, and we don’t know when it will happen. On the other hand, Isaac Asimov did predict that we would have something like the internet, and something like very big screens in our home that we would be able to watch and so on. Some people predicted that our cities will be covered by big bubbles to insulate them from the weather. Some people predicted that we would have flying cars, none of those have come to pass. So we need to have a little humility in making predictions. Some will happen and surpass us, others will not.
0:08:07.3 HA: In one of your interviews, you described that we use technology to actually make up for our limitations as a human race. One of the examples you talk about is, we made up for physical deficiency by building railroads, cars, airplanes.
0:08:22.5 IWB: Right.
0:08:23.1 HA: Then you talk about cognitive deficiency, where we cover it up by bringing information and data together. Now obviously, we are in the age of AI. Obviously, it’s rapidly multiplying itself. And so what deficiency are we making up today as a human race by adopting AI?
0:08:42.9 IWB: Well, the key… And let’s keep calling it the fish in sea, although maybe that’s not the right term. But the key augmentation that AI gives us is the ability to analyze vast and vast amounts of information of all types, not just structure information, not just text but videos, images, voice and so on, and extract insights out of all that information and then translate those insights into predictions of what may happen. And because of the huge advances we’ve had in the last 15 years in gathering so much data, the big data era, and because of the sophisticated algorithms like machine learning and deep learning, and because of the extraordinary advances in the power of the computing technology that we use to analyze all that data, the cost of making AI-based predictions has kept going down.
0:10:14.3 IWB: But now, there is a difference between a prediction and a decision. The tools can tell us different probabilities of what might happen, of what is likely to happen. It can also increasingly let us link together different data sets, different ways of thinking about a problem based on different kinds of data to make a broader kind of prediction. However, in the end, it is the humans who make the decisions in my mind. It is the humans who then take all of the various predictions that the AI technologies are making and translate them into a decision.
0:11:18.9 IWB: Let me give you an example. One of the most exciting uses of AI is to help radiologist detect potential growth of cancer in a way that they couldn’t have done with the naked eye. And it could be that a number of the mammograms that developing cancer had such tiny growth in them that the radiologist couldn’t have seen them. Whereas, now, with a properly trained machine learning algorithm, the algorithm can point the radiologist to say, “This particular mammogram has this kind of probability of developing into cancer.”
0:12:13.3 IWB: However, before the radiologist tells the woman what he thinks or she thinks is the probability, I would hope that the radiologist is now going back to his or her experience on how well the algorithm works and maybe has consulted with the other members of the medical team and is making the human decision in such a critical area as to how to proceed. So our tools, just like with cars and airplanes and the internet, give us incredible powers we didn’t have, but we have to be careful to rely on the tool 100% as opposed to look at the tool as a way of aiding the very important human decisions.
0:13:21.2 HA: Really fascinating, and I cannot help think about Boeing as a company and the crisis they’re in. In one of the articles, the article that you talk about, the coming era of decision machines, you highlight those difference that you’ve mentioned between predicting decisions versus decision-making itself.
0:13:46.1 HA: So in the MAX 737 crisis that we have, the plane was trying to make a decision on its own, and the experience that you talk about, which is the pilot, was trying to override it unsuccessfully. So obviously, there’s been a conflict there in terms of how the entire augmentation of the man-machine or the human-machine mix has been achieved there. What are your views on this?
0:14:14.5 IWB: Yeah, listen. From everything I have read, the issue with the Boeing 737 MAX was the assumption, and whether it was made because they just made a mistake, or whether they tried to rush the plane by making overly optimistic assumptions, other people will know more about that, but there was an assumption that the pilots didn’t need extra training to know how to deal with the new, let me call it, AI-based capabilities that the MAX had to be able to react on its own to different changes. And they made the assumption that because the MAX was an evolution of the 737, the pilots didn’t need to go through the lengthy and expensive training that you would normally do when introducing a new capability.
0:15:42.6 IWB: And Hari, from what I’ve read, pilots who’ve been properly trained in how to react and override what the plane was doing, would have been able to control the plane. This is based on what I have read. Whereas, if the pilot wasn’t trained, they were not able to, on their own, figure it out, especially in the middle of a life-and-death crisis where they’re trying to see what was going on. Now in engineering, especially in mission critical engineering, we expect engineers to be very fatalistic. An optimistic engineer is a bad engineer.
0:16:46.3 HA: In all of our conversation, you emphasized the critical role humans play, even in a very highly AI-driven environment.
0:16:55.4 IWB: Especially in a highly AI-driven environment, if I may add.
0:17:01.0 HA: Perfect. But in a recent write-up that you do in Wall Street Journal, CIO Journal, you also talked about why conceptualizing AI in human terms could be misleading. Could you explain that a little more?
0:17:15.8 IWB: Yes. If you look at the areas where AI has been most successful, like helping to recognize natural languages, machine translation, playing championship level Go, which people thought would take more years, and it was achieved by the AlphaGo algorithms of Google just a few years ago, or detecting very tiny emerging cancers in a mammogram, if you look at those examples, those are very focused, detailed examples in a very specific area. And the better you train the AI algorithms, the better they will do in a very specific area. Humans can be very good at specific areas, however, what distinguishes humans from AI algorithms is our general thinking and problem-solving capability, a lot of which comes from what we call common sense, which means that from very young, we learn how to deal with the world and we see a variety of issues in the world.
0:19:06.0 IWB: Now, if you ask me, how do we learn that? What allows a six months old baby to start learning languages and to start recognizing different animals and so on if the baby’s brain hasn’t been trained in the language? Well, as it turns out, [chuckle] our brains have had millions of years of evolution behind it. And I say millions of years because even though modern humans may be only a few hundred thousand years ago old, our antecedents, the primates, different levels of human go way back, and those ancestors of ours that made it, that were not eaten by saber-toothed tigers and were able to hunt and were able to find mates and raise children and so on, their brains had to evolve to be able to do that. And so, while AI has to be trained, our human brains have competed very nicely through evolution for many years, and that gives us an incredible capability of general problem-solving, of interaction.
0:20:49.9 IWB: In the article I used as a basis for a blog about the difference between AI and human intelligence, the article said that, okay, you can train through machine learning an algorithm to recognize dogs, and that will work very well. But then they found that if you point the algorithm at pictures of a cloud formation that more or less had the shape of a dog, and you know, cloud formation sometimes have different shapes, the algorithm will get confused and may say, “This is a dog,” because it’s consistent with its training. Well, no human would confuse a cloud that looks like a dog with a real dog, and that’s part of the huge difference between humans and AI for the foreseeable future.
0:22:02.2 HA: I appreciate it, Irving. In my view, I believe that every business is a platform, technology, and a data business, and these need very tangible investments. There’s allocated capital, then you have marketing investments, you have people, technology, all of that coming in. You talk about making intangible investments. What do you mean by that?
0:22:27.1 IWB: Well, let me talk about one of the most intangible investments of all, which is investments in research and investments in the kind of talent that conducts research. Let me explain. Usually, what really good researchers do is they try to analyze the major trends in technology, the major trends in market, the major trends going on around them, and then try to make predictions and decisions of what is likely to happen in the near to farther out future. Now, just as an example, from my past, as you may know, IBM almost went out of business in the early 1990s because of a major technology transition that made mainframes almost unaffordable because previously they had been designed with very unique expensive technologies, whereas now, microprocessors have gotten inexpensive and powerful enough that they could compete with the mainframes. Now, as it turns out, IBM researchers in IBM’s R&D labs had anticipated these changes, because technology changes don’t just show up one day and you say, “Oh my God,” major technology changes, you can anticipate if you’ve been watching what’s happening in technology.
0:24:38.3 IWB: I think of that as asteroids. Asteroids don’t just fall out of the sky one day. If you’ve been looking at the sky, you start noticing something that doesn’t quite look right, and eventually, you say, “Oh my God, it’s an asteroid that’s coming. This is its size, his is what it might hit, etcetera, etcetera, etcetera.” And this often happen in months, if not years, before the asteroid hits. Well, that’s the case with researchers and major events. And when you start anticipate a problem, then you can start preparing for it. And as it turned out, in the case of IBM, because we anticipated this transition, our research community had been simulating a new family of mainframes that was based on microprocessors and parallel architectures, and then, we were able to switch to these new models of mainframes and continue existing, whereas quite a number of other companies went out of business because they didn’t anticipate and prepare for the transition. So, research… Investments in research and investments in the talented people needed to conduct such research is one of the most important intangible investments in an era of rapid changes.
0:26:30.3 HA: Another intangible that I firmly believe in, is imagination. Imaginative companies have not just solved problems but have created new solutions that we never thought were possible. And I place a significant premium on imagination as a key leadership attribute. I remember you mentioning an example where an engineer somewhere near Stuttgart also put up Linux into the mainframe itself, and that changed a lot of the conversation around. So what are your views on imagination? Would you agree with me on putting a premium on leaders having imagination?
0:27:10.4 IWB: Absolutely, Hari. And in fact, let me tell you an anecdote about those engineers in Germany who on their own developed Linux on a partition of IBM’s mainframes. I remember being in that lab. The lab is in a town called Böblingen, which is near Stuttgart in Germany. And I was having lunch with a gentleman from a German company… And you know, SAP actually. I was having lunch at SAP’s in Germany. I was having lunch and I was mentioning the work on Linux on mainframe. This must have been in the early 2000s when Linux was still relatively new in the commercial world. It already been widely used in the world of research, supercomputing, and the internet, but we were just bringing it to the commercial world. And this gentleman said when he had heard that these people were putting Linux on the mainframe, and they didn’t get permission from, I don’t know, headquarters or wherever, he was worried that they would be fired because how dare they on their own put Linux on mainframes.
0:29:03.0 IWB: And I said to him, “Are you kidding?” Not only were they not fired, we consider them heroes because once we made the decision that IBM would embrace Linux across all its platforms, the fact that we were doing that, not just on personal computers and UNIX workstations where it was a more natural thing to do, but in our mainframes was one of the most important business decisions that we could be making. Had they not done this work, it may have taken us months, if not maybe a year to figure out if it would work. Because of their work, we immediately had the prototypes ready to go, and we were able to quickly get it to market, and it wasn’t long before Linux was about 25% of the market of mainframes, which is a significant portion of it. Now, this is what really good imaginative talented people do.
0:30:32.0 HA: Thank you, Irving. I have always believed that the best ideas come from conversations, and your examples have reinforced my belief in that. Thank you so much for being on the show. I’ve thoroughly enjoyed listening to you on a wide range of topics. We’ve covered talent, imagination, IBM, decision-making, predictive analytics and I’m sure our listeners will find that fascinating too. And when next time in New York, I promise you a good jazz evening, and look forward to catching up with you in person. Thank you so much.
0:31:10.2 IWB: Hari, I look forward to sharing a meal with you in New York or LA in the near future.
[music]
0:31:21.6 S1: Thank you for listening to Prepared with Hari Abburi. For more episodes and additional resources, visit PreparedWithHariAbburi.com. Let’s continue this discussion online. Follow along on Twitter @HariAbburi.