The AI Revolution-1: MIT's Rama Ramakrishnan Explains the AI Transforming Business and Society

I think AI is going to be one of the most profound transformations humanity has ever seen. The impact of AI on jobs is a very hotly debated topic and it's obviously of great interest to every single one of us. While it's very difficult to predict what the impact of AI on jobs is going to be with any degree of accuracy or certainty, I do think that there are a few sound economics principles that we can actually apply to make sense of what's happening and what's likely to happen. Our guest today is Rama Ramakrishnan. He's a professor at Sloan School of Management at MIT and has been a serial entrepreneur. We are going to talk to him today about AI. Rama, welcome to our studio. It's great to be here, Upendra. Thanks for having me. Sure. So let us start right away. How real is AI? You know, I think AI is both overhyped and underhyped all at the same time, in my opinion. That is the word I was looking for. Yeah. Yeah. I think overhyped because obviously everyone has essentially stapled AI to their whatever they are doing. And they're making a lot of marketing promises. On the other hand, I do think it's underhyped. I think it has got incredible transformative potential. I know that it's often compared to the internet or electricity or the industrial revolution, for example. I actually think that it's more profound than all of them. And there are a couple of reasons why I say that. First of all, if you look at the internet, it made communication, it made sort of the acquiring of information, searching for information much more efficient. AI, and of course, going back to the industrial revolution, electricity, it made sort of essentially manual muscle power. It made it, it amplified it. It removed the need for much of it. It gave us the ability to do physically hard things that we couldn't do easily before. So first revolution made physical things easy. The second revolution makes communication easy, makes information acquisition easy. AI, I think is interesting and more powerful because A, it actually piggybacks on all those things, right? It actually is sort of standing on their shoulders. That's very interesting. Yeah, that's the first reason. So it has the potential to be much more transformative for that reason. Secondly, perhaps even more importantly, I think AI actually makes intelligence much more cheaper, much more easily available. And if you look at how us Homo sapiens have sort of, you know, essentially conquered the planet and we are able to, you know, send people into outer space. And frankly, and I love this phrase, I just read it somewhere, you know, we have learned how, we have figured out how to take sand, which has, you know, silica and turn it into something that can think, right? Computers that can think. That's a wonderful analogy. I loved it when I read it recently. And so now all of it stems from the fact that we are intelligent. And so now we have the ability to actually have this, you know, external source of intelligence. So I think it's going to be a dramatic amplifier on what we can do. So for all these reasons, I do think that AI is perhaps the most profound transformation that we have ever seen. Rama, let us talk about AI from the consumer point of view. How real it is for a consumer and how much is all the marketing aspects and hype about this? What impact is it really having on the consumer and what are the issues consumers should be aware of? Yeah, I think, first of all, I want to say that it is already having a massive impact on people. Now, I use AI a lot, partly because I work in the space and I teach in the space, but I've spoken to lots of other people. I've seen a whole bunch of reports out there as well. Clearly, more and more people are using AI every day. Now, I think that a lot of people, for example, are using it to answer healthcare questions they might have. Obviously, health is very important to all of us. So it ends up being a very, very sort of common use case. I think people are using it for finance, people are using it for all kinds of things. So it's clearly taken the consumer world by storm. That much is very obvious to me. Now, that said, I think people have to be a bit careful in how they process the results of whatever the AI is telling them. As people might know, ultimately, chatbots, large language models, AI tools, they are probabilistic in nature, meaning that they are essentially trying to predict the next word one at a time, but then they do it so well. And because of a whole bunch of underlying factors, the end result ends up being very, very often, very coherent, logically consistent, very tight and correct, and so on and so forth. But the thing is, it's not guaranteed to be correct 100% of the time. So when you look at something, it looks really good, but it actually may not be correct. And so if you're actually seeking health advice, for example, and if you see something that looks like as if a doctor wrote it, how can you possibly be sure that it's in fact correct, and not something that looks correct, but may not actually be correct? I remember reading about this study that was done in the US maybe a couple of years ago, it might still be ongoing, where they were using AI tools to transcribe the doctor-patient conversation during a visit to the doctor. And then it will create a nice email note that the doctor can email out to the patient. Clearly a good thing, because it saves the doctor effort in having to write it and so on and so forth. But it turns out, apparently, the draft coming from the AI system had errors, which would have caused serious harm to the patient 7% of the time, if the doctor had not caught it and fixed it. So in general, I think one has to be very cognizant of the fact that the output from these systems are almost always very often correct, but not 100% correct. So Rama, what is the impact which a consumer is going to feel or a member of the society, as far as the jobs are concerned? There's a lot of already nervousness. Many people are saying that I'm going to lose my job. Some have already lost their job. Yeah. You know, I think this is obviously an incredibly important and very controversial topic. There is no shortage of views on the topic. You look at media out there, and you look at all these experts opining one way or the other, and they can't agree with each other. So I think it's a very complicated question. And it's very difficult to come up with a very definitive answer right now. Part of the reason why we can't actually make a very clear prediction with a lot of certainty is because it all depends on how companies use AI in the months and years to come. And the way they make those decisions is going to affect the economy and jobs and so on. So all those unknowables are actually in the future. That's one of the reasons why we can't predict it. But that said, I think that maybe there are a few, what I think of as sort of economics, one on one common sense ideas that I think one can keep in mind, as one sort of reads about and hears about all these things. And as one thinks about one's own job, right? I find those to be personally useful when I think about jobs, and perhaps others might find them useful as well. So I'd be happy to share them. So I guess the first thing I want to say is that when you're looking at a particular job and asking yourself what is the impact, likely impact of AI on the job, don't think of the job as a whole, monolithically. Always break it down into the individual tasks that make up the job, right? So for instance, I mean, I'm a professor at a business school. I do lots of different things in my job. Some of those jobs, some of those tasks, if you will, are actually AI automatable to some degree or the other. Others are not. So for instance, maybe I want to create a bunch of questions for a particular exam. And I want these questions to be roughly similar to past year's questions, but different in a few different ways. And I can probably get a pretty good draft from an AI system to come up with those questions. So that's sort of automatable, partly automatable. But then when a student comes to me for office hours, and I know that they're struggling with something, very often, I have to sort of figure out the body language and just this human empathy element to figure out how I could be helpful to them. And I don't think that's easily automatable using AI. So the point is that every job is a bunch of tasks. And so you need to ask yourself, what are the tasks in this job? And then look at each of the tasks and ask yourself, okay, is it automatable by AI or not? So now if you look at it in that lens, the task lens, not the job lens, then let's take an extreme case where maybe there's a job in which every task in the job is automatable using AI. I'll give you an example from a past technology development, which is elevator operators. I don't know if you remember in India, when we were growing up, there'd be a person sitting who would operate the elevator. And that was true in the US as well for many, many, many years. But we don't see any elevator operators now. Because if you look at the elevator operator's job, that basically every aspect of the job, every task in the job was automatable using that technology. And when everything in a job is automatable using a technology, it's going to disappear. That's one extreme. But fortunately, I think most jobs are a mix of different tasks. Some of them are automatable, some of them are not. So the real question is, which are the most important things in a job, automatable or not? That's really the key question. That's the key. That's the key question. So let's take a couple of examples. So one example, a lot of people are using these days is software engineering. Clearly, AI is making it very easy for us to build software. And so the fundamental thing you have to understand when you think of AI's impact on a particular task is that if AI can automate a task, it means that the cost of executing the task has become much smaller now. The cost of doing something has become much smaller. When the cost of doing something becomes much smaller, sometimes people demand a lot more of it. So if you take software, for example, a lot of people believe, and I do agree with this belief, that most companies out there, they would love to use more software to help run their businesses. But previously, they couldn't justify the business case for it, because it is just too expensive to write and build the software and maintain it. And compared to the benefit of doing that, it just wasn't worth it. But now, the benefit is modest, but the cost has dropped quite a bit, which means there is an ROI to be made. And so a lot of companies are like, you know what, I'm going to start writing more software, building more software to help automate more things in my company. Or if you're a startup, a software startup, maybe you wanted to build three other software products, and you never did that, because it wasn't viable for you. But now using AI, you can build it more cheaply, therefore, maybe you build more products. So these are all examples in which, because something has become much cheaper to do, there is going to be a lot more demand of it, because it has become newly affordable, or newly economically viable for a lot of people. So if you're lucky, your job might be in one of those categories, where because it has become, some aspects of it have become cheaper due to AI, there'll be a lot of latent unmet demand that rises up and says, hey, I want that. So this, by the way, is called Javan's paradox, where when the cost of some key task or input becomes much cheaper, there's sometimes a lot of latent demand for it. And when that happens, the overall number of jobs will actually go up as a result, right? Now, just to be clear, not every task has this characteristic. So you need to be a little careful. The classic textbook example from econ is salt. Let's say, for example, I tell you, hey, Upendra, the cost of salt, thanks to some revolutionary new desalination technology, has dropped by 90%. Salt is way cheaper now than it used to be. We are not going to suddenly start consuming more salt, just because it becomes cheaper, right? So I know that's an extreme example. But the general point I want to make is that some kinds of tasks and jobs, maybe are subject to this Javan's paradox dynamic, in which case, I think the number of those jobs is going to go up, right? Can you also explain that paradox, so a person who is not familiar can understand what exactly? Yeah, I think the basic idea is this, that, you know, you're doing something, and there's a key input that goes into doing that thing. And let's say it becomes much cheaper to do that. Now, what that means is that the overall thing you're doing has suddenly become cheaper, right? So you would think that, well, I used to do 100 of these things before for $100, a dollar each. Now I can do the same 100 for just, say, 10 cents each. So it has gone from $100, I can do just for $10, right? So now what do you do now? Do you take the extra $90 and just put it in your pocket? Or do you say, you know what, it has become so much cheaper, and I'm going to actually start selling it for not 100 bucks anymore, for maybe for $11, right? Now, suddenly, a lot of people who could not afford the $100 thing you had, they're like, oh, wait, you're selling it for $11? I want it, because I've always wanted it, I just couldn't afford it. So a whole bunch of demand is going to show up, because it has suddenly become more affordable for them. And now you start selling more and more, and it turns out now you have to employ 10 times as many people. So that is, it's called a paradox, because initially the thinking was that if something became much cheaper, people would just use it more efficiently, and the overall pie will not grow. But as it turns out, if something becomes cheaper, if there's a lot of latent demand for it, the pie will, in fact, grow as a whole as well. So that is why it's called a paradox. Can you use the example of like, you know, all these retailers have this, you know, the 50% discount, 60% discount, bring, is that something to do with? I mean, I think you are referring to the fact that like one of the basic ideas in economics is like the demand price curve. The idea being that as the price becomes smaller and smaller, the demand is going to go up, right? It's a downward sloping demand curve. And basically, the Jevons Paradox is just like a way to illustrate that point. Namely, for certain kinds of goods and services, if the price drops, the demand actually increases significantly. The technical term for it is elasticity. So if the elasticity is at least one, or greater than one, for example, then what's going to happen is the total output is going to increase as well. And therefore, the total employment actually may increase. That's really where it's coming from. So it's related to the whole price demand curve. So now, one question I have for you is, again, my experience here, we get a lot of press releases for our various publications. So this was late last year, I got one press release. And I started to read that. And I was amazed, totally blown away by the quality of the press release. The writing, the style, the language, the codes, it was totally flawless press release. And I was so impressed, so that I ended up calling the marketing person who had sent me the press release. I said, look, I have never seen any press release like this, because there is... And she said, thanks to AI. So I was very impressed. And actually, that was the one thing which really intrigued my interest in AI. And I started to really follow seriously and do some experiments. But just on that point, though, I think it's important to understand that when you take a job and break it into tasks, and look at which tasks are automatable by AI, and so on, and you apply this Jevons paradox argument, and try to figure out if this, what I'm doing, if it suddenly becomes a lot cheaper to do what I'm doing, will people want more of it or not? That's the fundamental question. If people are going to want a lot more of it, then my guess is that those jobs are going to increase in size, increase in number. That seems to be kind of true for software. It seems to be true for lots of other things. It's probably not true for a whole bunch of other things. It just depends on the case by case basis, I want to say. The other final thing I want to point out is that this whole, even if the Jevons paradox thing is true, it might take a little bit of time to play out. And so, for instance, some companies may say, you know what, before I needed 10 people to do whatever they are doing. But now with AI, for example, let's say you have 10 people writing press releases. Now I can actually have just two people crank out the same volume of press releases. I don't need 10 people anymore. And so those companies may decide, you know what, I don't need these extra eight people. So there might be a job loss. So here is a paradox, another issue. So this happened two months later. I'm getting so many press releases. And all the press releases are very similar. Similar style. I could look at them and say, oh, this is written by AI. Exactly. So again, the question comes at how you distinguish yourself. And I think that's a great segue. So if some companies are going to be like, okay, I'm going back to the example, 10 people, I only need two people, I don't need eight people, I'm going to lay off eight people. So they have essentially reduced their cost with fewer people, but the same kind of output. Now remember, that tool is available to everybody. So their competitors can also do the same thing. But some of those competitors may decide, you know what, everything is going to look identical. It's not going to stand out. So I'm actually not going to get rid of the eight people. I'm going to take the same 10 people, but I'm going to give them AI tools so that their output is much more valuable. Yeah, that is the key. Correct. That's much more valuable than what we used to do. Maybe we can do new kinds of things, new kinds of products, new kinds of services, even existing products can be made much richer, and so on and so forth. And when that happens, those companies are going to start taking market share from the companies that are only doing cost reduction, in which case these companies will go out of business unless they also change their strategy. So they will start hiring again at some point. And which is why initially you might see like a bunch of job losses. And then as the competitive nature of the market comes into play, and this lack of differentiation becomes an issue that they need to address, then I think jobs will start growing again. So I guess what I'm saying is that there might be a near term sort of dip, but hopefully in the medium term, you'll start job increases as well, as long as the overall Jevons Paradox argument holds, which again, just to remind people, if this thing becomes cheaper, will people want more of it or not? That's really the key question. That's where everything hinges on. I guess the final thing I want to say, Upendra, is that you might have a situation where the number of jobs actually is going up for all these reasons. But it's possible that the average amount of money you can make from doing these jobs, the wages or the salary may actually go down in some cases. So you need to be a little thoughtful about it. And that's going to depend on whether AI, because it makes it easier for people to perhaps do the job, maybe it expands the labor pool. And there are many more people who didn't need to be as qualified as the previous people. Now they enter the labor force, because with AI, they can actually do the job, which means as the labor supply increases, the pressure on wages might increase. So there might be more jobs, but each job may not pay as well as it used to. All these things are possible. And so just to step back from all this for a second, now you understand why it's very difficult to look at the whole thing and say, oh yeah, we're going to have more jobs, fewer jobs, less jobs. It all depends on case by case what's going to happen. But I think the general thing I would recommend people to do is to look at their jobs and figure out what part of it is automatable, what part of it is special. And the special parts end up being things which involve human judgment, empathy, persuasion, things like that. And then essentially be very cognizant of the splits, and then ask yourself, okay, how can I make whatever I'm doing much more valuable using AI, as opposed to using AI to just make life a little easier? I think that's really important. The other thing I want people to remember is that I think if you might have jobs, this job has 10 tasks, this job has 10 tasks, companies may decide, you know what, three of the seven here is going to be AI-automatable, four of the 10 here, and then I'm going to take the remaining seven and six, I'm going to put them together to create a new job. So I think new jobs are going to happen. It's not just that existing jobs are going to just stay as it is or reduce. I think new jobs will get created, new kinds of recombinations are going to happen. All these things are going to happen. And I want to point out, we haven't talked about the fact that brand new jobs are going to get created as well. Like a great example is when the internet was invented, did any of us think that there would be a job called social media influencer? Sure. Right? Brand new job, didn't exist, nobody predicted it. So similar things are going to happen in AI, there are going to be brand new jobs. So that's sort of the overall picture. And also the social media, all the content creators, they have to have a content for social media, video, writing, all this stuff. So I think overall, some jobs will disappear. If they're fully automatable with AI, lots of new jobs are going to appear, and all the jobs in the middle, they're all going to get reconfigured in crazy and interesting ways. And the way they get reconfigured will depend on what individual companies do, and how they respond to competitive pressure in their industries from other companies who are trying using AI to do even greater things. And that's why you have to watch your industry and company very carefully as this thing evolves, so that you're best positioned for it.