A Stanford Professor on How AI Will Transform Productivity

ELISE HU: That was professor, researcher, and writer Erik Brynjolfsson. He has spent a long time analyzing the ways in which data know-how is reworking enterprise and the financial system. His analysis has more and more centered on synthetic intelligence, and he’s going to share his perspective on AI’s potential to remodel work. Here’s my dialog with Erik. 

ELISE HU: Thanks for doing this, Erik.

ERIK BRYNJOLFSSON: Good to be right here.

ELISE HU: Why don’t we simply have you ever introduce your self and what you do.

ERIK BRYNJOLFSSON: I’m Erik Brynjolfsson. I’m the director of the Stanford Digital Economy Lab. I’m a professor right here at Stanford within the division of economics and enterprise college, and primarily the Stanford Institute for Human-Centered AI.

ELISE HU: Well, all of us learn the headlines, and it’s clear that one thing actually fascinating is going on with AI proper now. So Erik, how would you describe in layman’s phrases this second and what’s occurring?

ERIK BRYNJOLFSSON: Well, there’s positively one thing massive occurring. I believe a whole lot of the thrill is on ChatGPT, and possibly DALL-E, and these are each examples of a brand new class of AI known as basis fashions. That contains not solely these giant language fashions that may write tales or poetry, electronic mail, advertisements, and lots of different sorts of textual content, but additionally, like DALL-E, they’ll make photos, there are others that may make movies, audio, and even write pc code. These applied sciences have the potential for actually reworking the financial system, I believe, creating trillions of {dollars} value of worth, however they will also be very disruptive.

ELISE HU: What’s occurring precisely to make these AI breakthroughs doable?

ERIK BRYNJOLFSSON: Well, it’s actually a confluence of three issues. One is simply much more pc energy, orders and magnitude, extra pc energy than we had 10 or 20 years in the past. The second is much more knowledge; which may be an important factor. Over the previous couple a long time, virtually every little thing has been digitized. And that gives the uncooked materials for these machine-learning engines. And final however not least, we now have a lot better algorithms. People have discovered new methods of utilizing these knowledge and making use of pc energy to them to reply questions that we couldn’t reply earlier than. 

ELISE HU: It is thrilling. It additionally, after all, results in so many questions, which then dovetails along with your profession, which is targeted on the financial affect of digital applied sciences. What would you say is the through-line that connects your analysis and inquiries and this specific second that we’re in?

ERIK BRYNJOLFSSON: Well, for a very long time, I’ve been keen on how computer systems are altering the world, since I learn Isaac Asimov and different science fiction. And after I went to grad college, my professor requested me to plot pc energy within the financial system, and each time I plotted it, there have been these exponential curves rising actually, actually quickly, which was method again within the Nineteen Eighties. But I might see, if this continues, something like that, simply astonishing issues are in retailer. And certainly, issues have continued just about alongside these strains, and we’re starting to actually change the world.

ELISE HU: What type of astonishing issues are bearing fruit? What are we seeing proper now?

ERIK BRYNJOLFSSON: Well, recently, I believe these giant language fashions, or basis fashions, have been simply very placing. They’re capable of generate new sorts of content material that beforehand solely people might do. I talked to the inventors of those applied sciences, the folks creating them, even they’re shocked at a number of the capabilities. So these emergent properties, having it perceive proper from fallacious, or be capable to create new sorts of insights, or to talk in several voices, even to write down pc code or play chess. These are issues they didn’t anticipate it to have the ability to do, however have been surprisingly good should you ask it the appropriate questions. 

ELISE HU: Okay, this after all, is the WorkLab podcast, so we wish to know the way that is going to have an effect on workplaces. What do you’re feeling like leaders in firms proper now have to know concerning the potential of AI.

ERIK BRYNJOLFSSON: Let me begin by saying that these applied sciences are racing forward at actually an unprecedented price. The previous couple of years have been some breakthroughs. And our organizations usually are not maintaining our abilities, our establishments, even our legal guidelines are falling behind. And in that hole between what the know-how can do or what the know-how calls for, and what our organizations and our human creations are doing, there’s an even bigger and larger set of challenges and an even bigger and larger set of alternatives. We want to shut that hole. And we shouldn’t try this by slowing the know-how, we should always try this by rushing up our adaptation. The modifications which are occurring proper now, they’re affecting virtually each a part of the financial system, and lots of of them are fairly totally different than what occurred within the earlier 10 years. In the previous, we had comparatively gradual rising productiveness; I predict sooner rising productiveness. In the previous, we had rising inequality; I believe these instruments can, and hopefully will, result in extra extensively shared prosperity. If we play our playing cards proper, the subsequent decade could possibly be a number of the greatest 10 years ever in human historical past. 

ELISE HU: What ought to firms be enthusiastic about? What ought to leaders be enthusiastic about to be able to adapt shortly sufficient or be agile sufficient?

ERIK BRYNJOLFSSON: Agility is essential. So you’re going to need to have individuals who perceive the know-how, but additionally perceive the enterprise wants, what the purchasers are searching for. And that’s a uncommon mixture. In specific, I believe what we have to have is a capability to alter our enterprise processes and our organizations, and never merely bolt on the brand new applied sciences to the prevailing method of doing issues. Very not often, is it doable to have a plug-and-play use of the know-how. In virtually all instances, the massive advantages come from doing new issues that we hadn’t completed earlier than. And that requires much more creativity on the a part of managers and entrepreneurs than merely saying, what are we doing now? And how can a machine change an individual?

ELISE HU: You talked about there’s nervousness about AI eliminating jobs due to its potential for productiveness features. And as a result of machines can typically be substitutes for human labor, this might additionally imply that staff might lose energy and change into more and more depending on those that management the know-how. But you’ve sketched out a unique imaginative and prescient, one thing known as complementary AI. Can you speak somewhat bit about that?

ERIK BRYNJOLFSSON: I imply, that’s most likely the most typical query I get. And the fact is, is that sure, there are a whole lot of potentialities the place the know-how can change some present jobs. But I don’t assume that’s the principle impact, and that’s not the principle alternative. The greater alternative is that this complementary AI. What which means is enabling folks to do issues that they hadn’t completed earlier than. In truth, should you look by means of historical past, most applied sciences have ended up complementing people relatively than substituting for them. The individuals who have been weaving material early on have been apprehensive that the spinning jennies would change them and drive down wages. It turned out they have been proper, the wages for these expert artisans did go down. However, most often, the wages of staff have gone up, as a result of most often, the know-how has amplified what folks can do. The method, say, a bulldozer permits an individual to maneuver extra issues bodily, or software program has allowed folks to have an effect on much more sorts of issues than they might beforehand, and that signifies that they improve wages. So over the previous couple hundred years, have wages gone up or down? Well, they’ve gone up about 50-fold. I ought to notice that it’s not inevitable, it doesn’t all the time occur—the previous 20 years have in some instances been a divergence from that nice pattern from the earlier 200 years. Many sorts of labor have really seen the hourly price go down, the place individuals who have a highschool schooling or much less are incomes much less in actual phrases than they have been a few a long time in the past. Those with faculty or skilled or graduate educations have seen continued will increase in wages. So we’ve had a divergence. I believe the massive problem for us going ahead is whether or not or not we will use these applied sciences in a method that creates shared prosperity and doesn’t have this elevated polarization or inequality. 

ELISE HU: What are a couple of actionable issues that we ought to be enthusiastic about, or enthusiastic about doing, to assist shut that hole?

ERIK BRYNJOLFSSON: Yeah, I’m glad you requested that query, as a result of I believe that is the massive problem for our society developing. The cause I began the Stanford Digital Economy Lab was to assist shut that hole, you understand, not by slowing down the know-how, however by rushing up our adaptation. And there are a selection of coverage issues we will do by way of investing extra in schooling and focusing extra on having folks ask the appropriate questions, be artistic, and fewer on the rote duties that the machines can do rather well. There’s an enormous function for technologists to rethink the best way that they develop the applied sciences. Alan Turing was an important researcher, and he got here up with this evocative thought of the Turing take a look at, which is, can we make an AI that’s so human-like, so much like folks that we will’t inform the distinction between them. And I believe that’s impressed a technology of technologists. Also, I believe it’s precisely the fallacious factor to do. In truth, it will probably lead us right into a entice, which I name the Turing entice. I’ve checked out it extra intently, and when you may have a know-how that imitates people, it tends to drive down wages; when you may have a know-how that enhances people, it tends to drive up wages. So we shouldn’t be making machines which are shut photos of ourselves, we ought to be making machines which are as totally different as doable from us and permit us to do new issues. It’s a unique strategy to know-how. And most significantly, I believe managers and entrepreneurs have to rethink the best way they’re utilizing the know-how. Don’t simply take a look at your present processes and assume, oh, how can I change this employee with a chunk of software program or an AI? It’s okay to drive down labor prices. I imply, it’s nice for us to have the ability to get issues cheaper. But there’s far more upside in doing new issues, or delivering issues in a wholly totally different method. That takes somewhat extra creativity on the a part of managers however in the end leads not simply to extra complete output and extra worth created, but additionally results in extra broadly shared prosperity since you’re protecting people as a part of the manufacturing course of and never changing them. And if all three of these teams—policymakers, technologists, managers and entrepreneurs—every pursue that type of path, I believe we’re going to have a number of the greatest years forward of us that we’ve ever had.

ELISE HU: How might it will definitely result in adaptation of human capabilities? Technology modifications us, proper? Like, I don’t bear in mind cellphone numbers anymore as a result of I don’t need to. I don’t actually learn a map anymore. I imply, these are clearly very reductive ways in which know-how has modified me, however in what methods would AI change who we’re?

ERIK BRYNJOLFSSON: Well, this isn’t the deepest level, however I discover that it’s already affecting the best way I assign homework and the best way that the scholars do their homework. As you’ve most likely heard, these instruments could make it very straightforward to generate an essay primarily based on a immediate, and the essay could also be fairly good. So a whole lot of professors, a whole lot of highschool lecturers, are questioning, how can I assign essays for college kids to write down if the know-how is simply going to do it for them? I believe the reply is it will probably and will change the best way we’re doing it. I imply, a couple of individuals are saying, so we’re gonna need to discover a approach to detect them and ban folks from utilizing it. I used to be disillusioned, one of many massive AI conferences even had a requirement that not one of the submitters—AI researchers—have been allowed to make use of these instruments once they submitted their papers. I believe that’s the fallacious strategy. A greater strategy is to, as you say, redefine what it’s that we’re anticipating from folks. And so, if after I’m educating my class within the spring, I’m going to inform the scholars, go forward and use the know-how, however I anticipate your essays to be that a lot better than the youngsters’ final yr. In truth, I’ve already put the questions by means of the ChatGPT and the opposite device, so I do know what a standard reply could be. That’s your start line. 

ELISE HU: What type of roles or professions do you see probably being most reworked by the AI that we’re seeing today?

ERIK BRYNJOLFSSON: That’s an important query. It’s a trillion greenback query. I believe it’s gonna have an effect on virtually all of us. You know, having labored with giant language fashions, I see that a whole lot of artistic work is tremendously being affected. I used to be simply speaking to a CEO, who was attempting to determine what the appropriate KPIs have been going to be. So he went to ChatGPT and had it counsel some primarily based on his firm’s objectives, and it got here up with an important listing. He stated he didn’t use them verbatim, but it surely was an important spur to doing it higher. I’ve seen folks use it to assist design new sorts of swimming pools, new sorts of songs, give you all types of artistic work. I’ve used it myself in a few of my analysis writing. It’s serving to folks in any respect elements of the spectrum, not simply the much less expert data staff that have been affected by earlier applied sciences.

ELISE HU: Does this concept that people primarily have to actually lean into that which makes people distinctive, the power to ask the appropriate questions, the elements of humanity that we now have, like perspective and shock that machines should not have… does this portend a change to the bigger labor drive and the financial system?

ERIK BRYNJOLFSSON: I believe there are some actually disruptive modifications coming to the labor drive and the financial system. And I don’t totally perceive all of them, I’m attempting to review them. One of my massive initiatives is to go extra in depth to a number of the modifications. But I see a number of the broad outlines. And I believe, as you recommended, asking the appropriate query is essential. There’s a brand new career known as immediate engineering, which is actually telling the language fashions what you need it to do. And it seems that, relying on the way you ask these questions, you will get higher solutions, extra correct solutions, extra insightful solutions, extra artistic solutions should you construction it the appropriate method. And which may be actually the place people can add essentially the most worth. One of the issues that I stated in a few of my books is that as these instruments change into increasingly more highly effective, which means, virtually by definition, that we now have extra energy to alter the world. And that signifies that our values matter greater than ever earlier than. So it’s time for us to assume extra deeply about what it’s we would like the world to seem like and the way we wish to use these instruments to reshape it.

ELISE HU: What do you suggest that leaders be saying to their groups, their workers, their workers who’re apprehensive about this and fear concerning the challenges or the existential risk that’s posed by AI.

ERIK BRYNJOLFSSON: I wrote one thing with Andy McAfee in a Harvard Business Review article a couple of years in the past, I stated that AI will not be going to interchange managers, however managers that use AI are going to interchange managers that don’t. And I believe that’s much more true at present. So my first piece of recommendation is to have everybody in your workforce get acquainted with these instruments—they’re type of enjoyable to play with. And so I used to be speaking to at least one government and he was declaring a full day the place he requested everybody in his firm to simply spend the time enjoying round with the instruments and get a way of what they’ll do for his or her jobs and for his or her firm. That type of familiarity goes to create a whole lot of new alternatives—determining the brand new issues that the know-how can do for particular person staff, not simply taking stuff off the shelf that’s already been developed by some start-up. And in some instances, creating issues that go on prime of it to make it extra helpful for specific enterprise wants that you’ve.

ELISE HU: When you carry up enterprise wants, that jogs my memory of the potential for AI to assist us cope with possibly a number of the extra disagreeable elements of labor. I’m speaking about elevated complexity and tempo of change, data overload, too many conferences, simply the drudgery or extra tedious duties. Where do you see AI serving to resolve a few of these ache factors?

ERIK BRYNJOLFSSON: Well, AI can do lots to assist resolve and in addition exacerbate them relying on how we’re utilizing them. But a method they might assist resolve them is that, simply because the instruments can generate new textual content from easy outlines, it will probably additionally go the opposite method round. You can provide it an extended article, you can provide it even a e book, and it’ll distill down the essence of it. And you’ll be able to have it linked to that report in such a method that while you wish to double click on, zoom in, on one piece of it, it’ll carry you to the related a part of the doc. So it’s like an extremely sensible analysis assistant, or possibly your self spending weeks going by means of stuff, so that you get a set of notes which are very related. This is all completed in an automatic method. 

ELISE HU: Yeah, I can see this actually helpful to regulation companies, proper, which are concerned in massive litigations.

ERIK BRYNJOLFSSON: Absolutely. I imply, there’s already, for some years, there’s been instruments to assist with doc discovery, sifting by means of and discovering key phrases or phrases. But now it will probably go additional and perceive the ideas which are in there and summarize them and even give you counter arguments as wanted.

ELISE HU: Amazing. This dialog jogs my memory of that well-known John Maynard Keynes prediction 100 years in the past that we might solely be working 15 hours every week. Because over time, in keeping with Keynes, because of machines and know-how and new concepts, folks would get extra productive. And the machines and the know-how would take over for lots of the extra menial or tedious duties that people have been doing. Why was that prediction fallacious?

ERIK BRYNJOLFSSON: Yeah, that’s an important quote, and I encourage anybody who hasn’t already completed it to go forward and browse his essay, “Economic Possibilities for Our Grandchildren.” And first, I ought to say he received lots proper. So his prediction was primarily based on the belief that productiveness development of a pair % per yr would proceed exponentially enhancing dwelling requirements for the approaching century. And it roughly has. There have been some ups and downs, however by and huge, our development has matched what his prediction was. The distinction is a failure of creativeness of all of the issues that have been created. So rich folks of his period, they lived in manors, and possibly every now and then went foxhunting, however there wasn’t lots else they might do with their wealth, and so it sat round. But now we now have all types of different devices and enjoyable issues you’ll be able to spend your cash on. And one other factor is extra sociological. I’m an economist, however I’ve come to acknowledge that lots of people get that means from their work, and relatively than retire, many individuals really feel like they wish to proceed to contribute indirectly. Or possibly they really feel compelled to do it. So for each these causes, folks have continued to work fairly a bit, regardless that our productiveness is vastly larger than it was when John Maynard Keynes wrote these phrases.

ELISE HU: So if that’s the case, and there’s all the time going to be this sense of human striving, and AI helps improve productiveness, then what’s going to people be doing?

ERIK BRYNJOLFSSON: Well, for a very long time, I don’t assume we’re gonna have any scarcity of labor to do. There’s no hazard of mass unemployment. When I go searching by way of healthcare, youngster care, elder care, cleansing the atmosphere, invention, artwork—there are such a lot of issues that people are uniquely good at doing. Perhaps an important one is what Pablo Picasso identified is, asking the appropriate questions. There goes to be probably a jobs high quality drawback of, are we going to get the wages paid proper? Are we going to have sufficient of the sorts of jobs which are actually rewarding to do? And we will work on that and do a greater job, I believe, but when we use the know-how primarily to enrich folks relatively than substituting folks, I believe we will have a state of affairs the place most individuals have methods to contribute to society and the know-how is amplifying these capabilities. And too many individuals, I believe, for a failure of creativeness once more, assume we’ll be producing the identical issues however with fewer and fewer staff. And that’s actually one approach to improve productiveness. But we will do lots higher by rising the quantity that we produce. And that may be completed, not simply by way of portions, but additionally by way of high quality, or new sorts of output, as nicely.

ELISE HU: How can of us higher perceive the AI potentialities which are with us now and possibly play with these instruments?

ERIK BRYNJOLFSSON: Well, greater than two years in the past now, I used to be requested to touch upon a paper at a convention with the National Bureau of Economic Research. And I occurred to be the final speaker, and it was a convention about AI and the way it was altering the financial system. So I assumed I’d not simply speak the speak, however stroll the stroll. I put it by means of GPT-3, and I requested it to assist with my remarks. And I stated, do it within the type of Erik Brynjolfsson, and it got here up with some fairly good sounding stuff. I’ve to say, after I first learn it, I used to be like, hey, wow, that is actually good. And after I learn the second time, I used to be like, you understand, that’s not fairly proper what it’s saying right here. It positively did sound good. I had the viewers hearken to that. They thought that was type of cute, it was enjoyable what it did. But then what I did subsequent, I assumed was actually fascinating. I requested it to redo it within the type of Taylor Swift, only for kicks, and it wrote this stunning poem with all these evocative metaphors—

ELISE HU: So it’s very Swiftian.

ERIK BRYNJOLFSSON: Yeah, completely. And I used to be like, it captures what I used to be pondering, however in a way more stunning method. And I believe in a method, really, the viewers received higher than they might have with my prose. So that was an actual aha second for me. Some of these metaphors that Taylor Swift got here up with, I used to be like, that’s so good. I’m wondering the place, you understand, the place GPT received this from. No one had ever stated this stuff earlier than, it was fully unique, but additionally fairly artistic and evocative. So that’s one thing that I imagined extra folks could be doing. I’ve been getting, I don’t find out about you and your folks, however I’ve been getting poems from my buddies and family members about me or one thing else, and a few of them are fairly foolish. But we’re having enjoyable with it, and I believe it’s altering our lives for the higher.

ELISE HU: Okay, and only one final thing. What type of future do you envision if we leverage the potential of AI?

ERIK BRYNJOLFSSON: I don’t assume any specific future is inevitable, and so I believe the best way you phrased the query was precisely proper. If we do it fallacious, we might have elevated focus of energy and wealth, lots of people shedding their financial wellbeing. But if we do it proper, I believe that it’s going to not solely result in shared prosperity but additionally result in a better price of invention, extra creativity, and folks inventing new medicine, new science, new sorts of buildings, new supplies that hadn’t existed earlier than with the assistance of those instruments. And I’d not be shocked in any respect if the subsequent decade was one of the crucial productive a long time ever in historical past, as a result of these instruments will enable us to do new issues that we by no means did earlier than.

ELISE HU: And but, someway, we’ll nonetheless wish to work.

ERIK BRYNJOLFSSON: Well, lots of people will. I believe the definition of labor will change somewhat bit. I believe we’re lucky—most likely my nice, nice grandparents wouldn’t acknowledge what I do daily as work, you understand, they’d say, I don’t get it, you’re not lifting something. My hope is that going ahead, extra of the routine, the boring, the rote elements of the roles can be completed by machines, the elements we don’t like will even be handed over to robots. And we’ll be capable to spend extra time on asking the appropriate questions. Also on interacting with different folks, that I ought to underscore, that’s one other factor that I believe people are uniquely good at and most of us get pleasure from, which is interacting with different folks, relationships. I believe most of us wouldn’t wish to have a robotic caring for our infants or our grandparents. We need to have the ability to work together with them ourselves. And that is also a uniquely human ability.

ELISE HU: Well, Erik Brynjolfsson, I recognize the dialog. I’m certain you’ll be able to inform how a lot I like speaking by means of these ideas and potentialities. Thank you a lot.

ERIK BRYNJOLFSSON: It’s been an actual pleasure. I like speaking about it with you.

ELISE HU: Thanks once more to Erik Brynjolfsson. I beloved that dialog. And that’s it for this episode of the WorkLab podcast from Microsoft. Please subscribe and examine again for the subsequent episode, the place I can be talking with Gloria Mark. She’s an writer and professor of informatics, exploring how leaders will help their groups regain management of their consideration and restore stability. If you’ve received a query you’d like us to pose to leaders, drop us an electronic mail at worklab@microsoft.com. And try the WorkLab digital publication, the place you can see transcripts of all our episodes, together with considerate tales that discover the methods we work at present. You can discover all of it at microsoft.com/worklab. As for this podcast, please price us, assessment, and comply with us wherever you hear. It helps us out. The WorkLab podcast is a spot for specialists to share their insights and opinions. As college students of the way forward for work, Microsoft values inputs from a various set of voices. That stated, the opinions and findings of our company are their very own and so they could not essentially replicate Microsoft’s personal analysis or positions. WorkLab is produced by Microsoft with Godfrey Dadich Partners and Reasonable Volume. I’m your host, Elise Hu, and my co-host is Tonya Mosley. Mary Melton is our correspondent. Sharon Kallander and Matthew Duncan produced this podcast. Jessica Voelker is the WorkLab editor.