The impact of AI on the workforce, education and the economy

[Music] turns out that for many jobs if AI ultimates you know 20 30% of the task in a job then the job maybe is actually decently safe but what will happen is not that AI will replace people but I think people that use AI will replace other people that [Music] don't hello and welcome to Washington Post Live I'm Daniela bril a reporter here at the post who covers Tech how technology affects workers across Industries joining us today is Andrew a he is the founder of deep learning Ai and the co-founder of corsera an online learning platform and Andrew welcome to Washington Post Live thanks always great to be here before we get started I do want to tell our audience Andrew sits on Amazon's board of directors and Amazon's founder Jeff Bezos owns the Washington Post let's go ahead and dive in Andrew you've been a global leader in both Ai and education for decades when you first started Google brain in 201 were you thinking about the a impact AI might have on the workforce and the economy of the future I mean how would you describe the growth that you've seen in recent days um the growth has been tremendous I have to say those of us working on AI early days of Google brain even before I think many of us were very bullish about the impact that AI can have on the world and maybe it was a active naive day that you know 15 years ago I was thinking how can we have ai do many more things that humans can do today uh and it's been been been many years at work but it feels like the field is making good progress and so when you think about the impacts that AI is going to have in the workforce uh over the next five years I mean what would you see are sort of some of the biggest positive impacts I think we heard a little bit about that in in your intro here in that intro video um but if you were to really like you know look this out five 10 years what are sort of those biggest positive impacts and then what are some of your biggest concerns so I think be exciting massive upside with with a little bit of downside that needs to be acknowledged as well but already today every knowledge worker can get a meaningful productivity boost by using generative AI um part of the challenge is though most people will need a little bit of training to know how to use AI responsibly and safely but generative AI is impactful for I think pretty much all knowledge work today um uh but the training the the societal challenge of upskilling a lot of people to use these tools will be significant I know people are appropriately worried about AI taking away jobs um but it turns out that for most jobs AI can automate only a small fraction of their work and so training people to to use AI will be the main goal there is a very small number of jobs that maybe AI will fully automate and I don't want to minimize the suffering or the impact or our responsibility to make sure people are taken care of I think the number of jobs you know that will entirely go away is probably much smaller than most people think having said that it is really impactful on on individual and I think we owe it to them to provide the opportunities ups schill and keep on learning and keep on contributing to society as well I'm gonna continue with that a little bit um I just came from dreamforce and I asked this question to Mark Benny off himself um and and you're talking a little bit about it here and how these jobs you know maybe a small portion might be impacted but really it's portions of jobs rather than full jobs themselves I am concerned about um sort of entrylevel jobs and you know those first jobs out of college those call center jobs those lowlevel um jobs that don't require a whole lot of skill maybe they are repetitive in nature um but you know we all kind of start low level when we we get started in any industry or or we get our first jobs or we're switching careers and we're moving up the ladder it it does bring a question to people who are trying to get those entry-level jobs what kind of skill set you might have to have going in and and does that barrier become larger it raises the stakes and what you have to know to get a job in that industry I think the skill mix is changing and that is part of the challenge that training uh uh you universities as well as organizations like corer and D AI I think have a have an important role to play um and contact centers call centers is one of the job sectors most impacted but me to take another example that I know well software engineering ing it turns out that with generative AI it's easier than ever before for almost anyone to learn a little bit of coding and to use it effectively in their day-to-day work even if their job title is not software engineer so um it turns out on my team uh have people in kind of journalist or in marketing roles really not software engineering roles where I have a marketer that writes code a little bit of code to create web pages synthesize them to derive marketing Insight so he's a better marketer more effective marketer because he could use little coding have a reporter journalist work on my team that is writing code to um flag for him uh uh important articles that he should be paying attention to so what I'm seeing is the skill mix is changing and generative AI is creating more opportunities for a lot of people including people notthern software Engineers to learn a little bit of coding and it turns out with AI to help you code that easier than ever before uh but also uh to then use this you know in in these different job roles and you did mention call centers as being one of the sort of jobs that would be disrupted I'm curious is that the the one that you see kind of the target are there other roles you see that might be the first affected and and how that might change sort of what those jobs look like yeah so I think that um if we were to look at all the tasks involved in the job uh one thing that um there's a technique that one of my friends Eric ber offs in Stanford and colleagues at Pine which is a task based analysis of jobs and what that means is if you look at a job break the job down into tasks and then analyze which tasks are meable to AI augmentation or automation then you find that a lot of time you know like 20 30% of someone's job is um amable to you know AI automation or AI assistance and what that means is the nature of work will change uh because if someone Embraces these tools maybe with a little bit of coding or just buy your commercial off the-shelf solutions to use these tools then you know like a 20 30% boost in productivity is actually very significant but that work does need to be done to look at all the tasks in someone's job um and the you know take a look at which ones are amable to AI assistance and that would change the nature of some of these um these workflows and I know that uh I think J has been amazing you know I people use chity claw copilot Gemini uh uh llama other tools like that but there is a little bit of uh uh uh uh little bit of training and and thought needed uh uh uh to use that effectively so that kind of leads me into where I want to head next the uh company currently lead uh deep learning AI is an education technology company that provides AI training uh what's your advice to companies that are looking to upskill their employees and how should workers be thinking about what they need to do in terms of upscaling I know you talked about coding and I'm thinking of you know these genz that I talk to and you know they're graduating as software engineers and they're they're a little nervous you know what do I need to know do I know enough so that they hire me instead of having the AI do the lower level coding um so how should employers and also workers be thinking about that upskilling component I think in terms of workers I would say there are probably two segments to it uh one is technical rowes U where I feel like almost all software Engineers would be well served by learning about AI technology and learning how to build with AI um but then the broader set of jobs the largest set of economy is non-technical roles um I think last year the fastest growing course on corera was gen VI for everyone uh which I taught was offered by dear. a and I think that by providing training on what AI can and cannot do how to use it safely and responsibly what are the Privacy implications of using AI in different ways it turns out that um uh you know uh people could use AI as a brainstorming partner as a copy Editor to answer basic factual questions uh to summarize articles but a lot of things that we do that AI can assist with and in fact I think even reporters you know can can can AI help with copy editing or summarizing or no ticking so it turns out when you look at the job adapting the generative AI um to the task can can give us a big you know productivity boost and then employers have responsibility to provide the training um to help people uh understand what these tools can do and to adapt it to their toss so in terms of reskilling I mean it sounds like a lot of at least what I know of what it takes to sort of work with generative AI is pretty you know not a whole lot of complication in learning how to use it but I'm curious in terms of what you need to be good at your job to be working with AI in these in these ways whether it's coding or like you said um you know getting notes or or whatever you need to do Can employees res skill at the pace that AI is developing today I mean we're already seeing you know open AI talk about sort of like voice and talking um to to your AI and and it being able to talk to you I mean do we can we skill fast enough get the whole Workforce skilled up fast enough to sort of keep Pace with the how quickly the technology is developing yes we can but it won't be easy and we may or may not succeed um it turns out that I think the amount of training needed is not massive it's not that everyone has to go back to school to learn for years uh but I think a basic non-technical understanding of how does CHP or claw or gemini or LL actually work um and in some best practices on you know is it okay to type highly sensitive information is one of these things and what's the comfy policy on this and when you get a response back when should you trust it and when should you maybe double check his output before you go take action on it um I find that when people jump in and use the technology sometimes a few false thoughts where it say something hallucinates or make something work and you trust it ones that maybe we shouldn't have um but I find that the journey to start using it um it's very difficult to be good at gen unless you actually get hands- on with it I know that some sometimes people think using these tools is like an intellectual task you have to learn how to do it but I think there are more similarities to say riding a bicycle playing tennis then is broadly appreciated which is you know you can read all you want about riding a bicycle and that that could help but now you actually try doing it you really can't ride a bicycle unless you try or swing a tennis racket and I find that when people are jumping embrace the technology you know in a safe way make some mistakes see where it hallucinates or make facts up or uh I find that the learning curve let's people start getting value from this really quickly and then that continual usage hopefully drives continual learning um but I think that's what the individual usage there's actually one other thing I find very exciting which is um when the when the internet came up or when there was cloud computing when there were mobile devices it allowed people to write a lot of new software applications that were useful for all sorts of job roles right today the job of a reporter or marketer or or a recruiter is so different because of these you know Cloud host internet based tools that make this work more efficient so it turns out generative AI as a fundamental technology platform is also opening up tons of opportunities for businesses um or or for startup companies or for large companies to write new software applications that then people in different job RS can use into day-to-day lives and I think that um uh with the with the new tools that many teams are building uh for different job roles I think it be an exciting exciting future so given what you just said I feel like you kind of just answered my next question but I want to give you the chance to respond to it um in a pair of recent surveys from Boston Consulting Group about half of respondents and these are all SE Suite Executives said that they don't expect generative AI to bring about substantial productivity gains and that they're worried about the potential mistakes and data compromises arising from generative AI power tools so how do you respond to these concerns it sounds like you are pretty optimistic about sort of the possibilities it can bring to the table but these are some serious concerns yeah so I think I I say this with respect I think if there are um if someone does not think gen will bring massive privacity improvements um they're wrong uh and um it turns out that one of the economic you know calculations is over the past several years massive amounts of money has gone into training very large AI models we call them Foundation models so a lot of capital has been spent at you know companies like open AI anthropic you know Google and so on to train these very large AI models in fact soqu Capital had a really thoughtful article about you know filling like a $600 billion dollar right where where do you get massive amounts of revenue from to uh to to justify all the money you know going into companies like Nvidia for example but really other capex Investments and I think at the technology layer um I think could be really valuable but generating the returns at the the very short time scale that people are hoping for you know that's something that that needs a little bit of work but in terms of building applications on top of the technology um I'm seeing very good Roi and it's because some teams have spent so many billions of dollars building this AI technology that is then released on the internet you know sometimes free for anyone to use in the case of Open Source or open waste models and sometimes very inexpensively VI API calls in the case of a lot of the genifer AI companies it turns out that it's very Capital efficient now to write applications on top of this generative AI technology so for example my team AI fund uh we budget $55,000 to build a working prototype and we're building software prototypes to help with um process complex legal documents or to deal with complex government compliance requirements or to um uh you know uh help help review privacy documents uh but I find it or or to try to make Healthcare recommendations um but I find that with generative AI we can build really valuable prototypes that then can generate a significant Roi shortly after because it's so Capital efficient now to experiment and and to build applications and I think one important thing distinguish is the application layer versus the technology layer the technology layer is where a lot of capital has gone you know I think we have some work to do to generate the Returns on the time scale we're hoping but because of that investment technology is actually very efficient now to build applications and generate meaningful Roi on that so you know when you talk about the uses here um you know we're talking about in Industries where the stakes could be high right like something as Sim as simple as a summary might not be a big deal if there's a few errors or a hallucination in it if it's just summarizing a meeting but it might be problematic if it's um summarizing you know medical records or um something like legal documents are you at all worried about the current uh generative AI models that are out there given the fact that they still hallucinate yeah I think it's a great question GL young that um I want to say yes I'm worried but probably not as worried as you know um some others are so it is true that gen models make things up or hallucinate and I don't think the rate of hallucination or the rate of it ever making mistake will ever go to zero but it turns out that for a lot of applications the mistake rate is low enough that is just fine um because it turns out that uh humans make mistake as well um and I wish I could I wish I knew how to set up a contact center so that no one will ever say the wrong thing to a customer I wish that were possible and it turns out that the technology of um we call them God rails where we may have ai generating an output then have another AI double check the output before we send it to a user the mistake rate is never exactly zero it's not it's not that it's completely impossible for it to ever make any mistake but I find that um for a growing set of applications the mistake rate is low enough it's just fine I'll give you example I don't think we're there for healthcare yet I wouldn't want to ask AI what should I do about health condition and just blindly do that right which is not there yet but for lots of other applications uh low low stakes customer support or or we have other tricks like we call them confirmation flow where AI says you know do you really really want me to ship this for you and charge your $20 your credit card and then the user clicks yes right so things like that hope God againsts downside risk and so I think while Society people have broadly been appropriately worried about some of the downside risk if you look at the Practical practical day-to-day implementation is far from perfect but is also probably you know not as bad and more more things are signed to make into production than might be widely appreciated um well Andrew we only have a minute left but I want to see if I can squeeze this question in really quick if we can keep it brief um openai recently hired an executive from your company corsera to be its first general manager of Education uh the Hope here is to bring AI to more schools and uh and to the classroom are there any concerns you have about introducing this so early for children are still kind of learning the basic skills and how do we address like the situation with Equity here I think that um geni has the potential to be an extremely democratizing Force because today only the you know frankly relatively wealthy may be able to afford to hire personalized tutors highly skill personalized tutors for the kids so I think effs like um KH academies kigo and corera CER coach are excellent efforts uh I actually had the privilege of working of Leia bski is I think who you're referring to for many years uh and I I think that uh um I think that it's wonderful that open AI is trying to do more in education um I feel like so long as we do it responsibly we can build a brighter future and part of it is the reality of the future of work is people will be working with generative AI finding the appropriate way to teach school school children how to use it responsibly and well um to give everyone the you know personalized AI that helps them learn faster and take workload off the human teachers they can then do other things I think there's a lot to gain by thinking through responsibly how use gen responsibly in schools well we'll continue to follow this story Andrew thank you so much for joining us in Washington Post Live and for the wonderful insights you just provided thank you please stay with us for our next interview in the program when the world of business is constantly changing you need to stay ahead so what about AI to generate new possibilities machine learning to predict what customers will like or expand your business by forecasting delivery routes with real time data with the most experience in the cloud imagine how AWS can transform your business alternative AI is already revolutionizing the workforce by automating tasks and enhancing productivity joining me to discuss this very topic and how companies Can upskill employees to leverage AI is Amazon web services director of Enterprise strategy ISU vro johni thank you so much for joining us welcome well to be here with you Ruth fantastic well let's Jump Right In First can you give me real world examples of how artificial intelligence can increase productivity well in a very short period of time we have seen some remarkable impact of generative AI on employee productivity whether it is helping them find information very quickly for research and the task that they need to do from documents and data bases or helping them turn data into Insight by allowing them to interact with data using natural languages or automating repetitive task and I'll share with you two examples uh the first one is from LA County public defender office now that's a 109 year old office and one of the things that they did is they worked with us to build an AI based case management system that put their client the person at the center center of the system rather than the case and the calendar it reduced the manual entry by 85% and really alerted public defenders of things that they would have otherwise missed uh resulting in avoiding restr for one of the clients the second example is from Amazon uh if you think about software development one of the most tedious tasks but it's necessary is the foundational software upgrade uh so at the end of these upgrades you basically your customers get exactly the same thing that they had before but these are necessary upgrades you have to do them our developers used Amazon Q which is our generative AI powered assistant and we're able to upgrade thousands of java applications saving 4,500 years of development work and it's years not hours and that's $260 million in efficiency gain so we seeing these productivity gains making room for employees to then focus on things that truly differentiate and create value for customers absolutely well how then do you uplevel and educate employees so that they can reap the productivity benefits of artificial intelligence well it's super important to uh invest in upleveling and upskilling your existing employees because they know your domain they know the company the culture they have the context of your customers and I would say three things there first is clearly communicating the why to your employees how do you see the new skills that you are asking them to acquire but also investing so that they can acquire going to help them do their job better and how do how does that actually help them achieve the outcome the second thing is treating generative Ai and AI skill as a core competency for the whole company not just for Technical and Ai and machine learning roles so that everyone in the company whether they are in marketing or Finance or HR or customer support can benefit from them and the third thing is that you got to have some real word pilots and projects that people can very quickly try as soon as they acquire this new knowledge so that they are able to see the value of their newly acquired uh skill being used in the business looking internally how is Amazon tackling the cost barrier to upleveling and what benefits do you ultimately hope to see well look our uh Focus has been to democratize and make Ai and generi accessible for companies and institutions of all sizes and maturity uh and that also includes everyone from students to developers to Middle managers and Senior Executives uh and we are making substantial Investments uh We've announced an initiative called AI ready uh under which we have committed to train over two million people globally uh on AI skills and there are about 80 plus free courses that we have made available for everyone to use so I'll encourage you to check that out uh we also uh this is on top of 29 million people that we have committed to train on cloud computing skills by the way we have exceeded those Target by training over 31 million people and then the second thing that we do is a lot of customers want to know from us what are the best practices because we have been using Ai and machine learning throughout our business whether it is retail or our warehouses and fulfillment center Alexa or Amazon Prime video and we are always happy to share those lessons and best practices with customers as well absolutely well that brings me to my final question how does Amazon Advance the field of AI research well Amazon has been at uh Forefront of AI research for over two decades now our scientist and thought leaders uh advaned the field of research through publishing papers contributing to the community uh Amazon science publishes a number of technical papers that are available uh uh they also have been running a contest uh called Amazon trusted AI challenge uh which not only promotes the responsible de velopment of AI but also train students to acquire those skills we are also making substantial investment in research we've announced a $25 million 10-year commitment with University of Washington as well as University of sucuba in Japan and Nvidia over the next 10 years to fund Advanced research including post-doctoral scholarship and uh Fellowship as well as funding entrepreneurs who are doing some fascinating work in this area well in this area of Rapid transformation now is certainly the time to develop AI skills it sh thank you for such a robust conversation and now back to the Washington Post thank you Ruth [Music] I am pretty certain that some firms already are rethinking their workflows and they're rethinking uh their uh organizational processes in such a way that these Technologies create new task and New Opportunities uh and potentially also new jobs [Music] welcome back for those of you just joining us I'm Daniela bur Tech work writer for the Washington Post I'm now joined by Matt Bean a technology management assistant professor at UC Santa Barbara and rafhaela sedun business administration professor at Harvard Business School Matt and Rafaela welcome to Washington Post Live delighted to be here thank you for having us absolutely so Matt let's start with you uh you've been making the argument for years that AI could keep young workers from getting skills that they need can you explain a little bit more about that sure and um you don't have to be young to be a novice actually strictly speaking it's whether you're new and trying to learn how to do this new thing which in many cases means you're over 60 and and you're in your job but practically we're getting a lot of the gains I found this early in robotic surgery in say 2013 14 and have since found it in many other places we're getting a lot of the gains out of these new intelligent Technologies in part because they Empower a single expert to get that much more done at higher quality that much faster which the expert loves uh in general that's part of being an expert is solving problems the organization that is hiring them loves that too better Roi on that investment what it means though is that the novice becomes more and more optional sometimes entirely optional in the world work and when's the right time to involve someone who is slower and makes more mistakes the research I've done now across more than 30 sectors in the economy and occupations and so on says never basically and so what you see is a dramatic reduction uh really across all the places with very rare exceptions uh that I've looked in novice involvement in the work and there's a lot of talk about the value of training and training is of course important for building skill but if skill is your ability to reliably produce results under pressure on the job the research and the word on the street is you mostly get that on the job doing the job and so if you can't get it up to bat so to speak your prospects for skill and in fact your occupation's future your organization's future capability really are in a new kind of trouble so rafhaela I'm going to turn to you now I'm curious about your thoughts on Matt's take on the matter I know that you've made the argu M that there are responsible ways companies can incorporate AI as long as they're thinking about employee growth and communicating clear expectations but what do you think about this sort of loss in ability to get the novice involved here yeah so first of all I want to say Matt wrote a wonderful book about this where he really makes a grand a great argument for this lack of connection between the novice and the rest of and the more expert and I completely agree but I would say if I can build on what he just said said this I think reflects a misunderstanding of what training is uh we we think about training as something that you plug and play in the organization and maybe you buy a software that allows people to acquire new skills separately from what they do on the job but in fact training is part of a larger set of organizational practices that have to do with talent management and so I think that you know part of these practices for example are coaching and mentoring by middle managers and uh you know it's Direct related to the point that Matt was making the importance of human connections on the job uh across different layers of the Hy and so what I'm seeing is that there are certain organizations that have embedded training in this larger organizational culture and for example they are do a great job at motivating the middle managers as well as the employees in being part of part of it and others that I think a little bit more naively just do the stop- down approaches or approaches that don't really um engage uh people and you know the very human skills of connecting and mentoring and encouraging people in the process and I think that this is something to surface and something to work on um as long as we are aware of it Matt so some firms are already increasingly using AI in a variety of areas um we're hearing about cases in evaluation sales marketing particularly concerning recruiting um and I want to talk to you a little bit about how you describe the pros and cons here um I mean we've done actually a lot of coverage on the recruiting side and and how you know it could be concerning I'm curious about your thoughts on the benefits and and some of the cons in all of these um positions right well I think the main thing to say is we don't really know yet we need data and we need new metrics new ways of measuring the effect of deploying and predicting the effect of deploying these kinds of systems into organized life um we tend to focus on productivity we tend to focus on do we get the job done at higher quality or faster uh because we're used to measuring those kinds of things and uh if you buy the argument that I just levied and that thank very kind of you to mention my book rafhaela that that's in that book um then we have another metric to be concerned about which is Workforce Readiness like what is using this new tech in this new workflow new workflow going to do to you and your skills and your Readiness for tomorrow the if you repeatedly do that new job all day long we just don't really know in some cases that's going to be better for you it will upskill you your career will be brighter uh but again if uh the data I see are default uses of these systems our default modes of deploying them uh are going to do the opposite um it's going to create needless waste in terms of human potential uh and we just need better data is what I would say yeah I guess it's kind of a wait and see on that one um rafhaela I want to turn to you in the end when all this kind of plays out and we see more adoption of AI across different jobs different Industries different companies do you see AI as a net creator of jobs or do you see AI as a subtractor of jobs right look I mean I think that first of all at the individual level there is already massive adoption today uh or maybe yesterday a new paper came up about the US Workforce and David ding who is an economist at the canem school reported that 35% of American adults are already using Ai and uh at the individual level these are these are rates that are higher than PCS so I think that it's very very interesting to note at the individual level this is already happening at the consumer level now you contrast that with the statistics that we have about about the adoption of AI on in firms a sensus paper that came up a few uh months ago reported that the numbers were about 8% of firms and Incredibly heterogeneous between all the new firms small and large firms so you know to answer your question my sense is um first of all it's very difficult to give an answer that is average because the adoption of these new technologies as we as we've seen in previous technological waves will be incredibly different across different organizations organizations that will be able to understand how to experiment as Matt was saying and learn from these experiments and change tasks and change occupations I would expect that AI would be a net benefit not only in terms of productivity but also in terms of empowerment and term growth now on the other hand My worry is the lack of imagination if you like in some parts of the uh in some organizations where as a you know building on what I told you before a little bit naively we are just going to think about this is a technology that you plug and play and in that case I would expect you know perhaps more automation but also uh you know warningly not such a big boost in terms of productivity so my my my um uh main point here is it all depends and it all depends on what we do not only about the technology but what we do about the organization in which these technolog is embedded yeah and I think we're seeing a lot of that experimentation and people going all in and then realizing oh maybe I got to wind it back or implement it in a different way so that that makes a lot of sense Matt uh this is a question we asked Andrew in the last uh segment but I want to sort of see your opinion on the matter do you think it's possible to skill the workforce um quickly enough to keep up with the pace of Technologic development here I mean we are talking about a technology that you know in the past two years has completely just ballooned um and people are still some people have never even touched it um what are your thoughts on the matter failure is not an option and will occur unless we take aggressive action so is it possible sure but it's strictly necessary now like we we are inadvertently sacrificing skill development Workforce learning on the alar of productivity in many many cases it is needless and um so we must take great sort of directed action to try to solve that in my view and most of that as rafhaela said and as the research shows will not have to do with formal training it will have to do with the design of jobs and the design of tasks because if you buy the argument that most of your skills comes from just doing your job then Ground Zero for this effort is do you design work in a way that just by doing it you're better off in your skills that's if you want to think of an engine for training that could actually make a difference that's it it's the design of jobs it's the way we automate and that is I think very good news because the people who are making investments in their 20,000 co-pilot licenses for their Fir and thinking what do I do um they are the people who have their hands on the lever uh that can shape the trajectory of whether we sort of get the best of both worlds here or not um training is a necessary accelerant to that we can build awareness about tools sort of basic familiarity this is really obviously necessary that's what school is for um but it's not the main event and so thoughtful automation efforts uh involving experimentation and Gathering data on both fronts on the impact on the workforce as well as impact on productivity um it's all one effort now and AI actually in my view is also strictly necessary to make that effort possible so the the very thing that is now part of the our default use of it anyway is part of the issue is part of the solution we have to um it's too complicated a optimization problem in mathematical terms not to use AI to try to solve that very problem so uh that's why I'm a techno Optimist sort of in that sense but maybe a techno necessi or something like we there's no way to get where we need to go without the tool and yet the tool is currently the default use of it not the tool itself is part of the problem that's very complicated but I'm I'm following um R rafhaela according to research for Microsoft and Linkedin only 25% of companies are planning to offer training on generative AI tools like chat gbt and Microsoft co-pilot is that concerning to you or is the hesitation founded uh what's your advice to CEOs who want to incorporate tools like this in their businesses but they don't know where to start and they're very much worried about the effects of hallucinations misuse things like that yeah so look I and I come clean here I teach strategy in a business school so I'm going to say that the first thing that this should do is think about about how the incorporation of these Technologies eventually builds their competitive advantage and I feel that this is a point that is often missed in the attention that we pay to the uh to the application uh and the technology itself but I think that a very fundamental point is why are we doing this and how is this going to help us differentiate from other organizations I think this matters and U uh it's something that matters because that's how you explain to your organization why you want them to learn new tools because you know one of the issues that comes up in these training programs is incredibly low take up often and that is part of the part of the story is that we don't really explain to our workers and to our middle managers for that matter why there should be um upskilling the second point is I think it would be very important at this point to build a basic literacy across all levels also non-technical levels because we are at a point of extreme uncertainty where although the consumer side may be learning how to use these Technologies the application of this technology in the specific you know business model as is Sil un certain so you know this happens in the industrial revolution this happened in the ICT Revolution this is a point in which there is tremendous value in in experimentation if possible control experimentation where we can actually measure the returns and compare different approaches and for that to happen you need to give people the same l language and the if you like the familiarity or the lack of fear of engaging uh engaging with a new technology I'd say however that another Point here is that perhaps on the on the flip side even people who are familiar with the technology um should perhaps think about upskilling not so much their technical expertise but their soft skills because those people have the keys to interact with other parts of the organization that might not be as technical but they may have the right expertise to put the technology at use and so I think that this is you know part of part my advice to the SE would be to consider this as a strategic problem that involves different parts of the organization and involves also their uh communication skills Matt I want to move to a question from the audience Anna long from Virginia asks what recommendations do you have for a company AI policy uh so in a way actually Raphael and I have sort of been laying the groundwork um on this one which is focus intelligent experimentation that is also aggressive is probably necessary for most firms right now right you saw that stat on Executives feeling like they have to do something about AI um it's clearly the new general purpose technology that's going to reshape the economy uh in the way the internet n did and so on no matter the trajectory and so um that and yet it's interesting intelligent well-designed and sort of very visible experimentation we know a lot the science there is very good on how to do that in organizations and that is mostly not what we see right now mostly we see two modes one is don't do it in other words we're not buying that stuff we're not going to do it you're not allowed to use it and yet folks of course are on the slot on their cell phones um the alternative approach is just turn it on uh and it and just see what folks come up with which is sort of uh on its surface not harmful but with previous technological revolutions say the the electric Dynamo there's a marvelous paper on this by Davis research paper showing you know the first uses for this are remarkably unimaginative you know it's just kind of dro it in where you had the old thing and it's not going to hurt things per se but but you're not going to get that Quantum Leap Forward so go back to your MBA textbooks or whatever source of learning you've got about like how do you run a good experiment that's aggressive and that will be very public and show people the results failure or and success included in your organization in your community in your occupation um and and Bone up on that and run them involve people across multiple roles levels you raised a question of equity and education earlier with Andrew um there's an equity problem and opportunity inside organizations too if you really want to learn what's going on quickly it's usually your Frontline folks it's usually somebody somewhere you don't expect who has invented something that they themselves don't even think of as remarkable it's like yeah whatever I just did this thing but it turns out they made the critical discovery that can enable your strategy going forward so now is the time to learn from each other inside a company more than ever rafhaela I want to just give you the opportunity to add to that I know you were talking about transparency and and being really open with your employees about what you aim to do do you have any other suggestions um for a company policy yeah I mean I think too one is that the organization is uh you know it's an organization that is made of humans and often you know politics so I would say what I would add to uh what M just said is be very sensitive about the politics that that may be involved in this sort of organizational structure I want to be a change I want to be specific um the deployment of AI Technologies often relies on you know having good data and having centralized data now the process of building this data legs or you know sharing information across the firm is not of is not often the default mode for organizations I would say how do you think about convincing people uh to be part uh to be part of that um um of that effort and the second point is um I I think we make a big deal of um experiments as if we are you know for mat and I I I think we are on the same wavelength uh this maybe because we are researchers this seems you know how you should do it when I talk to business people I hear resistance because you know maybe this is adding uh too much time but in fact the reality is that companies run experiments all the time and they also have information uh would allow them to evaluate the outcomes of this of these experiments but because we are not thinking about them in a structured way or you know as an opportunity to learn a specific thing I think that we're actually wasting a lot of learnings that just get you know um they reside within one person rather than being shared across organization so I would say you know should maybe I don't know maybe we should create a SE position which is a chief experimental or something experimental officer I don't know another type of coo but it's really a different type of mindset I think these are some great suggestions and a great place to end it on uh unfortunately we are out of time so Matt Bean rafhael sadun thank you so much for joining us on Washington Post live thank you my pleasure thank you and thanks to all of you for joining us as well for more of these important conversations sign up for a Washington Post subscription you can get a free trial by visiting washingtonpost.com I'm Daniela bril thanks for being a part of the conversation

2024-09-27 12:52

Other news