Loading summary
Transcript65 lines
- [00:07]
A
Autonomur europa met nordea asset management.
- [00:14]
B
Ladies and gentlemen, our keynote speaker for tonight, Andrew McAfee.
- [00:19]
A
Julie, thank you so much. I want to talk about the question who's going to survive in this era of ridiculously fast technological progress?
- [00:28]
B
Jat AI all on the jobs wech Automatisehre days is stech denker Andrew McAfee van at Massachusetts Institute of Technology specialner brusolge vlog om obns deit event New Insights.
- [00:41]
A
Thank you for having me. I'm honored to be here tonight.
- [00:43]
B
Kommer prater overdemat schapplke impact van Artificial
- [00:47]
A
intelligentsi Science fiction is becoming business reality
- [00:51]
B
every day Cafe is MIT initiative on the digital economy the world Verndered and hevtar o kalverschellen buchen over his reven keynote of New Insights Condor noch efede podcast studio Interke Poronkesprek over the job Vernied the artificial intelligence. And watpe driver New best. Welcome by this extra flavoring Van der Seifer an exclusive interview with Decdenker Andrew McAfee. Mr. McAfee, welcome. We're only getting 20 minutes, so let me get straight to it. Will AI take all our jobs?
- [01:47]
A
No. One of the things that happens every time we get an extremely powerful technology coming across the economy, coming across our societies, coming across the business world, one of the things that happens very reliably is that that we have a panic about technological unemployment. And this was the case with the steam engine. This was the case with the internal combustion engine, with electrification. When we first got these things called computers, people immediately started worrying that they were gonna take all the jobs. The PC. We've seen this over and over again. Now, it is true that this modern bundle of AI that we have is astonishingly powerful and can do things that feel like science fiction. I very often feel I have to remind myself that I'm interacting a piece of technology when I work with AI these days. I get that this is very, very powerful. It's also new. I understand that as well. However, I think the burden is on people who say this time is different, to articulate why this time is different, to be very clear about that and then to have some evidence to back that up.
- [02:55]
B
Okay, fair enough. So, like in the Industrial Revolution, AI won't take all jobs, but some.
- [03:02]
A
It's going to probably take some jobs. Previous technologies have taken jobs. Let's be super clear about that. All the automation that we had early in the 20th century reduced the number of people who work on farms. We need fewer people to generate all of the manufacturing output in America than we did 30 or 40 years ago. And think about particular technologies like the shipping container that reduced the number of longshoremen. Technology changes jobs, and it's going to generate jobs as well. That's the other part to keep in mind, of course. But technology takes jobs, automates them. And what happens is innovators and entrepreneurs grab these new technologies and they realize that they can do novel things with them. And they start up companies, they start up entire industries. Those companies and industries need people to help with the work. They need labor. 50 years from now, are we going to have an extremely productive economy that doesn't need very much labor? Maybe. I don't know. But five, I'm pretty confident that we are not over that time horizon.
- [04:06]
B
No. But in the meantime, AI will take some jobs. Can we already see signs of what types of jobs are more vulnerable?
- [04:17]
A
Okay. We're early in the era of generative AI. ChatGPT only debuted about three and a half years ago. Three and a half years is enough time for us to start seeing some evidence of AI's impact on the labor force. And it's right where you would expect it to be. It's right in the jobs that are most exposed to AI, where a lot of the tasks can be done very well by AI. A great example here is computer programming. It used to be the case that all of our technologies were terrible at writing software. And if you wanted software written, you had to have people doing it. That's not true anymore. It's just not true anymore. I write a lot of software these days. I haven't written it by hand in 30 plus years, but I'm writing a fair amount of software these days. So that is different. And there is some, some research that shows that for jobs like software engineering that are pretty exposed to AI, labor demand might be slowing down compared to previous trends, Especially for the youngest entrants to the workforce.
- [05:17]
B
Okay, the youngest entrants. So starters will definitely be having a tougher time because of AI.
- [05:24]
A
Yes. Which makes sense because those are the people who you have to teach how to be a professional software engineer, a professional coder. They need kind of an apprenticeship. Apprenticeship period where they're learning how to do their jobs. And so if you have AI that can do all that work and you don't have to go through the apprenticeship period, maybe you can see why that's the part of the labor force that would be most exposed to AI and where we would see softening labor demand show up. So there is some of that evidence, there's a lot of other evidence that shows that AI is not yet causing massive joblessness anywhere. And one thing I always try to keep in mind is that when creative destruction happens, and we're in a period, we're early in a period of, I think, pretty deep creative destruction. When creative destruction happens, it is always easier to see the destruction and foresee the destruction, predict more destruction than it is to see the creation that goes on. It's just a very, very easy trap to fall into. I'm trying not to fall into that trap these days. Okay, focus.
- [06:33]
B
Andrew mcafee. This was by the industriale Revoluci of the StartFund computer diet. Bergsechti Dun was either in Okbung for overweldigen chopferlies Nardad Blake. Over the Angsteit Collectiv in her walle. If it'll be creative destruction rather than employment destruction, then why are people so scared that AI will specifically take their jobs?
- [07:27]
A
Because we prefer the status quo. We like things to be recognizable. We like them to be consistent with the past. And this is not consistent with the past. AI is new. It is novel. It is different. It is scary. In some ways, it appears to have intelligence. Artificial intelligence is not a bad label for what this is. And when we human beings face periods of deep, profound change, we very often don't like it. We want to cling to this past, cling to nostalgia, cling to the status quo. So I'm not surprised anymore at this large scale freaking out about new technology. But I take some comfort from history.
- [08:06]
B
I understand. And I'm thinking it's weird because the cliche is that people mostly get excited about new technology, right? So does AI specifically, does it have a PR problem?
- [08:20]
A
AI has a huge PR problem. The leaders of a lot of the companies and a lot of the very good pioneers of AI, who I've learned a ton from and who I respect a great deal, have been going around stoking fears about massive unemployment, about risks, about harms. Some people believe that AI either will become sentient and rise up against us, or we will make a mistake and it will become ungovernable and uncontrollable and unstoppable. These are science fiction stories. Literally. They're the plot of science fiction novels. And again, I fall back on things that I've heard that resonate with me. And I think it was the physicist Carl Sagan who reminded us that extraordinary claims require extraordinary evidence. And if you're making a claim that AI is gonna lead to Massive joblessness is gonna be an extinction level event for humanity. That is an extraordinary claim. That's a huge claim. The evidence is not on me to say. No, it won't. The evidence is actually on you. The evidence not just that. Well, here's my irrefutable logic. I need evidence about why this is a thing that we should be worried about. And so far the discussion about the existential risks and harms of AI, to me that discussion has been very, very light on evidence.
- [09:35]
B
That's fair enough. But still, I can't shake the feeling that it does go a bit further than that. Right. I think people might get scared because several companies like Amazon or Meta, or even Belgian companies have announced massive layoffs and have in doing so, overtly linked those layoffs to AI integration and AI driven efficiency. Meta dasupdrive, both on the Facebook and Instagram.
- [10:06]
A
Meta has laid off thousands of workers as the tech company is heavily focusing its effort around AI investment. Amazon just laid off 16,000 people daily.
- [10:17]
B
Combat Drive, Proximus, Scratch Proximus, MER, AI Hatchebraeke. So how do you square this? Because this makes the link between AI efficiency and layoffs, job losses feel very direct.
- [10:37]
A
Again, let's be clear. My claim is not that AI is going to automate nobody out of existence. I don't think that's accurate. It's also the case that these companies hired at an astonishing rate. Remember back in the early years of the pandemic 22, 23, the companies could not find enough people and they had hiring trends that were lined up with their revenue trends and that they expected those to continue. And in some cases they were overly ambitious in their hiring and they're cutting back. It's a better public relations mover. It makes you sound like a more forward looking leader if you say, wow, we didn't anticipate AI and I can now do all of this so we don't need people. That's better than saying we are hiring plans, we're over ambitious a couple of years ago. So I understand why people reach for AI as a reason for why they're changing their workforce composition. And I'll say it one more time, AI is automating some work. It is going to cause some level of job shuffling.
- [11:37]
B
But still above all else, AI is going to make us more productive.
- [11:42]
A
It's in the process of making us more productive. I think we also have some decent evidence that it's leading to increased rates of entrepreneurship because people are looking around and saying, wait a minute, I don't need to Hire a big team of people with all these skills that are very hard to find and very expensive. If I have some kind of vision for what I want to have happen in the world, I can just put that vision to work. There's a slogan going around Silicon Valley that I love, and the slogan is very simply, you can just do things. That is a slogan of empowerment that I love. And I think it's accurate. Every time I sit down, I've got some kind of vague ambition or a thing that they want to work on, and I start interacting with AI to say, hey, can we do this with this work? The technology goes, yeah. Hold on. Here's a first draft of that. This is amazing. And I find it weird that our dominant response so far to the appearance of a technology that is this clearly empowering for a lot of people is one of fear and one of concern and negativity. I think that's inappropriate. But I do want to say new technologies bring new risks, new harms. This is not new. AI is not an exception to that. New technologies have also advanced human flourishing recently. They have helped us lighten up on our planet. And AI is going to continue both of those trends, I believe.
- [13:08]
B
Andrew McAfee is the Stydulken optimist. Ocal heft AI. Well, undoubtedly pr problem. And as dalesdescht van de groote AI basen. So whilst mark zuckerberg van meter and jeff bezos van Amazon die nied aldete eve direct masa und slagen in hundbetreife linken an AI integrassie marikola over echte winston mcafee worked by mit naran in the global media.
- [13:50]
A
So there was a recent study that came out of MIT that's gotten a lot of attention
- [13:57]
B
onusoxinstitute mit.
- [13:58]
A
Wall Street's biggest fear might have just been validated by an MIT study. It seemed to suggest that 95% of investments in generative AI were not yielding any positive return. That statistic was shocking to some of my sources on Wall Street. Reality check.
- [14:33]
B
The reality check. Marit Wirt Al Snell Daedel nied bepalt vanwass van die Publicasie. So if AI is supposed to be promising in terms of productivity and profits, then how come a widely referenced study from your institution MIT from last year showed that 95% of companies that have heavily invested in AI?
- [14:53]
A
Yes. Stop. That was not research. That was a bunch of interviews conducted at one conference. That number gets tossed around. I beg you to stop citing that work. There's no.
- [15:03]
B
So it's incorrect then?
- [15:04]
A
It's pointless? No, it's absolutely nonsense. Well, it's not even research. It's just I interviewed a bunch of people and here's what I heard and we attached a 95% number to it. MIT does not release research. People at MIT release research. So I appreciate the temptation to call it an MIT study. MIT has not, in any sense of the word, said 95% of AI efforts fail. We need to rely on sounder work than that.
- [15:30]
B
Okay, so this publication has truly globally been misinterpreted.
- [15:35]
A
I mean, it was brilliant marketing. It's a big number like that. It grabs headlines. It's not worthy of discussion.
- [15:39]
B
Okay, fair enough. So you do believe that AI is generating extra productivity and profits for companies.
- [15:46]
A
As of today, there's very good research been published in some top journals that look at productivity gains, look at customer service improvements. There's a wonderful piece of work done by my colleague and co founder Eric Grynjolfsson and his team that found that in one customer service environment, AI had a bunch of benefits for the customers. The interesting thing, it reduced turnover among the customer service reps. And what we think, what I think is going on is that customer service is a difficult job in large part because customers get unhappy. When people got access to that technology, they stayed in their jobs longer. And that phenomenon was strongest for the youngest, the least experienced customer service reps. So we're getting very good, well done research that shows showing different levels of benefit. Again, this technology is very, very early.
- [16:39]
B
Okay, so very early. Too early to say if it's going to generate real profits, you know, out of the heavy AI investments that companies are putting in.
- [16:48]
A
It is the case that companies are spending a lot of money. They don't know what the ROI of those investments is. It reminds me a lot. We always talk about AI by analogy to something else. One of the analogies I like is an analogy between what's going on with AI right now and the early days of aviat. Because when the Wright brothers flew their plane, it was immediately clear to people all over the world that the world had changed. If you were in the transportation business, if you were moving people around, if you were engaged in national defense, the airplane was a world changing technology. So all over the world, people started doing exactly what they should do, which is flying airplanes. This thing is so new, it's so powerful, we don't understand all of its capabilities. Get planes in the air and let's figure this out. So we started a Lot of experimentation and a lot of investment. One of the very first things we learned, in addition to the fact that these were amazing technologies, is that they needed instrumentation. We were flying blind, literally. There are these wonderful books written from the early years of flight. Americans only know Antoine de Sainte Zupery as the guy that wrote the Little Prince. He was actually one of the pioneers of French aviation. And he writes about a trip he was taking over the northern coast of Africa at night. And he was literally flying blind. He had no idea there were no lights below. He had no radar. He didn't have any. He was just dead reckoning, trying to orient himself. And he thought he was about 1,000ft in the air. And he actually crashed into the ground and very nearly died that way because he was blind. And when I look at the uncertainty that companies experience these days about the returns on their AI investments and are we spending way too much and our token budgets are huge, it strikes me that we're in this same phase where we're flying blind too much and the experimentation is exactly appropriate. But then when you figure out what's working, do more of it. Get instrumentation, get better visibility over your portfolio of efforts. Understand if they're going up or going down or what the trends are, and then if try to fix things that aren't working well and do more of what is. That's what we're going to do in the years ahead. It's a very natural progression.
- [19:03]
B
In AI that he. In the primary Experimente Fase Citru Newark met AI Driver Mutodus, Experimentere Welke Tips Heved Mage Finuch vorbe drivesleiders, Deite omever in the Sune van de CEO Dran Stan. What kinds of companies will benefit most from AI?
- [19:50]
A
My very short answer is it's the same kinds of companies that have been successful at everything else so far in the 21st century. The 21st century has seen a real strong value concentration among companies that know how to work with fast changing technologies and are good at harnessing them.
- [20:07]
B
So basically the most flexible ones. So is not using AI still an option for companies?
- [20:14]
A
I guess so. I mean, you could still have all of your financial records in physical spreadsheets. A spreadsheet was a description of a big long piece of paper where you kept your books for the company and you called them a spreadsheet because you had to unfold them to look at the, you know, several years worth of data. You could still do that. I doubt you would be in business. I very much doubt that in almost Any part of the economy, I don't know how far in the future. But if you decide not to use AI, maybe if you're an artisanal butter maker, maybe if you're offering is a completely human artisanal experience, I guess there's room for that still. If you have ambitions to scale and to grow and to become a large enterprise, then no, no, you simply have
- [21:03]
B
to integrate AI into your company processes. Now what if I'm leading a company that hasn't yet implemented AI? What if I haven't really even discussed it with my employees in earnest, but I'm considering it? Where do I start?
- [21:21]
A
Let me give you three principles that I see the successful companies doing much more than the less successful companies. The first one is plan a little and iterate a lot. And what I mean by that is fine. You have to devote resources to a small set of efforts that are meaningful to you. That's the planning part. When a technology is this new and this fast changing and this unfamiliar, you cannot plan your way to success with it. You have to start doing stuff. You have to get the airplanes in the air. Even though that's risky in some ways and uncomfortable in some ways, you cannot plan your way to success here. The second principle is enable bottom up user led innovation. When you deploy this technology into an organization, okay, fine, your software engineers will love predict that you will also find that there are people in other parts of the organization that are geeks that are just going to grab this, adore it and do really deeply interesting things with it. We think about innovation very often as a centralized activity or a top down activity. And it can be that way. But one of the things I'm really excited about with AI is that it's going to enable bottom up innovation and energy. And not trying to tap into that I think is a pretty big mistake. And the third principle we've already talked about, stop flying blind. You need instrument of what you're doing.
- [22:37]
B
Okay, Those are some very clear guidelines. Now for some companies though, I imagine the instrumentation, this measuring the success rate of AI integration is easier, right? If you're a data company or a finance firm, you can clearly see to what extent your employees are using AI and what that is yielding for your company. But what if you're, let's say a hotel service or a news desk? Desk. Then it's much harder to measure, isn't it?
- [23:05]
A
It is. And it's easy to see if an assembly line in a factory is doing well or doing better. You know, look at your quality numbers look at your output numbers, look at your cycle time numbers. That's, that's pretty clear.
- [23:17]
B
Exactly.
- [23:18]
A
In a hotel, sometimes it's a little bit more vague. But you've got customer satisfaction measures, you've got financial measures. You can look at whether or not different, different advertising, marketing campaigns were effective. You have ways to judge the health of your business. What we say is great. You also can experiment with AI and it's not too difficult these days to say, okay, we intervened here with AI, did the things that we care about, did the trends start to get better or not? And if you want to get really sophisticated, you can just do an a B test, which people do all the time with their websites and other online tools. Let's give half the people AI and half not, and then we'll see if their performance diversify. There are ways to get at this question.
- [24:03]
B
Some companies, like Meta from Mark Zuckerberg, obligate employees to use AI. Could that conceivably be a good strategy?
- [24:14]
A
Yes. Yeah. You know what I mean? If I talk about OKRs.
- [24:18]
B
No.
- [24:18]
A
An OKR is an objective and a key result. And lots of very successful companies say we have a handful of things that we are going to concentrate on for the next year, for this year at the company. It could be a financial target, it could be a growth in customers or a growth in revenue or whatever, or it could be something like a safety initiative or a quality initiative. And OKRs are things that are expected of you in your job. A point that I make is make AI an okr.
- [24:48]
B
Right. Okay. So how should I, the manager of the company, go about this?
- [24:53]
A
There are more or less effective ways to do that. There are some pretty dumb ways to do it. So if you said to me, andy, I want you to double your AI usage or you're going to be in trouble, I'd say fine. And I would double the amount of time I spend prompting AI in meaningless
- [25:09]
B
ways by what we call token maxing.
- [25:11]
A
Yeah. And you would see that I hit my goals and I probably cost the company money and didn't really accomplish anything. So you can get the incentives wrong. But the message is, I think if this is a way you want to go, making an AI an OKR is a very effective thing to do. We are going to use AI at this company. If that doesn't work for you, best of luck. People at this company are going to use AI in the exact same way. I don't know if any company ever actually had to say this, but I'm sure some did. If you are going to be an accountant at this company. You are going to use the digital spreadsheet, not the paper spreadsheet. We're not doing that anymore. I don't think you're obligated to keep the accountant around who really loved his paper ledger and didn't want to change. Okay,
- [26:12]
B
Implementer. Would it be so bad if in the end, AI did take over all of our jobs and we could just lean back under a glorious system of universal basic income?
- [26:42]
A
One of my favorite science fiction series is a set of novels called the Culture Novels by Ian Banks, the Scottish writer. And they're exactly about the far future that you just identified. It's thousands, I don't know if it's tens of thousands, thousands of years in the future. And we humans are living in a post scarcity society, scarcity of any kind. And we are in partnership. That might not be the right way to say it. There are these pieces of non human intelligence called minds, which got started as AI and then they turned into something even more spectacular that pilot these gigantic ships and fake planets all over the galaxy. And the humans are kind of along for the ride.
- [27:24]
B
Okay.
- [27:24]
A
And I love those knowledge because they're very smart. They make you think about about a lot of things. But I find them to be more utopian than dystopian about a far future where we have extraordinarily powerful technologies. But here's the key thing in your question. We are post scarcity. We are living in complete material abundance. Nobody in the culture novels worries about their housing, about food, about the necessities of life, or about, you know, can I have a fulfilling life? It's all about how people manage to have fulfilling lives in a future where scarcity does exist. Now, I appreciate that. I'm sure the same is true for you and me and a lot of the people listening. Their work means a great deal to them. I get that. It's hard to imagine what my life would look like if I didn't have my work. But work in the narrow way that we think about it right now, a thing that you do that gives you money, man, maybe we need to broaden that definition out. I don't think that we need scarcity to have meaning in our lives.
- [28:26]
B
Andrew McAfee, thank you.
- [28:28]
A
Thank you.
- [28:32]
B
Vormer insecte van Andrew McAfee Leset verslag van new insights off had interview met MIT on Zukit van Menkollega Stefaniedesmet in the show notes business news,
- [29:38]
A
Asset management.