Latest / Merryn Talks Money / What If AI’s Biggest Impact Isn’t Jobs, But Minds?

Transcript

What If AI’s Biggest Impact Isn’t Jobs, But Minds? · Merryn Talks Money

• 0:02
  Speaker 1: Bloomberg Audio Studios, Podcasts, Radio News.
• 0:19
  Speaker 2: Welcome to Meren Talk's Money, the podcast in which people
• 0:21
  Speaker 2: who know the markets explain the markets. I am Meren
• 0:24
  Speaker 2: Sumset Web, and this week I am speaking with Tom
• 0:27
  Speaker 2: Slater of Bailey Giffert. He is an investment manager inside
• 0:30
  Speaker 2: the private company's team.
• 0:31
  Speaker 3: He's head of the US equity team. But crucially, I.
• 0:34
  Speaker 2: Think most of you will know him as the manager
• 0:36
  Speaker 2: of the Scottish Mortgage Investment Trust, which I strongly suspect
• 0:39
  Speaker 2: most listeners hold, and by the way, I also hold.
• 0:42
  Speaker 2: I invited Tom on specifically because I wanted to talk
• 0:46
  Speaker 2: about an essay he's published recently called AI Isn't Coming
• 0:50
  Speaker 2: for your job, is Coming for your Mind. It's about
• 0:54
  Speaker 2: how AI is reshaping the way we think at a
• 0:57
  Speaker 2: fairly extraordinary speed, and what that means for us, for
• 1:01
  Speaker 2: workers in general, for companies, and crucially for society. It's
• 1:05
  Speaker 2: a really interesting We can put in a link, can't
• 1:07
  Speaker 2: we Tom at the end and people can read it
• 1:09
  Speaker 2: for themselves as well.
• 1:10
  Speaker 4: Oh absolutely, yep.
• 1:11
  Speaker 3: Okay, great.
• 1:12
  Speaker 2: Anyway, that wasn't supposed to say that I'm supposed to
• 1:13
  Speaker 2: say before I start talking to I'm supposed to say, Tom,
• 1:16
  Speaker 2: welcome to Marin Talks Money.
• 1:18
  Speaker 4: Thank you very much for having me. It's great to
• 1:20
  Speaker 4: be here.
• 1:21
  Speaker 2: I'm really interested in this. I can't tell you how
• 1:23
  Speaker 2: much it resonated with me. Normally, when we're talking about AI,
• 1:26
  Speaker 2: we talk about the investment, we talk about is there
• 1:28
  Speaker 2: a data center bubble, We talk about how it's going
• 1:31
  Speaker 2: to take everyone's jobs, and we talk about how it
• 1:33
  Speaker 2: will reshape the economy. And while there are occasional conversations
• 1:37
  Speaker 2: about how it might actually reshape our brains and even
• 1:40
  Speaker 2: our consciousness, that is not people's focus.
• 1:43
  Speaker 3: And actually, the truth is probably the most important thing
• 1:46
  Speaker 3: of all.
• 1:47
  Speaker 2: It's an awful lot more important than the valuations of
• 1:51
  Speaker 2: AI companies and AI adjacent companies. Right, what's happening in
• 1:54
  Speaker 2: your head? So let's talk about how you started thinking
• 1:58
  Speaker 2: about that.
• 2:00
  Speaker 4: I guess it's a topic that is at the top
• 2:03
  Speaker 4: of people's minds at the moment in one way or another.
• 2:05
  Speaker 4: We're all being affected by this technology, thinking about what
• 2:08
  Speaker 4: is doing to society, and I guess from books that
• 2:13
  Speaker 4: I'd read, I just started wondering about whether we were
• 2:17
  Speaker 4: focusing on the wrong question, because this is very obvious,
• 2:20
  Speaker 4: you know, is our job is all going to be
• 2:23
  Speaker 4: automated away by this technology, and we can get into that.
• 2:27
  Speaker 4: But the more I thought about it, the more it
• 2:28
  Speaker 4: seemed to me, actually that wasn't the issue it because
• 2:32
  Speaker 4: it assumes sort of stasis. It assumed that humans would
• 2:36
  Speaker 4: stay the same and this technology would get better and
• 2:39
  Speaker 4: would be displaced. But actually all the evidence suggests that
• 2:42
  Speaker 4: it changes the way we think, So there isn't some
• 2:46
  Speaker 4: sort of stable balance between and us and the machines
• 2:49
  Speaker 4: we adapt. So simple example for me, you know, I
• 2:53
  Speaker 4: lived in Edinburgh for thirty years and it used to
• 2:56
  Speaker 4: be the case that I could think, I want to
• 2:57
  Speaker 4: go to this place and visualize whereas going in my head,
• 3:00
  Speaker 4: and I'd set off. But just because I turn the
• 3:04
  Speaker 4: sat nave on every time I go somewhere in the
• 3:06
  Speaker 4: car now and it's got to the point where I
• 3:09
  Speaker 4: actually can't do that anymore. And I haven't suddenly lost
• 3:12
  Speaker 4: that skill, but it's neglect. And each time I've not
• 3:15
  Speaker 4: made that effort, that that piece of my brain has
• 3:18
  Speaker 4: become less utilized as shrunk, and now I can't do it,
• 3:22
  Speaker 4: and I think my decusively, I does the same thing.
• 3:25
  Speaker 2: Tom in the beginning of your paper, you talk about
• 3:28
  Speaker 2: how our brains literally rewired when we learned to read.
• 3:33
  Speaker 2: So you say, hundred years ago, so only abound twelve
• 3:35
  Speaker 2: percent of the world's adults could read. Today it's about
• 3:38
  Speaker 2: eighty seven percent. And we all look at that and
• 3:41
  Speaker 2: we go, well, this is it's all about education, and
• 3:43
  Speaker 2: that's amazing because it leads to different livesly, democracy, et cetera,
• 3:47
  Speaker 2: et cetera, all these things. But it wasn't really just
• 3:50
  Speaker 2: about that. It was about our brains rewiring. And I'm
• 3:53
  Speaker 2: really interested in this bit because I actually gotten some
• 3:55
  Speaker 2: rouse on social media about this recently, about whether if
• 3:59
  Speaker 2: you read very fast, which I do, that affects your
• 4:06
  Speaker 2: ability to recognize faces, because it's the same part of
• 4:10
  Speaker 2: the brain. And what you say here is that is
• 4:13
  Speaker 2: that the connection between the hemispheres thickened. I'm not quite
• 4:18
  Speaker 2: sure what that means. I'm going to leave you to
• 4:19
  Speaker 2: explain that. But this intellectual shift led to a biological shift.
• 4:23
  Speaker 4: Yeah, that's right. And I find this fascinating because I
• 4:27
  Speaker 4: think it. Certainly I had thought about changing physiology as
• 4:30
  Speaker 4: a being a function of evolution. It happens across generations.
• 4:34
  Speaker 4: But it was reading a book called The Weirdest People
• 4:36
  Speaker 4: in the world. That got me think about this, and.
• 4:39
  Speaker 3: We had the author of that on the podcast a
• 4:41
  Speaker 3: while back.
• 4:42
  Speaker 4: It's a fascinating book, such a good book, but he
• 4:44
  Speaker 4: talks about this way that sort of culture hijacks your
• 4:48
  Speaker 4: physiology and changes changes the way you think, and in
• 4:52
  Speaker 4: particular this learning the skill of reading and writing, it
• 4:56
  Speaker 4: caused these changes in the brain. We repurpose this area
• 5:00
  Speaker 4: of the brain that was used for recognizing faces. There's
• 5:04
  Speaker 4: a bit of a trade off there. You do become
• 5:06
  Speaker 4: less good at that, but in return you get this
• 5:09
  Speaker 4: incredible skill which is so useful. It's just a really
• 5:11
  Speaker 4: stark example of the way that you use technology. In
• 5:15
  Speaker 4: that case, writing reshapes your physiology, and so we shouldn't
• 5:20
  Speaker 4: It is absolutely possible for a cultural technology to reach
• 5:23
  Speaker 4: inside your skull and change the organ that makes you human.
• 5:28
  Speaker 2: Interesting because we tend to think that, as you say,
• 5:30
  Speaker 2: that's evolutionary and it should take thousands and thousands of years.
• 5:32
  Speaker 2: There should be generations of tiny little bits of change
• 5:35
  Speaker 2: until the result appears that this can happen in no
• 5:38
  Speaker 2: time at all.
• 5:39
  Speaker 3: And reading spread quite slowly, right, Not everybody.
• 5:42
  Speaker 2: Learned to read in a year, but everyone who can
• 5:45
  Speaker 2: read now has access to AI immediately, so it's much faster, right,
• 5:50
  Speaker 2: And what is it doing?
• 5:51
  Speaker 4: I think that's right. So I think what you can
• 5:53
  Speaker 4: what you can see is that you get this sort
• 5:56
  Speaker 4: of These cultural changes operate in the same way that
• 6:00
  Speaker 4: genetic changes work. The three factors of variation, transmission, and selection.
• 6:06
  Speaker 4: If you think about you go back to the weirdest
• 6:09
  Speaker 4: people in the world. The example there is the Catholic
• 6:11
  Speaker 4: Church and how that controlled variation and ideas. It controlled transmission,
• 6:16
  Speaker 4: the way ideas spread and controlled selection. Now, if you
• 6:21
  Speaker 4: apply that to what's going on in the world today, AI,
• 6:26
  Speaker 4: we can supercharge variation because the number of ideas that
• 6:30
  Speaker 4: you can test, that you can generate grows exponentially, and
• 6:34
  Speaker 4: ideas that humans just wouldn't have got to. It completely
• 6:37
  Speaker 4: changes transmission of ideas. If a million children ask their
• 6:41
  Speaker 4: parents why is the sky blue, they'll get close to
• 6:44
  Speaker 4: a million different answers. But because we're all started using
• 6:47
  Speaker 4: two or three of these really dominant systems, everybody's getting
• 6:51
  Speaker 4: some variation of the same answer. And yes, those systems
• 6:54
  Speaker 4: are built on the whole corpus of human knowledge, but
• 6:57
  Speaker 4: the centralization is incredible. You've changed the transmission mechanism for ideas.
• 7:03
  Speaker 4: And then these AI algorithms that power media sites now
• 7:07
  Speaker 4: completely controlled selection, which ideas get amplified, which quietly disappear.
• 7:13
  Speaker 4: So we've allowed technology to dominate those three key vectors
• 7:17
  Speaker 4: of cultural evolution, variation, transmission, selection.
• 7:21
  Speaker 2: Yeah, and you say as well that you now understand
• 7:23
  Speaker 2: why Elon Musk bought Twitter.
• 7:26
  Speaker 4: Yeah, it was quite at the time. It was challenging
• 7:29
  Speaker 4: to understand the economic rationale. Now that we can debate
• 7:32
  Speaker 4: whether that has changed subsequently. But when you think of
• 7:35
  Speaker 4: it in terms of the impact that it has on
• 7:38
  Speaker 4: cultural evolution and shaping ideas, you can see why that
• 7:41
  Speaker 4: would be a valuable asset to somebody.
• 7:44
  Speaker 3: Interesting.
• 7:44
  Speaker 2: Okay, so it sounds when you put it like that,
• 7:47
  Speaker 2: it sounds as though civilization itself will be transformed by
• 7:52
  Speaker 2: these three things.
• 7:52
  Speaker 4: I think that it does bring change. And the question
• 7:56
  Speaker 4: to my mind is are we going to passively and
• 8:00
  Speaker 4: unthinkingly let this change happen or do you have a
• 8:04
  Speaker 4: debate about how these tools are used where they used.
• 8:08
  Speaker 4: Do we passively accept our fate with this or do
• 8:10
  Speaker 4: we proactively seek to control it?
• 8:13
  Speaker 2: Okay, well, let's look at what happens when we allow
• 8:16
  Speaker 2: it to happen passively.
• 8:18
  Speaker 3: Right.
• 8:19
  Speaker 2: So one of the things you look at is this
• 8:21
  Speaker 2: MIT study that looked at brains of participants.
• 8:25
  Speaker 3: While they wrote essays.
• 8:27
  Speaker 2: So some of them wrote essays with absolutely no assistance
• 8:30
  Speaker 2: at all. Some were able to use search engines, and
• 8:33
  Speaker 2: some were able to use AI assistance.
• 8:35
  Speaker 3: And that was an.
• 8:36
  Speaker 2: Experiment that showed us what will happen if we don't
• 8:38
  Speaker 2: bring some focus into this conversation.
• 8:41
  Speaker 4: Yeah, absolutely, So what you saw was that if you
• 8:44
  Speaker 4: looked at what happened the two people, if they were
• 8:48
  Speaker 4: given AI, they would essentially outsource the writing process. And
• 8:53
  Speaker 4: the result was that about over eighty percent of that
• 8:56
  Speaker 4: group that used AI to write the essay couldn't provide
• 8:59
  Speaker 4: a single correct quote from the essay they had just
• 9:02
  Speaker 4: written a few minutes earlier. So the effort that goes
• 9:06
  Speaker 4: into the durable learning that writing an essay had just
• 9:09
  Speaker 4: been bypassed entirely, And that was completely different in the
• 9:14
  Speaker 4: group that you search the group that wrote themselves. If
• 9:17
  Speaker 4: you just employ these tools to do something you would
• 9:20
  Speaker 4: have done otherwise done yourself, you will gradually lose the
• 9:23
  Speaker 4: capability to do that. You will never develop the skill
• 9:25
  Speaker 4: to do that. But if you engage positively with these systems,
• 9:29
  Speaker 4: you don't you have the essay to write. You don't
• 9:31
  Speaker 4: just say to the AA write the essay, but you
• 9:33
  Speaker 4: go and research the question using AI as a tool.
• 9:37
  Speaker 4: I want to understand why this happened. I don't understand
• 9:39
  Speaker 4: the relationship between factory and fact B. You actually interact
• 9:43
  Speaker 4: with it and think about the output and engage with it,
• 9:46
  Speaker 4: then you get the same results cognitively as if you
• 9:49
  Speaker 4: if you just hand written the essay. So it's that
• 9:51
  Speaker 4: there's something about that effort, full interaction, that seeking to
• 9:56
  Speaker 4: gain knowledge that is the crucial part in actually developing
• 10:00
  Speaker 4: the cognitive skills. And if you passively let the IAI
• 10:04
  Speaker 4: do it, then it's not simply as bad as you
• 10:08
  Speaker 4: never learn it. But actually you will start to mistake
• 10:13
  Speaker 4: your ability to use AI tools with your mastery of
• 10:17
  Speaker 4: the subject that you are trying, and so you will
• 10:19
  Speaker 4: gain no knowledge or little knowledge and skill in the
• 10:23
  Speaker 4: subject that you're focusing on. You'll just get better at
• 10:26
  Speaker 4: using the AI, but you will confuse that with thinking
• 10:29
  Speaker 4: that you're good at the actual topic.
• 10:32
  Speaker 2: Okay, so, as you would put it, you've made a trade,
• 10:34
  Speaker 2: but you haven't recognized that you've made that trade, yes,
• 10:38
  Speaker 2: traded efficiency, saving time, etc. And the result has been
• 10:43
  Speaker 2: that your brain has become weaker, actively weaker.
• 10:46
  Speaker 3: It's worse than it was.
• 10:48
  Speaker 4: Yeah, that's right. It's because you. I think it's called
• 10:51
  Speaker 4: Dunning Kruger effect, this sort of this as you put
• 10:55
  Speaker 4: more effort into something, but if you are you don't
• 10:58
  Speaker 4: know what you don't know effect me And as you
• 11:00
  Speaker 4: put more and more effort into it, you become more
• 11:03
  Speaker 4: aware of how little you know of the topic. But
• 11:06
  Speaker 4: you completely break that effect, that understanding of the limitations
• 11:09
  Speaker 4: of your own knowledge when you start using these tools,
• 11:12
  Speaker 4: and that's the real weakness, the real interruption of the
• 11:14
  Speaker 4: learning process, and you.
• 11:16
  Speaker 2: End up without that base knowledge for yourself. And so
• 11:19
  Speaker 2: that brings us to the world of work, or and
• 11:23
  Speaker 2: in fact the entire world. Right, there's been so much
• 11:25
  Speaker 2: talk about AI removing the bottom level of jobs so
• 11:29
  Speaker 2: I can do the simple stuff. It can do anything
• 11:31
  Speaker 2: an intern could do. It could do anything a junior
• 11:33
  Speaker 2: could do. You can argue about whether it can and
• 11:35
  Speaker 2: can't an extent, which it makes mistakes, since we've done
• 11:37
  Speaker 2: podcasts on that, of course, but for the moment, let's
• 11:39
  Speaker 2: just take it as read that it is possible the
• 11:41
  Speaker 2: AI to remove these lower levels of jobs. That then
• 11:47
  Speaker 2: turns into a real problem because if the next level up,
• 11:51
  Speaker 2: if the job market starts at the next level up,
• 11:54
  Speaker 2: that base knowledge is never embedded and created. You have
• 11:59
  Speaker 2: to judge whether what your agent has told you is
• 12:01
  Speaker 2: correct or not. But you no longer have the foundational
• 12:04
  Speaker 2: skills to be able to do that.
• 12:06
  Speaker 4: Yeah, absolutely so. I think one way of thinking about
• 12:09
  Speaker 4: this technology is that it's not it's tied that lifts
• 12:13
  Speaker 4: all boats. In some ways, you can think of it
• 12:15
  Speaker 4: as a force multiplier, and what I mean by that
• 12:18
  Speaker 4: is if you are experienced in a topic and have
• 12:22
  Speaker 4: a great deal of knowledge, then these tools can massively
• 12:24
  Speaker 4: enhance your productivity. But a key part of it, as
• 12:27
  Speaker 4: you highlight, is you have to be able to understand
• 12:31
  Speaker 4: interpret the output of these systems. And if you're a
• 12:35
  Speaker 4: scientist who's never struggled through a statistical analysis manually, then
• 12:39
  Speaker 4: you're probably not going to spot the results that are
• 12:43
  Speaker 4: conceptually meaningless that are being spat out by a machine.
• 12:46
  Speaker 4: It's that prior mastery which allows you to genuinely evaluate
• 12:51
  Speaker 4: what's coming out of these systems rather than just giving
• 12:53
  Speaker 4: it a superficial check. And so for those that have
• 12:56
  Speaker 4: not gained the skills prior to using these tools, you
• 13:00
  Speaker 4: don't have the ability to do that you have. There's
• 13:03
  Speaker 4: a completely different power balance between the AI system and
• 13:06
  Speaker 4: the human.
• 13:31
  Speaker 2: So a quote from your paper, the central paradox is
• 13:34
  Speaker 2: that AI reliably improves immediate task performance while degrading the
• 13:39
  Speaker 2: underlying human capabilities that produce that performance. You get better
• 13:43
  Speaker 2: results today, but become less capable tomorrow. And you talk
• 13:46
  Speaker 2: about that as well in terms of productivity, and that
• 13:49
  Speaker 2: we get a productivity revolution upfront while undermining the foundations
• 13:55
  Speaker 2: of that productivity enhancement. So it can't continue indefinitely unless
• 13:59
  Speaker 2: it's used in different way.
• 14:01
  Speaker 4: Yeah, that's it. So you get this one time benefit
• 14:04
  Speaker 4: if you like that, you have a workforce who's been
• 14:07
  Speaker 4: through the effort for struggle to gain capability and knowledge
• 14:11
  Speaker 4: and mastery and then can use these tools. But unless
• 14:16
  Speaker 4: you invest in the next generation and ensure that they
• 14:20
  Speaker 4: have those skills, they don't simply use these tools, or
• 14:23
  Speaker 4: they don't simply oversee these tools, then they can never
• 14:26
  Speaker 4: have that same relationship with them, and you don't get
• 14:29
  Speaker 4: the same productivity and benefit.
• 14:31
  Speaker 2: And then even worse you get this confident trap that
• 14:37
  Speaker 2: you talk about as well. I was really struck for
• 14:39
  Speaker 2: which read absolutely horrified by a study you talk about
• 14:43
  Speaker 2: from last year in the Lancet which tracked endoscopists.
• 14:48
  Speaker 3: How do you say that word? Indoscobus words? I can't
• 14:52
  Speaker 3: even say.
• 14:52
  Speaker 2: Look here we are. I'm destroyed by AI already. And anyway,
• 14:56
  Speaker 2: they went through periods of working with AI when they
• 14:59
  Speaker 2: hadn't been much came with AI, and the results for appalling.
• 15:02
  Speaker 4: Yeah, that's right. So you saw if you had this
• 15:05
  Speaker 4: experienced practitioners using AI systems, and then after you took
• 15:09
  Speaker 4: the AI systems away, there was one key indicator, for example,
• 15:14
  Speaker 4: where the detection rate fell twenty one percent with the
• 15:17
  Speaker 4: same group before and after they'd had access to these
• 15:20
  Speaker 4: AI tools. But then there was this, and I think
• 15:24
  Speaker 4: this was probably the more worrying result, was that people
• 15:27
  Speaker 4: tend to trust the output from the AI, and so
• 15:31
  Speaker 4: when the experimenters went in and tweaked the AI to
• 15:35
  Speaker 4: give confident but incorrect interpretations, what you saw was that
• 15:40
  Speaker 4: human performance collapsed. So humans were much more likely to
• 15:44
  Speaker 4: accept that the answer that they were being given was right,
• 15:48
  Speaker 4: even though it was not.
• 15:50
  Speaker 3: Even when they were experienced and should know.
• 15:52
  Speaker 4: Even when they were experienced. Yes, and AI literacy was
• 15:56
  Speaker 4: not sufficient to protect against that, so that the tool
• 16:00
  Speaker 4: is fallible, did not prevent over alliance.
• 16:05
  Speaker 2: We've seen this happen. These studies in the medical industry
• 16:08
  Speaker 2: are really intrue. But we've seen this happen in real
• 16:09
  Speaker 2: life in the legal.
• 16:10
  Speaker 3: Industry, haven't we.
• 16:11
  Speaker 2: We've seen lawyers actually on several occasions. In fact, we've
• 16:14
  Speaker 2: had examples of this in Scotland, le seen lawyers actually
• 16:17
  Speaker 2: relying on AI and producing false stuff, false quotes, false cases,
• 16:22
  Speaker 2: false presidents.
• 16:23
  Speaker 4: Yeah, and it's back to this. If you get good
• 16:27
  Speaker 4: results from the system, then you fall into this pattern
• 16:31
  Speaker 4: of not questioning the output. And that's true of experienced
• 16:36
  Speaker 4: professionals and its and it's problem that only gets worse
• 16:39
  Speaker 4: if people don't get that sort of underlying experience and
• 16:42
  Speaker 4: understanding before they start using the tools.
• 16:45
  Speaker 2: So we could end up in a position where most
• 16:48
  Speaker 2: people do not have the foundational knowledge required to spot
• 16:53
  Speaker 2: errors or to really see what is right and what
• 16:55
  Speaker 2: is wrong. And even people with solid foundational knowledge can
• 17:00
  Speaker 2: still fall into these traps where they don't use their
• 17:03
  Speaker 2: foundational knowledge to spot mistakes.
• 17:06
  Speaker 4: YEP, and that's and this is this back to this
• 17:09
  Speaker 4: sort of erosion of human skill or changing the way
• 17:12
  Speaker 4: that there just isn't this stable balance between us and
• 17:15
  Speaker 4: the machines. Is use these tools. As they become more
• 17:17
  Speaker 4: and more effective, your relationship with them changes, and there
• 17:22
  Speaker 4: is we have this tendency to to but when they
• 17:25
  Speaker 4: confidently say things about the world to accept it as true,
• 17:28
  Speaker 4: we give them a false inappropriate amount of respect.
• 17:33
  Speaker 3: So what do we do?
• 17:34
  Speaker 4: I think there's a couple of things. There are. There
• 17:37
  Speaker 4: is some evidence that in sort of expert domains that
• 17:43
  Speaker 4: if you use these tools deliberately, they can make you better.
• 17:46
  Speaker 4: One one example of that was in the game Go,
• 17:51
  Speaker 4: where AI famously beat a human player. But what you've
• 17:54
  Speaker 4: seen subsequent to that that sort of landmark event was
• 17:59
  Speaker 4: that human players have learned from the AI. It's developed
• 18:05
  Speaker 4: novel strategies, novel ways of going about the game, and
• 18:07
  Speaker 4: you've seen an improvement in human performance. And do you
• 18:11
  Speaker 4: see that result in sort of expert fields where these
• 18:15
  Speaker 4: tools are used in a very deliberate way to be
• 18:19
  Speaker 4: complementary to the human behavior. And so it's I think
• 18:24
  Speaker 4: the key to it is do you do you actually
• 18:28
  Speaker 4: deliberately and thoughtfully engage with these systems or do you
• 18:33
  Speaker 4: do you just allow them to take over tasks that
• 18:35
  Speaker 4: people are doing, And all the evens suggests that if
• 18:38
  Speaker 4: you don't really tightly control that, people will just let
• 18:41
  Speaker 4: the machine do it and human capability will atrophy.
• 18:45
  Speaker 2: Yeah, and the worry is that we are seeing that
• 18:48
  Speaker 2: in a big conversation of course at the moment is
• 18:50
  Speaker 2: around employment and about the disparance of junior jobs, which
• 18:54
  Speaker 2: we talked about just now. But that's not theoretical that
• 18:56
  Speaker 2: that's happening. And you read about that in this paper
• 18:59
  Speaker 2: as well, about hollowing out of white collar professions. So
• 19:03
  Speaker 2: it looks that we are already taking the wrong path here,
• 19:07
  Speaker 2: and that we should have just as many junior lawyers, accountants, doctors,
• 19:13
  Speaker 2: customer services.
• 19:14
  Speaker 3: People, whatever it is.
• 19:15
  Speaker 2: We should have as many as we had before for
• 19:17
  Speaker 2: the knowledge building.
• 19:18
  Speaker 4: I think that's the real challenge, right because people look
• 19:20
  Speaker 4: at this technology and say, it can allow me to
• 19:22
  Speaker 4: reduce costs and I don't have to hire junior people,
• 19:27
  Speaker 4: I don't have to train them, I can use these systems,
• 19:30
  Speaker 4: and there's real pressure to do that. But I just
• 19:32
  Speaker 4: think that is a very dangerous path to go down.
• 19:36
  Speaker 4: I think there's real value in having humans learning to
• 19:40
  Speaker 4: do these tasks which have white collar profession, and having
• 19:45
  Speaker 4: them do it without access to these systems, and so
• 19:50
  Speaker 4: that you then retain that ability to oversee override these
• 19:56
  Speaker 4: systems when they're not producing the right answers. But you
• 19:59
  Speaker 4: have to invest in the human capital first. So I
• 20:02
  Speaker 4: guess the analogy would be with pilots. Pilots still have
• 20:06
  Speaker 4: to learn to fly manually before they can learn to
• 20:09
  Speaker 4: use autopilot because if the autopilot at some point fails,
• 20:12
  Speaker 4: they need to be there to land the plane.
• 20:16
  Speaker 2: This is why everyone has a bias towards older pilots, right.
• 20:19
  Speaker 2: I don't know about you, but when I got on
• 20:20
  Speaker 2: an airplane and I see that the pilots like maybe
• 20:22
  Speaker 2: over forty five or that is still over fifty, I'm thrilled,
• 20:26
  Speaker 2: absolutely thrilled, because I know that it means that they've
• 20:28
  Speaker 2: got a strong base of knowledge that are to actually
• 20:31
  Speaker 2: fly a plane because they probably had to do it
• 20:33
  Speaker 2: for a good while without autopilot before now.
• 20:36
  Speaker 4: Absolutely, and it applies. An accountant who's prepared hundreds of
• 20:39
  Speaker 4: tax returns by hand can spot the error in an
• 20:41
  Speaker 4: AI generated filing. AI the doctor who's made the diagnoses
• 20:46
  Speaker 4: without AI can override the confident but wrong prediction with
• 20:49
  Speaker 4: in the right circumstances, with the right support. And so
• 20:52
  Speaker 4: I think what that says to me is that professional bodies, universities, employers,
• 20:56
  Speaker 4: they need to preserve those training pathways to building it expertise,
• 21:01
  Speaker 4: even when the AI makes that look slow and inefficient,
• 21:06
  Speaker 4: because you're not going to be able to orches straight
• 21:08
  Speaker 4: to the AI.
• 21:09
  Speaker 2: Tough sell to people trying to cut costs, yes, and
• 21:12
  Speaker 2: a tough sell in the world where you're let's say
• 21:14
  Speaker 2: you're the company who says I have to preserve this
• 21:16
  Speaker 2: and I'm going to keep hiring and we're going to
• 21:18
  Speaker 2: learn this properly and then we'll use AI later. And
• 21:21
  Speaker 2: your competitors are saying so that we're going to fire
• 21:23
  Speaker 2: everybody around down at the bottom will certainly not higher anymore,
• 21:26
  Speaker 2: and our cost be lower, in our margins will be higher.
• 21:29
  Speaker 2: And as you say, that's going to be a bit
• 21:30
  Speaker 2: like a sugar rush, right, That'll work for a certain
• 21:32
  Speaker 2: period and it won't work anymore. But that transition period
• 21:35
  Speaker 2: of learning how we work with AI is going to
• 21:40
  Speaker 2: be tricky for a lot of people.
• 21:42
  Speaker 3: A lot of companies.
• 21:43
  Speaker 4: I think so. And the and the challenge with it
• 21:44
  Speaker 4: comes that it's happening so fast. It's in previous transitions
• 21:49
  Speaker 4: you've had time to work through these issues. And the
• 21:52
  Speaker 4: challenge here is that we don't have a lot of
• 21:54
  Speaker 4: time because this is coming at us so quickly. But
• 21:57
  Speaker 4: it's I do think as the way the Scottish more
• 22:00
  Speaker 4: which invests is over really long time periods, and there
• 22:03
  Speaker 4: isn't a great deal of value in something that benefits
• 22:06
  Speaker 4: the next handful of years at the expense of all futures.
• 22:09
  Speaker 4: And so I think what you have a question of
• 22:11
  Speaker 4: time horizon here is yes, you can save money in
• 22:14
  Speaker 4: the short run, but if it's going to have a
• 22:16
  Speaker 4: real long term cost to your ability to operate your business,
• 22:20
  Speaker 4: then that that's a false economy.
• 22:23
  Speaker 2: Okay, So let's think about what we can do, both
• 22:27
  Speaker 2: as individuals and as companies. If you are a young
• 22:30
  Speaker 2: person entering the workforce today, what do you need to do,
• 22:34
  Speaker 2: what are the skills you need to acquire, and how
• 22:36
  Speaker 2: do you need to look at your employment. I know
• 22:38
  Speaker 2: there is one path to take, which is simply to
• 22:40
  Speaker 2: say I'm not getting involved in this and I'm going
• 22:42
  Speaker 2: into I'm going into a craft I'm going into roofing
• 22:45
  Speaker 2: or stone carving or we have a huge shortage by
• 22:47
  Speaker 2: the way of craftsmen in the UK. So this is
• 22:49
  Speaker 2: a fantastic thing if people are decided to make their
• 22:52
  Speaker 2: choice in that direction as well. But nonetheless that's not
• 22:54
  Speaker 2: going to absorb. But that's not going to absorb a
• 22:56
  Speaker 2: million kids a year.
• 22:58
  Speaker 3: What else? How do you look at your path as
• 23:01
  Speaker 3: a young person?
• 23:02
  Speaker 4: I guess the way I fame it is the people
• 23:05
  Speaker 4: who will thrive are not those who use AI the most,
• 23:09
  Speaker 4: but those who can still think without it. And the
• 23:13
  Speaker 4: messages don't take the easy path, don't take the shortcuts
• 23:18
  Speaker 4: to productivity that these tools give you. It's putting in
• 23:21
  Speaker 4: the hard yards. It's doing the things that you don't
• 23:25
  Speaker 4: want to do, that effortful struggle and generate build, investing
• 23:29
  Speaker 4: in yourself and your own capabilities that will be most
• 23:33
  Speaker 4: important over the long run. And it's it's master a
• 23:38
  Speaker 4: task first before you start utilizing these tools to make
• 23:41
  Speaker 4: yourself more productive at.
• 23:43
  Speaker 2: It and at the company level. So that's individuals. But
• 23:46
  Speaker 2: what does a company need to do? I think we've
• 23:49
  Speaker 2: slightly come up with the answer to that. They need
• 23:51
  Speaker 2: to keep paying up to make sure that their employees
• 23:53
  Speaker 2: have a base of knowledge and skill beyond that.
• 23:57
  Speaker 4: Well, I do think it's it's really hot because you know,
• 24:01
  Speaker 4: you have this this constant battle for survival in the
• 24:06
  Speaker 4: corporate world. You have to be gast effective. But if
• 24:09
  Speaker 4: it seems to me very short sighted if you if
• 24:12
  Speaker 4: you allow the current generation of experienced professionals to get
• 24:16
  Speaker 4: a huge productivity bump from these tools, and that you
• 24:20
  Speaker 4: don't think about how how you replace them. You know,
• 24:24
  Speaker 4: how how the next set of you know, the next
• 24:27
  Speaker 4: the next generation of professionals will actually operate because because
• 24:32
  Speaker 4: we also we can all see the flaws in these tools,
• 24:35
  Speaker 4: you know, a deep understanding of how they work, an
• 24:37
  Speaker 4: ability to you know, there's a whole set of of
• 24:40
  Speaker 4: cognitive skills which will be necessary for the next generation,
• 24:43
  Speaker 4: which which weren't necessary for our generation. You know, the
• 24:46
  Speaker 4: the ability to orchestrate these tools, the ability to break
• 24:50
  Speaker 4: down tasks and in in such a way that you
• 24:53
  Speaker 4: can use these tools to answer the sub questions and
• 24:57
  Speaker 4: recompose that information. You know, there's there's a whole set
• 24:59
  Speaker 4: of cognitive abilities which are necessary to use these tools effectively.
• 25:03
  Speaker 4: But it's don't neglect the basics, the fundamental understanding of
• 25:08
  Speaker 4: the job that you're trying to do and the ability
• 25:10
  Speaker 4: to solve problems and investing in people with the capability
• 25:13
  Speaker 4: to do that, because otherwise it's a false productivity.
• 25:18
  Speaker 3: And what about investment implications.
• 25:22
  Speaker 4: Honestly, I don't really think that there is a particular
• 25:25
  Speaker 4: investment implication from this. I went into it because it's
• 25:27
  Speaker 4: something I was interested in. I was interested in. Now
• 25:29
  Speaker 4: it is affecting me. I was thinking about what it
• 25:32
  Speaker 4: means for my kids. I think that this sort of
• 25:34
  Speaker 4: narrower investment standpoint, it is this really tricky balance between
• 25:39
  Speaker 4: driving productivity and making sure you have the capabilities for
• 25:42
  Speaker 4: the future.
• 25:44
  Speaker 3: How nervous does it make you?
• 25:45
  Speaker 2: Because I think people listening to this conversation will come
• 25:47
  Speaker 2: out of the other end thinking themselves that it's almost
• 25:50
  Speaker 2: a given that the majority of people will use AI
• 25:53
  Speaker 2: passively in the dangerous way that you have suggested, that
• 25:57
  Speaker 2: this will be the default, This is what most people
• 26:00
  Speaker 2: will do, and there will be a small group of
• 26:03
  Speaker 2: people who will not, and who will engage with it
• 26:06
  Speaker 2: in the other way that you suggested, using it as
• 26:09
  Speaker 2: an assistance, a learning tool with which they engage, but
• 26:13
  Speaker 2: still think for themselves. So it rather sounds as though
• 26:17
  Speaker 2: that leads us into a world of very unequal outcomes,
• 26:20
  Speaker 2: even more so than the world we live in now.
• 26:23
  Speaker 4: Yes, I completely agree with you, and that's come back
• 26:26
  Speaker 4: to the title of the paper. AI isn't come for
• 26:28
  Speaker 4: your job, It's coming for your mind. If you passively
• 26:30
  Speaker 4: accept that outcome, then the question is what job are
• 26:34
  Speaker 4: you suitable for in this world where there's some set
• 26:37
  Speaker 4: of activities. And I think a useful model for this
• 26:41
  Speaker 4: is don't thinking. Don't think of AI replacing jobs, think
• 26:44
  Speaker 4: of AI replacing tasks. And a job is composed of
• 26:48
  Speaker 4: multiple tasks. Now it's AI will replace humans staying some
• 26:53
  Speaker 4: of those tasks. That is inevitable. But it's like, what's
• 26:56
  Speaker 4: your worth as a human worker. It depends what you've
• 26:59
  Speaker 4: let AI do your mind? But I think that you're
• 27:03
  Speaker 4: in equality point. Do you take the example of Meta,
• 27:06
  Speaker 4: the owners of Facebook. It was last summer that they
• 27:10
  Speaker 4: were going they went through this hiring spree their superintelligence,
• 27:14
  Speaker 4: their AI systems unit, and the rumors were that they
• 27:17
  Speaker 4: were paying individuals hundreds of millions of dollars to come
• 27:19
  Speaker 4: and do this. And the rationale was that you're spending
• 27:23
  Speaker 4: billions of dollars on these systems. What you want is
• 27:26
  Speaker 4: a small group of engineers who can really keep that
• 27:28
  Speaker 4: whole system in the head. And even if you're paying
• 27:31
  Speaker 4: them hundreds of millions of dollars, that sort of is
• 27:33
  Speaker 4: dwarfed by the amount that you're spending on the hardware
• 27:36
  Speaker 4: within that company. You suddenly get this big discrepancy, this
• 27:40
  Speaker 4: big inequality between those small number of individuals and the
• 27:44
  Speaker 4: engineers across the rest of the firm. But what you
• 27:47
  Speaker 4: also see is that in other parts of the business,
• 27:50
  Speaker 4: you're reducing the number of people because you're getting efficiency
• 27:54
  Speaker 4: from using these tools, and so you're laying off people.
• 27:57
  Speaker 4: And you go from this world where there was great
• 28:00
  Speaker 4: equality in the workforce and what people are getting paid
• 28:02
  Speaker 4: to suddenly you're paying people hundreds of millions of dollars
• 28:04
  Speaker 4: and firing people at the same time. And you get
• 28:07
  Speaker 4: that within a single company, but then you start to
• 28:10
  Speaker 4: get that across companies, which are the ones that have
• 28:12
  Speaker 4: been able to adopt these tools and embrace them and
• 28:15
  Speaker 4: win competitively, and so you get the inequality between companies.
• 28:19
  Speaker 4: And then it's between countries as well, because whose AI
• 28:23
  Speaker 4: are we using. In the UK, we're mostly using AI
• 28:28
  Speaker 4: systems from the US and from China, and cheap power
• 28:32
  Speaker 4: then becomes another really important factor in that. And we
• 28:34
  Speaker 4: have very expensive power. So you can see these fall
• 28:37
  Speaker 4: lines of inequality in widening rapidly.
• 28:41
  Speaker 3: Yeah.
• 28:42
  Speaker 2: Yeah, final thought, kids going to university these days. We
• 28:45
  Speaker 2: talk about this a lot on the podcast, because there's
• 28:47
  Speaker 2: the extraordinary cost of student loans and the handicap that
• 28:50
  Speaker 2: gives you once you leave, effectively paying nine percentage points
• 28:53
  Speaker 2: of extra income tax every year, et cetera.
• 28:55
  Speaker 3: And the question now is.
• 28:56
  Speaker 2: Are you learning anything valuable at university? And I hear
• 29:00
  Speaker 2: more and more from students that their exams are at home,
• 29:04
  Speaker 2: open book, twenty four hours, etc. And a lot of
• 29:08
  Speaker 2: them will then use AI for that, I'm sure, which
• 29:12
  Speaker 2: sounds like it makes the entire university experience not just pointless,
• 29:16
  Speaker 2: but worse than pointless. So it's only worth going at
• 29:21
  Speaker 2: all if you engage with your learning in the way
• 29:25
  Speaker 2: that we've just discussed. So for an awful lot of people,
• 29:27
  Speaker 2: this may be a pointless expense.
• 29:29
  Speaker 4: I think that's a real risk. I'd say my point
• 29:34
  Speaker 4: of optimism on this is I think if you look
• 29:36
  Speaker 4: back through some of these previous technological waves if it
• 29:39
  Speaker 4: takes social media again. For our generation, we were hopeless
• 29:45
  Speaker 4: users of those systems initially, and people made all sorts
• 29:49
  Speaker 4: of faux paers, etc. But then the next generation who'd
• 29:51
  Speaker 4: grown up with those tools were very very smart at
• 29:54
  Speaker 4: how to use them and didn't encounter the same sort
• 29:57
  Speaker 4: of set of issues that they were native. They were
• 29:58
  Speaker 4: digitally native, and I observe similar pattern with ai telogy
• 30:04
  Speaker 4: in my own children. I used to worry about are
• 30:07
  Speaker 4: they doing exactly what you said? Are they taking their
• 30:09
  Speaker 4: school work and just outsourcing it? But I actually see
• 30:12
  Speaker 4: them using these tools too, not because I'm leaning on
• 30:15
  Speaker 4: them the whole time, but I'm just watching what they do,
• 30:18
  Speaker 4: and they don't outsource the tasks they're actually these are
• 30:21
  Speaker 4: fantastic tools for learning if you use them appropriately. And
• 30:25
  Speaker 4: so you know, it's it's back to that sort of
• 30:28
  Speaker 4: personal responsibility point. Yes, you can go and skivee your
• 30:31
  Speaker 4: way through university and outsource all the learning and end
• 30:34
  Speaker 4: up with lots of.
• 30:35
  Speaker 3: Come out with your worthless two to one.
• 30:37
  Speaker 4: Yeah, yes, exactly, yea, but you can have a very
• 30:39
  Speaker 4: different experience, and you want to have a much better
• 30:43
  Speaker 4: experience because effectively you have an always on infinitely patient
• 30:49
  Speaker 4: tutor who can answer you every.
• 30:51
  Speaker 3: Question, sometimes incorrectly, but I do think that you.
• 30:57
  Speaker 4: Know, you have to learn how to think fundamentally to employable,
• 31:01
  Speaker 4: you have to be able to think for yourself, and
• 31:04
  Speaker 4: that's what you need to go to university to learn
• 31:06
  Speaker 4: to do.
• 31:07
  Speaker 2: Okay, there we go. There's your takeaway. Everyone, keep learning
• 31:10
  Speaker 2: to think for yourself. Tom, thank you so much. That
• 31:13
  Speaker 2: was absolutely fascinating and just to repeat to everybody, we
• 31:15
  Speaker 2: will put the link in the show notes so you
• 31:17
  Speaker 2: can go and read this for yourself.
• 31:19
  Speaker 3: That's really interesting. Tom.
• 31:20
  Speaker 4: Thank you, Thanks Marian for having me.
• 31:23
  Speaker 3: Thanks for listening to this week's Marin Talks Money.
• 31:25
  Speaker 2: If you like our show, rate review and subscribe wherever
• 31:27
  Speaker 2: you listen to podcasts, and keep sending questions or comments
• 31:30
  Speaker 2: to Merriorn Money at Bloomberg dot net.
• 31:31
  Speaker 3: You will also follow me and John on x or Twitter.
• 31:34
  Speaker 2: I met Mary nurse W and John is John Underscore Stepic.
• 31:38
  Speaker 3: This episode was hosted by me.
• 31:39
  Speaker 2: Mare zumzep Web was produced by Sammersadi and Moses, and
• 31:43
  Speaker 2: sound designed by Blake Mabels and Aaron Casper.
• 31:45
  Speaker 3: A special thanks, of course to Tom Slater.