Oliver Spalt

Universities should be, if in in an ideal world, future labs. It should be that you know the smartest people should have a room to come together and to try and tackle tomorrow's problems with state of the art thinking that we have today. So it should not be actually that also different from you know some people say well universities are there to just

you know, teach stuff that we already know. I don't think I I think there's a role for that, but I don't think you need universities for that really. It's like what ideally we would have the smartest people come together and thinking about the toughest problems based on the knowledge state of the art that we have today. So future labs, right? The university as a future lab.

Welcome to Tastebench. I'm Philip Zahn, my co host is Bruno Marnette and this is a podcast about taste and the creative side of AI. Our guest today is Oliver Spalt, Professor of Finance at the University of Mannheim, where he studies how people and firms actually make financial decisions. His focus is on behavioral deviations from the rational paradigm and how concerns beyond profit maximization are relevant.

Thank you for joining, Oliver. It's a pleasure to have you. we start

Thank you for having me.

out. This is we want to talk about research and AI this this time. as a starting point. Can you say what your research is about these days?

yes, so my research is about financial decision making of households of firms, the biases in decision making, the stuff that goes wrong and the things that influence decision making. So I am these days most interested actually in what we call non pecuniary utility. So morals and finance, how does that go together? How does that influence decisions?

so that's that's a that's a big one. Of course, has you know economic implications, political implications, super interesting field, ton that we don't know yet. this is what I'm trying to work.

Alright. Now, a couple of years ago, there was this big splash with ChatGPT The whole thing has dramatically evolved, and there is now a lot of influence, a lot of discussion on that science changes. What is your personal perception of this currently? How has science already changed? How is it changing? Where

are things going?

Yeah.

Well I I I I think there can be no doubt that this is a huge change. and we are in the beginning, but already now, you know, it's it's like there's this huge wave coming and you already now all all boats are lifting. So it's like

Mm-hmm.

it's it it's really big. So I think if you look at our PhD students, I mean they

They are no now all using AI tools. If you look at you know more senior or call it older researchers like like you know people like me, it's you know you're you're trying to also you know sta stay stay ahead of the curve a bit and sort of get acquainted with what's possible and so you know of course, you know, I'm also looking

at Claude Code and thinking how can it help me with freeing up some time and of course I do realise that everyone else is doing the same so the time I free up now I can enjoy for a few weeks and then everybody else is doing the same and then people will expect more output and all that stuff and then all this nice advantage is gone again. but yeah everybody is is is doing it, you know, when people are writing papers.

you know, spell checks were yesterday. now, you know, write me a draft and I'll critically read it. Might be more common. and hopefully the critical reading part is something that people don't forget. You know, I think the big concern is that you'll have people just churning out papers after papers and quality goes down. I think that's a valid concern and it's an important one.

and there's and there's a lot of broader

Are you ready? Yeah.

questions that you guys I I know you guys want to talk about,

Mm-hmm.

which is has something to do with where is this all going.

Can I can I interject one or two questions? Sorry, Bruno, if I made two. The quality has it's been going down. Couldn't you couldn't you argue that has been the case before?

say again.

is it really new that the quality has been going down? In a sense of we there is a massive amount of papers already before. Maybe the the Yeah.

I don't know. I I I should

I so I I should be more careful when I say this. I I think

Okay.

currently, s as of now, AI has two effects. and that has something to do with the two tails of the distribution. So the skill distribution has people that, you know, try to write good papers, were not really able to write good papers. And for those people, probably their papers are getting better.

Okay, because

Hmm.

it's like it's like if you're a s if you're a student and you don't know how to write a thesis, I I think, you know, AI helps you and it gets you over the hump. Now, on the top tail, like for the top people, AI is, at least that's my perception, not yet better than humans at the top. And I I don't mean only the tibbidy top, but like

you know, there there

Mm-hmm.

there's like a a a large percentile in research where, you know, I I think top say a top five econ journal, a top three finance journal, AI currently still has a hard time writing a paper at that level. it's not clear that you know I my personal I I I think this is the state now and probably in in in in sh in a short period of time this will change, but right now

For the people on the top, if you write that's a trade-off, you can write more papers, but the quality will go down just because the AI is not able to write the best possible papers. So this is yeah,

Thank

you know, it depends on which shale of the distribution we'll look at.

How would AI help write better papers? How would it have to change to do that?

Well it I mean it can already do so now, right? AI I mean

Yeah.

AI it Claude Code is a fantastic research assistant, right? It can it can pre structure data, it can you know, find papers for you that are potentially interesting, it can help you try and craft arguments or find hole in arguments.

I mean not to speak of the more mundane like just spell check this for me and update all section numbers and all that stuff. that's all that's all good and that helps. but right now what you d I'm I mean you know, I I I I cannot say that I'm you know, it's it's like I I I'm not a a a native in in all this, right? I'm I'm just a user trying to s stay with with the pack as everyone else does. But I d I think that mm

As of now you need the human input and if you i especially if you want to do absolute top stuff, you still need the the human and the relationship is more like RA and professor. but I but you know this this will change, I have no doubt about

Are you already seeing signs of people using AI to come up, not just with the implementation, but the ideas themselves? Do you think we're already letting AI take care of coming up with ideas? Or is it just...

I'm I you know I know for a fact that you know all our PhD students are trying that. The the

Okay.

pr the the problem is that and this is what I was saying, if your goal is top journal level research, I have not yet seen AI generated ideas that would be good enough to make it.

So you can say, okay, maybe I yeah, you know. I I I I i if it turns out a million ideas, probably there will be a few that are good enough. But like you know, on on average it's it's it's not at that level yet. And I think it's not a close call as of now. But it's you know, the stuff now is light years better than the stuff that it turned out a year ago. So in that sense.

I I mean I personally would not use it as ideas, but again you can use it as mm and I know that s I know people who are using it that way is as a conversation partner, right? You're you're trying to chat about stuff and you're saying, Well, what about this? and then you know, you just go back and forth and so you are well essentially again you're using it as an as a conversation partner as an RA, but in a more fast paced conversational

Can I ask maybe to, because I'm not really sure in your field actually, what

Mm.

does a good idea look like? Is this, you know, what makes, I guess the ideas coming from AI not so interesting.

well a good idea is one you know, and it's always this trade off. So you want something that is big and profound and you know s is very fundamental, but at the same time it has to be if we're talking about empirical papers, very well identified

measurable. I mean so so data availability, that's that's a big constraint, right? What is the data that we have? And so to come up with new ideas based on old data that people have been working on for 30 years, that's not easy. And so a great idea is one that makes a profound impact with available data in a well identified manner. And

Where AI struggles is is is really two fronts. One is it is a great idea, but there's really not the data or the method to do it. That's one of the mistakes that it makes. And the other mistake is it comes up with something that's implementable, that is doable, but that is it's just not so profound. You know, it's just an epsilon improvement about over something, but the Journal of Finance would not publish.

An epsilon improvement. They're looking for more.

Actually, can you maybe, it must be very hard, but can you try and define what profound would mean?

Well profound you know, it means does it change

people's priors in a significant way. You know, it's it the biggest

Mmm.

papers say, okay, this is how we thought you should invest, and now all of a sudden this paper comes along and it's like, oops, we should perhaps be thinking also about another thing. You know, that that type of that that that type of of of benchmark. I mean that's of course the highest level. It's like we we thought it's we thought it's X, but actually it's Y.

And then you know, people's priors change and that changes the direction of the field. Those are the really big, profound papers. Smaller papers are like, well, we we're in this very niche thing and we thought these seven variables matter and I show you that these seven matter and an eighth one matters too. It's like, yeah, okay, that's perhaps interesting but

So does it change priors? That's that that's the that's the standard.

So if you're a PhD student that gets into a field, at least that's my experience, one of the hard things actually is understanding what is profound and what counts as profound, right? It's not it's not like that you look you read paper. I you can get an understanding for reading the journals, but it's not enough. How do you get there as a student?

Well it's it's very hard. I mean it's like I think and science is not different than, you know, say arts or or doing you know, entrepreneurship. It's like how do you get good ideas that get people excited? And you know, how do you come come up with a new symphony where people say, wow, you know the the Beethoven thing. People people hear it and like, wow, this is

First they say it's crazy and then they think it's it's just brilliant and it's taking us somewhere new or you know you you have the iPhone and it's like wow this is a totally new product, how did you get the idea? It's like that So it it's part I think it's part art and part science. So you know, how did how did Steve Jobs and and Beethoven and and you know John Nash get their ideas? It's like well, you know, part of it is just it happens.

And it's it's sort of skill and a knack for for finding it. But I think there's also just a a ground you know, ground ground so sort of background that you need. One is like just to know the literature, to if you read a lot of papers and you've been exposed to a lot of the top stuff that that you really get an appreciation why that is good, why that is profound.

That you have also been doing a lot of research yourself, that you've been running into enough dead ends to actually appreciate when something is really you know, and not obvious that you can pull it off, but then somebody can pull it off and you get an appreciation. So there is certainly learning by doing, but there's also just some innate skill that some people have and have not, and the

The the the ladder is of course unfortunate in science because you cannot really learn it. But the learning b by doing component is nevertheless s significant. So I do think you have to just try and get in there and and and and work on it. And and I think the exposure is is important. So, you know, trying to get yourself to come up with ideas at the highest level is

It's also dependent on the teachers you have. Let me put it like this. If you have you know, if if if if you have people that teach you that aspire to do top work, then of course your perspective is very different than if i i if your teachers also don't really know how to do it. I mean it sounds trivial but I think that's that's part of the

Yeah, the ambition. No. It's the same with entrepreneurship,

Yeah.

I think, right? I this is

many people make the argument about Silicon Valley that it's the ambition which counts there, which makes a big difference overall. will we have still research papers in the traditional sense in five years? Maybe or actually in three years, let's make it shorter.

In three years I I think yes, but not because I think that AI will not potentially be better than humans in three years in writing papers, just because I think there's a lot of inertia in the in the process. You know, there's still a lo there' there's a lot of old professors that like to write their papers and go to conferences and meet and chat and and university incentive systems are also not changing so rapidly. But

so so yes, I think there will still be papers. It's but but I think the underlying question is do I think and that I do I think that human papers will still be better than AI papers in three years? I I was I as I was saying, I I think actually for the run of the mill

paper towards the lower tail, I mean AI is already better now, right? the question is will AI be able to write top journal pa level papers? And I'm I'd be curious what you guys think, but I I've been thinking a lot about this and I do not see any structural reason why that should not be so. I think you would have to tell me there is something that humans can be better than the machine that i you know

People have have been saying, well, it's ideal generation, but I d I I don't know. I d I don't see any any structural reason why people why the machine could not generate fantastic ideas. How do I generate ideas? I have resources of stuff that is in my head, I'm trying to sort of you know connect the dots and think in crazy directions sometimes and then I'm like, oops, maybe this is interesting. There is nothing in the machine that that suggests to me, at least as a layperson

that it couldn't do that. So I think it's th you know, three years or not, I I I think it will not be long.

until at least with you know, there's a chance. I mean who knows? I don't have a crystal ball, but I think my my prediction is that it there's a good chance it will be like you know like chess computers. There w the the the machine will be better than humans and then the difference between chess and science is of course in chess, you know, we we still watch Magnus Carlsen in play, right?

We d we don't want to see two engines play each other. Why? Because it has entertainment value. That's different in science. If we c if we care about the cure for cancer, we probably you know, we couldn't care less if a human does it or a machine does it.

Will you write papers or will you write actually for AI?

what do you mean? Say again. Yes.

If AI is becoming powerful and is

creating new ideas and it's basically frontier research,

Yes.

will you actually write papers for humans or will you write them for the AI? And how would that look like if you do that?

I I

I I think I wouldn't write for the AI. I think the AI would write. I would be I don't know, teaching students for a while. Yeah, then no.

What do you do then?

I I I think this is serious you know. These are of course the the big question. If if if you take the

Take optimistic case and say the models very soon will match and at some point surpass even the best humans in generating ideas and generating knowledge in our field. Then the question is you know, is is there a role that humans can play? And I see I see one one that has to do with

data. So if there's proprietary data that the machine cannot access, then even with the you know with the best idea, then it's like you know so in the networks will probably so so very well networked people will at least for a while have an edge because you know suppose your government introduces a new program that needs to be evaluated.

If you are the guy they trust to evaluate this, you would basically have the AI write the paper, but since you are providing the link to the data, it will also be your paper. So there is still something in it. So the skill is becoming different then, right? You the the skill is to have the network and to just get the access to the data. So that's that's one way in which it it might work. And then there of course there's for a while there will be probably the

You know, the the best people that that have sort of the key to the AI lab that that that just sort of but you don't of course you you don't need thousands of them probably, right? You you you need a few.

So it's more like a superstar system, even more than it was before.

I I yeah, I think so. I mean a as long as you as long as you need them as as long as you need humans anyway. Right again, chess. I mean ch if if if our goal was to play the best game of chess, there would be no need for Magnus Carlsen And and so in that sense you could also say, Okay, maybe, you know, some institution, I don't know, be it a state, be it an enterprise, be it whatever, you know, sets up

these AI agents and this is what generates knowledge.

I'm curious to you speak of entertainment in chess as if there was no entertainment in academia. But wouldn't you say that there's also that part of the purpose of the paper when you write it is to kind of entertain the audience to some extent.

yes. I well no. Mm no. So so I I was it I think the entertainment is is if it if you're lucky for yourself, right? a lot of scientists are passionate about what they're doing. You get satisfaction from pursuing a project. just you would get satisfaction from playing chess. But I I don't think so

if you had a cure for cancer by an anonymous system or

an a good attempt at a cure for cancer by a human that fails. I mean we would not you know, I

Yeah

mean it's like we would not want the nobody cares about the human and his failings, right? It's in that case. It's like we want the result. And so I I think the entertainment aspect is is it's I I don't I don't know, I I see it I see it less. Of course you might have some people whose job it is to translate

what the AI find, you know, into language that humans understand, perhaps politicians, perhaps the broader public, right? There will be a few guys, pretty much like it is now. You have a few guys, everybody has their favorite, I don't know, X or Substack or whatever whatever it is they're looking at, their favorite guys who is sort of talking about what's going on in their respective fields. I think

Th those things and that is entertainment in a s it's it's information but it's also entertainment. I think th this will probably stay for a while. But but the pure science, I think the one the people who are un entertained are the scientists. You know, science is a great profession if because a lot of people genuinely love what they're doing. But it's a personal yeah, that's a personal value of entertainment. It's not I don't think that many people would like to pay me

to see me being happy about writing my paper.

Actually, maybe just because it relates so much to the question of taste. I think

Yes.

if you look at breakthroughs in physics, for instance, sometimes comes through a path which is quite theoretical. People have played with theories and part of the drive was the beauty of parallel equations. There's some notion of aesthetics in science sometimes, which can guide towards originality.

Yes. Mm-hmm.

And I guess in my mind, these humans today are maybe better than AI at appreciating this sort of aesthetics potentially. Or do you feel there's again, that there's no reason why an AI wouldn't just develop a similar taste?

Yeah, I don't know. Do you s do you do you think

the question is what is what drives our ability to perceive and evaluate something aesthetically? Is that something that's structurally different between a human and an AI? So is there any reason to believe that, you know, if we're looking at an equation or a or a picture or a piece of music, is there anything that

suggests that humans are have su what would be this innate ability that would not be copiable by a machine? I mean couldn't it just read a million you know reports from from critics or to to figure out what peop what it is that people find aesthetic or what are the criterion and it just apply it? I I I think you're right. You know I I think

You know, in in in finance, you know, this may be a very financy thing to say, but a lot of people find the Black-Scholes formula to be extremely elegant. You have a very complicated

Mm-hmm.

problem and you break it down into s a thing with four components, you know, it's like or you know, Markowitz's way to think about portfolio diversification. It's it's beautiful. Or you know, the CAPM equation.

it it it collapses something very complicated into a very, very simple equation. It's it's beautiful in its own way.

But I but f first of all, you know, the the question is of course and I'd be curious, I don't know. I the what is the fraction of big insights that have come on the back of an aesthetic intuition rather than

yeah, just the the the the the hard sciences

Mm-hmm. Hard science. Yeah.

I I I I don't know and and the second is you know it

I don't again I I and I'd be curious. I might be totally you know, I I I just I don't see a s a structural reason why the machine could not also replicate that part. But again, you you guys are the experts. I do you have do you see a reason?

I know some of the arguments. Some of the arguments to say machines cannot do it often relate to embodiment. There's this theory

Mm-hmm, yes.

that without a body you can't really have emotions.

Yes.

Another theory could be that machines are disinterested and that you kind of need to almost care about the specific things to develop a taste in that aspect. But again, that's more of a philosophical argument. I don't think

Mm.

that's a very kind of harder.

decisive argument. And then I guess it's just empirical evidence that so far when it comes to judgment, sometimes machines may not be that good yet at distinguishing a good idea from a bad idea. They may be very good at generating

Yes. Yeah, yeah, yeah.

a lot, brute force. mean dice could already generate good idea, right? If you throw them enough, often enough. But the judgment

I think also

part is where...

potentially we can empirically feel that there is a gap.

Yes, but but I think it depends on the gradient, right? So it's like I I I'm totally with you that right now, you know, I think hum the the good humans are better to to to to evaluate ideas. But but but the gradient is very steep, so the question is what yeah, Philip asks about three years. So my my personal I don't know. I would say that the chance is at least fifty fifty that

Yeah.

Go, Philip.

I

mean on the sorry, on the on the a aesthetic side as well. There is there I think there is a there's very often a package that we get which something looks which looks beautiful is also simple, right? And simplicity, I think there are some operational ways even now that we can think about this in terms of computational or information theoretic aspects of it, right? To be compressible, etc. So I think that part is in some sense there is and this is the weird thing, right? We have this kind of coincidence of something very often looks that looks also symmetrical, for instance, is

more beautiful for us than often, not all of us, than other things. And the same is for simplicity, right? And parts of it will be just captured even within the current models that can kind of go for that. maybe let's switch gears a bit because we talked about publication and the creation of it. Publication also plays a huge role in the allocation of resources in universities and the organization. How do you see universities changing when AI produces the papers or the research?

Yes. you know, it's I mean again, I I can only speculate, but but it's

Please.

of course conversations that that I'm having with people around me and so

So first of all let's let's let's assume that there is this world where very soon the AI will surpass humans in its ability to write papers. I think the natural response would be that you know, your university would not feel the need to pay researchers substantial amounts of money to do something that a machine can do better. Right? So there is i

Unless you're you're saying is there some value of having humans generate the knowledge? And as we talked about, it's it's it's not clear if it's if it's knowledge that we want. You know, in Germany there's state universities. Should the state pay people to engage in their hobby and do research when machines could do this easier and generate all this knowledge? Probably not. And then

And then you you know, there's there's inertia in institutions and there's a lot of that. So, you know, it will take a while before these changes take into effect. Private universities, different thing. If you're a private university, you could at least in principle make these changes much quicker than state universities. you know, you you just cut back hiring

People in research positions substantially. Actually, not only at universities, right? You you have all these institutions that have big research departments, think about central banks and think tanks, also those guys. It's like perhaps even more. Because at universities, the other part of what you're doing is of course the teaching part. And there I think I mean we haven't talked about the teaching yet, but there I think

Some of the same forces are playing out, but clearly the time frame is much longer. So right now and there's a fighting chance that we have an argument what humans you know, I'm always on the lookout for this argument. What is it that humans can do better than machines? And if the answer is pretty much nothing, well then we're doomed anyway. But but

Yeah.

what what what could what could be a thing? One one thing might be that

You know, as humans were programmed by evolution to be social animals and to have people I mean you mentioned embodiment, like you know, to have to have people with bodies in the same room is something that that is that that gives something to you that cannot be replicated by staring at a computer screen. now I would

think that some of my current students would disagree with that statement because the knowledge they're getting or seem to be getting is coming only from screens or from you know an audiobook

Hmm.

that ChatGPT generates on the basis of some set of lecture slides or something like that.

Mm-hmm.

I I have my doubts that is so effective on the whole, but it's you know it's it's what some people do. But if we

If we if we keep this idea and it's I think it's an open question if that's an idealistic idea by old people like us, you know. It's like you know, the horse carriage will always be around or something like that.

Ha ha.

Perhaps not, right? But but if you think there's something about our bodies and just humans are social animals and they

Mm-hmm.

want to be together. In that circumstance there will

be a need for teaching going forward that that you need humans to be filled as at least as long as the human humanoid robots are not here, right? If that happens then you know

Yeah, good night. At the margin, what are you changing right now when you're looking at let's say this year, next year in your teaching?

well this year next year I mean well th th th there's a few sort of immediate things that so one thing is we're bringing back oral exams.

Mm-hmm.

Because you know, writing a thesis

You you're at home, you're writing a thesis with AI. I mean, that at a minimum it means, you know, i not long ago a thesis could be, you know, replicate this old and important paper with new data. And it took people three months. Now

And they learned a lot.

and and they learned a lot. And now it I mean you you know yourself, right? It it may take them, I don't know, half an hour to get the same thing.

Hmm.

And so

But but of course just just telling Claude Code to replicate PayPrex is not teaching you anything. So one thing we do is we bring back the oral exam. So we say you can use AI because that's also what is going to happen in the future if

Mm-hmm.

you're going to work. But you have to be able to explain what it does and how it works. that's that's one of the changes in in in diseases.

I have a, yeah.

Yeah.

I want to go back to the social aspect. you were asking the role of a professor in a university where they get hired

Mm-hmm.

and so on. I don't know how things are in Germany these days, but I know in the UK, one thing professors complain a lot about is that usually they get hired to bring grants to university. part of their job is to come with money into the university so that the university has more

Yes. Mm-hmm.

money.

And in many ways, professors are becoming salespeople, right? Like part of the job is to really like sell a specific research direction. And you could also imagine on the student side, part of the job of a teacher is to attract more students by providing quality of service almost. Do you feel the... Would you agree this is a similar theory to kind of the social theory? we would have like a social...

Essentially the staff at university would have more of an almost more commercial position.

I f well look I I think this is already happening now, right? that every university has is on social media and is trying to broadcast what their researchers are finding. and and and to be fair, professors always had to be salespeople as well, right? Part of your job is to write papers, but part of your job is to promote papers. You get invited to other universities, to other conferences, you you want to

You know, you have the product and and you d you you want to tell people it's fantastic. So par i part of it has always been a sales function. Now the the the third money thing that depends very strongly on the institution, how strong the emphasis is and it's also very different across fields. I think in some fields you can basically not survive without third money. in in equal and finance I think the

Pressure is not so high compared to these other fields. So yeah, I don't know, but but I I I don't think there's a fundamental change other than of course, if AI is writing all the papers, what would you be selling after all, right? I mean for a while,

That's the question,

of course, you can for a while, for a short while, you can of course sell the more papers that you are writing. But but again, you know, you have to also think about

What is the downstream environment going to look like? What will journals look like? What will journal space look like? Who's going to read these papers? Who's going to referee these papers? Who's going to edit these papers? there will be also profound shifts in that space. So I I have heard people say that some of the top journals are already experimenting with AI referees, AI editors.

You know, if if that happens that cuts humans out of the picture there as well. And I think what would what would remain, so i in my view is well in my view what what what this all means is that it's an opportunity to rethink the role of universities and and sort of with it, you know, what what we think the role of humans is. And I do think

Well here's my two cents. Okay, that's the best I could come up with. The the other is the bleak okay, there's no use at all and we we we'll we'll just see if we we end up in the good AI future or the bad AI future. But I th I would think that

Universities should be, if in in an ideal world, future labs. It should be that you know the smartest people should have a room to come together and to try and tackle tomorrow's problems with state of the art thinking that we have today. So it should not be actually that also different from you know some people say well universities are there to just

you know, teach stuff that we already know. I don't think I I think there's a role for that, but I don't think you need universities for that really. It's like what ideally we would have the smartest people come together and thinking about the toughest problems based on the knowledge state of the art that we have today. So future labs, right? The university as a future lab. And if you have the university as a future lab, I think AI could be fabulous if you do it well.

It's like it it it gives us the ability to sort of just research so much more at much higher speed and look at a ton of perspectives. So I think you what you could do is you know, you you could also I mean a a a a very obvious thing is you could take inverted classrooms much more seriously. You y you let people just research on their own and then they they come and and together you're developing something.

So but and that's a big but, you know, it would be fantastic to have these future labs, but of course these future labs do not go together well with I don't know, a big lecture hall filled with a thousand students and one professor. So there's also profound

Hmm

changes in how you would actually have to set this up.

How

would you set it up? So if you if you have resources and you can create the University of the Future, what would be the first thing you

I I think

do?

well well you know if I could think of what what would be a an institution that I'd love to be at, it would be one that is is s small, has the opportunity to have I don't know, if you have a seminar with twenty twenty five highly motivated people, that's that's an optimal size and you could really do on the fly big stuff.

if if they are motivated. that that would be that would be an instrument. I mean, you you have to really think about it. So in our finance curriculum and not just ours, like you know, you you pick your favorite university in the world and it's the same in in all of them. It's like a big, big, big part is like, you know, we explained the capital asset pricing model to you. And already today

Why should you?

that's actually not clear why we should

Right and and in the future this will this will become even more backward looking and it's you know, because AI is also an opportunity. I mean that's it's just fantastic, you know, the the amount of of stuff I can do with a data set with Claude code is amazing. And I can do it in in short order and I don't have to wait for a week for an RA to get back with something. I can get it done in seven minutes and then

thinking on it and I could also you know if we had a seminar we could turn our conversation into a into a seminar with students and we could say you know let's let's try and get up to speed let's come together let's let's perhaps in real time use ai to help us make progress and I I I think that would be that would be fantastic and then that I think those are the opportunities if I i and that the big if is if we have the you know

this this this human as a social animal. So if humans want to be together. But note, we are in this podcast now and we are not in one room, right? And I I think i isn't it true that some the machine is already not so bad in producing the image of a human.

while talking. So if that's true, you know, I I could be in AI and then we

Hmm.

would have this podcast, you know, you you would you need me? Right? It's like but yeah. No but I but but I think it would be nicer to yeah. Yeah.

Not yet. We still need you.

Yeah. We wouldn't care as much about an AI's opinion. I think that's the main thing, right? We wouldn't

Yes.

care that much about what they say. Maybe... Yeah. Yeah. There's

Yeah, the question is would anybody care about what you care about, very soon.

Let's not please not go there. I I still have PTSD from from this higher order stuff.

one thing you said that...

really speaks to me, which is this idea that teaching things that everyone knows that you can find in a book or ask a GPT about feels like a waste of time. And I completely agree that with the right implementation, the AI itself could be as engaging as talking to someone on a video call if the AI does this well. So there's all this great argument to believe in AI teachers. The one thing I'm still unsure about is curriculum.

I think there's value in having someone, you mentioned earlier, right? Like when you have great teachers that have high ambition, they would steer you towards learning the right stuff, right? Whereas if you're completely self-driven and you just get charged GPT to make a curriculum for you, it's probably not going to give you the top curriculum. It's going to assume you're an average person and give you the kind of the average learning, right? Do you imagine a way for teachers to be...

guiding what you learn as opposed to how you learn it.

Well again that depends on on on on how we think of of humans, right? So if I think in principle, what

Mm-hmm.

prevents you from saying, you know, design a course for me and assume that I'm top student in Harvard? It's like and and it might do that, right? I I don't know. So so so so actually that's one of the arguments for AI based tutoring, right? That it can say, Hey, you know

I'm a very visual person, so you know, make these arguments to me with graphs or I I I I learn more with equations, so give me equations and it can tailor this exactly to who you are and to your strength and what you need. but I think you know in this more old fashioned way, it's like you know, it's if you if a teacher is also someone you you trust and you look up to and you

you know, if there's there's there's this deeper human connection, then I think that's what the teacher, you know, to just to convey that something is I think a part of wha what makes a a a good leader or a good teacher is a leader in a sense. it is that inspiring people and a a la you know, giving them the sense that what they're doing is worthwhile and giving them sort of

hope that they can also do something great. And perhaps that's I I don't know I I I don't know, but perhaps humans are better in you know in giving other people this this this hope, these these ideas,

Spark

the spark, yeah, this inspiration. And that would be a role for teachers. That would be an optimistic case. And that, you know, then then there would be yes, yes. Yes.

as a coach essentially.

Somebody who is emphatic. I mean you said this earlier, right? It's like it's also that I it it's not just somebody who says, yes, I feel for you and this must feel horrible. You know, if ChatGPT writes this but you read it and you're like, it helps a bit, but not it's it's still different if another human says.

Yes. Oliver, we have already taken up a lot of your time. I I have one topic question, whatever you wherever you want to take it to close. I know that you have a deep appreciation for art and also for the creative process. When you mentioned the idea of how to recreate the university, the frame you gave it was basically as a university. But the option that I see is obviously why just le why just stop there, right? Why not cross it with you know, artists that are painters, musicians

Maybe getting back to an ideal from the old times of the the Bildung as we say in German, that the full formation of the individual. Any thoughts on this?

think one way to think about the AI revolution, if you want to call it that, is that it lowers the cost of acquiring knowledge also in fields where it was very hard to get knowledge in the past, right? so w with lower costs in that regard, it should be easier to actually reach across the boundaries and and to talk more with with other people. And and and and you know irrespective of that I would say that

i even in the pre AI world there was a lot of opportunities that we missed by not speaking more with each other. But perhaps AI could be sort of this also an a a bridge that that just helps us to well not just bridge so so

understand each other across disciplines.

Yeah, well ultimately the question is what does it mean to be human? And what does it mean to be creative? And what does it mean to be inspired? And what does it mean to put something in the world that makes sense to you and which you can live with? And those are not scientific questions alone, those are not entrepreneurial questions alone, these are not artistic questions alone. These are universal questions. And I think the

you know, one hope would be that we would understand all these things. all these as you said, all everything creation is a a productive inspired process based on ideas. And so if AI would allow us to to s to speak with each other i you know easier and break down boundaries more

That would be fantastic and I think you're right, a futures lab need not stop in finance or in economics. A futures lab, if you take it serious, most of the big questions are multidisciplinary anyway, right? how we want to live, how us you know f from climate change to migration to polarisation to I mean you you name your favorite topic and it's all

It's all has something to do with you know, there's a challenge, can we solve it? What do we need to solve it? It's creativity and being inspired. And that's that's something for the future slab. So yeah, I think you know, perhaps perhaps it frees up a lot of resources because then the mundane nitty gritty just repeating stuff that we already know, type of demand is is is lower because the machine takes some

Off that of our back.

Oliver, it was a pleasure. Thank you for your time and thanks for being part of this.

Thank you so much, guys. It was a great pleasure. Thank you.

Thank so much. That was fun.

Thanks for watching. We have got more episodes in the oven, so please like and subscribe to get notified when they are out. And please check out tastebench.com to read about the Taste Benchmarks and the creative engine we are building towards.