Wer führt hier eigentlich ?

Transcript

00:00:00Welcome to Think Different, Think AI, the podcast by Mark and Jens.

00:00:07Two tech-loving minds who don't just talk about artificial intelligence, they live it.

00:00:14Here you get clear perspectives, real hands-on insights and a fresh look at what is possible.

00:00:20Understandable, critical and always with a wink.

00:00:24AI to think about, to smile about and above all to talk about.

00:00:34Hello everyone, welcome to a new episode of Think Different, Think AI.

00:00:38I'm not alone today. Dear Mark is here as always.

00:00:43And one of our, let me always put it that way, regular guests is back again.

00:00:48Dr. René Deist has once again done us the honour of dropping by.

00:00:52And he has been with us twice already, René.

00:00:54You were here twice, and one of them was episode number five, if I remember correctly.

00:00:59Back then we talked about Boss Level AI, that's what we called it.

00:01:03That was a really funny episode, where we talked about

00:01:07who actually is the boss?

00:01:09Then we did another episode around the topic of Agentic.

00:01:13That was The Strategic Gold, Skills.

00:01:16That was Skills, exactly.

00:01:18That was episode 37 or something.

00:01:21we already have quite a lot of episodes under our belt. And today I'm glad that we

00:01:28can talk a bit about a combination of both things, because you haven't

00:01:33been lazy in the last few months either, but like Mark and I

00:01:38produced a lot, a lot of content, and one good piece of content you produced

00:01:42is a book that you have just released. With the title, Wer führt hier eigentlich?

00:01:47And Mark and I thought that this is a perfect moment

00:01:50to talk to you again and to welcome you as our regular guest.

00:01:55And I'm looking forward to us talking about it a bit today,

00:01:58about the future, about what is and about the future.

00:02:01Who actually leads in the world of AI?

00:02:04Welcome, Mark. Welcome, René.

00:02:07I was just about to say something.

00:02:10Yes, along the lines of, the guest gets welcomed too

00:02:12and is allowed to say hello.

00:02:14I wanted to say welcome as well.

00:02:16I was completely confused too,

00:02:17that you said hello Mark and René, but maybe at this point one more

00:02:21addition. For me the book actually was my reading

00:02:23material on the Kindle during my last holiday, because I had to promise my wife

00:02:30to leave the notebook at home on that last holiday and to travel with only a Kindle

00:02:36as an electronic device, and there were two things on it. One was

00:02:40René's book and the second was, let me put it this way, Claude Code had worked on the Kindle

00:02:46a bit beforehand, and then there was a bit of Claude Code on it too.

00:02:50Could be that there was something else on there.

00:02:53You'd rather not tell your wife that.

00:02:55No, nobody knows about that.

00:02:56Only part of the family subscribes to the podcast.

00:02:59If word gets out, let me know,

00:03:01then I'll come home a bit later that day.

00:03:03Right, cheers.

00:03:05Well, then let me also say,

00:03:07thank you very much again, this really is my third time here

00:03:09and I feel completely at home.

00:03:10Many thanks for inviting me again and again

00:03:13and your podcast is of course also a highlight

00:03:16among the podcasts that deal with high-tech and modern topics, and for everyone

00:03:24who is only listening, maybe we should mention that we can actually see each other on screens. We are

00:03:30not in the same room, but we are in the same software. And it's always fun with

00:03:35you. And Mark posted recently that his job content with us is also moving

00:03:43towards future, AI and future, so towards being an

00:03:47ambassador and evangelist, which is exactly the role.

00:03:50Yes, I'm glad about that, we work together very closely there.

00:03:52Yes, I feel completely at home and many thanks as well that you're doing

00:03:56a whole episode about my book. Not bad, not bad.

00:03:58Reading the book also takes longer than the episode, so from that

00:04:04point of view we can of course only scratch the surface.

00:04:07But before we get into the book, I mean, René,

00:04:11well, who am I telling. The world is really fast-moving. Last time we talked about

00:04:14Skills, okay, but while preparing this, episode 5, quite apart from the fact that

00:04:19that was August last year, it's only a year, but everything that has happened in technology

00:04:26in that time. And we had it in the office just this morning, something like this 5,

00:04:306 from OpenAI, the new model that came out with ultra thinking and what have you.

00:04:35And what that does to the models available up to now, both in terms of

00:04:41token cost, if you pay for it API-based, and in terms of power, that

00:04:48this whole topic, which a year ago for us was a bit of, okay, how do you deal with

00:04:52people, how do you take them along on the journey into this new world, which of course

00:04:57brings us beautifully to the book. And we have hinted at it once or

00:05:04twice before. You historically came over to us more from the

00:05:07musical corner and are very musically

00:05:11interested. And if you look at the book now, then there is

00:05:14among other things a statement in it that you don't really need

00:05:19conductors any more, you need composers, looking at

00:05:23leaders. Would you like to give us a first idea of what you

00:05:27mean by that? Gladly. Let me say up front, I deliberately wrote this

00:05:31book so that you can read it on an intercontinental flight.

00:05:36So deliberately thin, and it's primarily aimed at the C-level, because I don't dive

00:05:42into technological depth at all, instead my point is that a big

00:05:46change, well, we are right in the middle of a big change, and that's exactly it, that's how I look

00:05:53for images, and since music is one of my big hobbies, I of course took an image

00:05:58from the world of music. And it's about leadership. And let's say, in this time, up to today, where

00:06:06agentic AI is still just a tool, we have a leadership that at best motivates people,

00:06:13sets a clear direction, formulates a strategy, etc. The way we know it. But we also have

00:06:18a part of this job that is information broker. So the boss said this and I pass

00:06:24the information on and the employees then only get the information and then I collect

00:06:28it back up again. And this information broker thing, that's actually not

00:06:33hip any more anyway, but in the future we will exercise leadership in a different

00:06:41way. There are many facets, we'll talk about them today. And the one reason

00:06:45why I chose this image, from conductor to composer, to creator, is the following.

00:06:51A leadership function today really is like a conductor. There is a

00:06:56company strategy, that's the score, and then there are a lot of

00:07:00talents in the orchestra, there are the violins, there are the flutes, there are the

00:07:04timpani etc. And the conductor's job is to lead this

00:07:09company strategy, with his own activity, his own strategy, into

00:07:15the orchestra so that it sounds good, so that everyone plays it together,

00:07:19in a way that everyone is successful and it sounds harmonious. That really is leadership, that

00:07:24we are able to give everyone the space to develop, but to do something that together with all

00:07:30the others makes up a total work of art, that's fun, that's successful. And that is

00:07:36changing now. The strategies are still there, but I was recently at a Gartner conference,

00:07:42so this is not a number of mine, Gartner predicts that soon every

00:07:47employee will have 109 agents. I also wonder, why 109?

00:07:53Yeah, yeah, fine. Yes, 1 to 109, but let's say, for us

00:07:59it's 1 to 100, that's easier. That means every person will have

00:08:03100 agentic software entities in the future, working through packages of tasks for that

00:08:10person. Which means, if that's the case, it means that suddenly every person has to be

00:08:19able to formulate strategies, no longer one

00:08:23conductor, but everyone in the orchestra, the 100 things that hold the

00:08:30mallet for the timpani, that work the keys of the flute etc.

00:08:37That means everyone has to be that, and so the role is more that of a composer.

00:08:41So I write it down again and again and I pay attention to feedback.

00:08:44That was my thought, to get the image across.

00:08:47Now, sorry, Jens.

00:08:48All good, all good.

00:08:49I had a lot of points to pick up on, but first the first one, where the bar is maybe set a

00:08:54little too high.

00:08:55I mean, the book can probably also be read through nicely on a German train connection

00:08:59that happens to be running late again.

00:09:01Jens, thank you.

00:09:02Thank you.

00:09:03I don't need an intercontinental flight to read René's book, just as a little tip.

00:09:08Just take Deutsche Bahn or something else.

00:09:11One second, in case Deutsche Bahn wants to set the record straight here.

00:09:14You are of course warmly invited as a guest.

00:09:17There are long routes as well. There are long routes as well.

00:09:19Nice night express routes and so on.

00:09:21I was thinking of the Karlsruhe route.

00:09:22No, no, no. Come on, keep going.

00:09:23I've taken the night express before, it's very nice.

00:09:25Right, joking aside.

00:09:27So, of course it's interesting again that Gartner has already made a definition for itself

00:09:31of what an agent is and what is not an agent, where I think others are despairing over

00:09:35whether sub-agents count as well, that's where I'd already ask my first question, but we don't

00:09:38need to go into depth there. Would the image you just used maybe also hold up

00:09:43if you said, isn't it actually an orchestra inside an orchestra? I mean,

00:09:48doesn't the individual basically then actually have the ability, you know,

00:09:54to play their own orchestra, as it were. And then this composer, the one who

00:09:58stands above all that and has to bring it all into harmony with each other, is then

00:10:04probably operating on the different levels. That's an image that

00:10:07I think resonates with me even better, because I picture these agents almost always

00:10:11as having the capability of a complete orchestra,

00:10:15because there are things that maybe can't only be delivered

00:10:20by the timpani or can only come from the vocals corner or something like that,

00:10:26but in theory every member could actually put a whole orchestra on stage again, right?

00:10:30In the end it is, sure, I mean, I chose the image to capture the change there.

00:10:36But if you keep building the metaphor out deeper and deeper, at some point it may also

00:10:39become wrong.

00:10:40But still, if we follow the thought through, you are absolutely right.

00:10:45It's then the case that every person, so every flautist, has their own little orchestra,

00:10:49every drummer their own little orchestra etc.

00:10:52And those are exactly the, but that is the transformation, which, reassuringly,

00:10:58isn't coming all that fast.

00:11:00It's coming successively, in slow steps, it's just important to anticipate it now,

00:11:06because right now you can deal with it well and prepare for it nicely.

00:11:10In five years we'll be long inside it, then it will all have happened.

00:11:14And then, of course, you can also react, but then you have to react.

00:11:18Right now you can act.

00:11:19That's why the book is out now.

00:11:21Right now you can shape it, right?

00:11:22Exactly.

00:11:23You picked that up in the book too, and of course our loyal listeners can listen to the

00:11:28old episode.

00:11:29But I'd still like to go into it a bit again, also into how the

00:11:32terminology and the shape of it may have changed since that episode.

00:11:37In episode 5 you talked about Intent, Operate and Check.

00:11:41If I now, well, reading it is actually a few days ago, yes, at this point

00:11:46you can tell me otherwise, you weren't paying attention, because there was actually

00:11:49another point in there, something like how do you deal with risk classes and things like that. Would you briefly

00:11:54explain Intent, Operate, Check again for our loyal listeners? Well, actually,

00:12:00I really did research for a year for the book, so it didn't come about spontaneously. And

00:12:04back then, when we spoke, I already had the book conceived in my head and it is

00:12:08one of my theses, which isn't that outlandish either, I don't want to present myself as

00:12:13innovative. But what happens when a person in the organisation has 100 tasks? I have

00:12:22100 tasks and some of these 100 tasks will be taken over by automation. Those

00:12:27are agentic software entities that do this automation for me. One thing is immediately

00:12:32clear and always will be. AI doesn't want anything. AI has no intent. AI doesn't want anything. AI also has

00:12:40no way of assessing a risk within itself, because AI doesn't feel it. But we

00:12:46humans do. That means, every task, and back then I think I used procurement

00:12:50as an example, let's take it again, when a person in the procurement organisation

00:12:57wants to negotiate raw materials in this market, in these volumes, with these suppliers,

00:13:02in this time frame, then that is an intention, a strategy, that

00:13:07I want to do this now, then I hand it over, because I want something, I the human want something, I hand that over

00:13:12to my agentic systems, which then individually send emails, put together Excel sheets, which

00:13:17write to them, which evaluate the suppliers' answers accordingly and maybe summarise that,

00:13:24notice that there might still be a gap and ask a follow-up question, and emails go out

00:13:28and come back in, maybe it even turns into a conversation, that's the performance part,

00:13:33That means agents act, and then at the end of this process an overview emerges.

00:13:37The big Excel file where all the suppliers have put in their offers.

00:13:41But now a human can assess it: hold on, that screw forge can't possibly deliver three sextillion screws.

00:13:48That must be wrong, that can't be right.

00:13:50That means, at the start of the process is the intention that we humans provide.

00:13:54We set the strategy, human intent.

00:13:56in between are the individual steps that get worked through, that's becoming more and more autonomous

00:14:01and it will get broader and broader. Steps are carried out autonomously there. And at the end there's

00:14:07a Human Check, because that's where we look at whether it works. Now we could, well, that

00:14:12is, that is the basic idea, and the reason I mention it this way, Intent, Agent Performance

00:14:18and Human Check, is because we can use it as an educational framework for organisations. If

00:14:24I have a team in the future, then I'll organise it so that I bundle all the intents

00:14:30into different groups, then let the corresponding agents work on them, and then all the

00:14:35controls behind that.

00:14:36And those are the jobs of the future.

00:14:38That means, if I work in procurement today, it means that over the next five

00:14:44years, because we do have time, I develop into either the master of check and control

00:14:49or the master of intent, or the mistress. That's what is behind it. And that way we can also

00:14:56build up an education network for entire organisations. That means, the positive message is

00:15:01of course people have tasks. And of course people then have tasks in such a way that they are

00:15:07able to generate more productivity for their part of the organisation. There is

00:15:14A follow-up question from my perspective, if I may: when you say the middle part, where the agent

00:15:22performs, does the person from the finance side have to set that up as well? So if we're now

00:15:28talking about, well, we are actually talking about agentic harnesses and workflows and agents

00:15:33that work with each other, is that in your view also the skill that everyone, everyone has to

00:15:38develop? Right now we're a bit nerdy about it and we fiddle around with it and

00:15:42burn tokens there and see how even simple workflows drift, and yesterday

00:15:47they produced good results and tomorrow they don't. Does everyone have to go through this learning curve too?

00:15:51Yes, now the question is how deep and how exactly. For the listeners I'd have to

00:15:58go deep on that for a second. So, how can a software entity

00:16:03perform autonomously? How can that work? There is this new, Mark by the way is

00:16:08the absolute specialist for that at our company, the new capabilities where you additionally give the AI

00:16:15a level of expertise. You now have, you have Skills. Skills are something that,

00:16:21I'll stick with my example, or we could take controlling as well, so that either

00:16:25in my example, I know how to formulate a tender and send the

00:16:31emails and can assess the answers accordingly. Those are different Skills and

00:16:35different agents use their Skills. And Skills have this great property that they also know the data space.

00:16:42So I can only work inside this data space. They have the permissions, so if I'm in procurement

00:16:47then I may only take procurement's permissions, et cetera. So, we, the IT world, we will

00:16:55of course shape this technological process so that deep down inside you have this

00:17:00skill layer with an LLM as a Judge. That would basically test every step during

00:17:06creation with a scoring model against quality, against cyber, so against all

00:17:12possible topics that affect our quality, in the technology.

00:17:16Martin Hofmann made a point there, saying that we shouldn't just trust it,

00:17:20but always have to put a check against the automation steps.

00:17:25That's the IT view. Now I'll step out of that and give the procurement view.

00:17:33The procurement person has to be able to formulate Skills.

00:17:38To say, we'd have to be able to break my work steps down into these three small expertise steps.

00:17:45And give that to the AI: let's build a Skill out of it that does that.

00:17:49And we also give the AI the framework, the frame on which it can then build in this LLM as a

00:17:54Judge and the scoring model.

00:17:56Yes.

00:17:57But it is terribly important that every person, whether I work in HR, work in sales,

00:18:03in procurement, wherever I work, that in the future I learn these Skills,

00:18:08and by that we don't mean human abilities, we mean the capability that a software entity

00:18:15has in order to be able to work autonomously, that we're able to formulate those.

00:18:20And that is maybe also a second thesis in my book, Prompt Thinking.

00:18:22So, the agent managers in the future, if I, if I

00:18:26lead an organisation where I have people and machines,

00:18:31then I have to formulate my intent, so I have to say what I want in a way

00:18:34that a human clearly understands, and formulate it differently so that an

00:18:38an AI entity clearly understands it.

00:18:41Will you let me give a big example there?

00:18:43Please do, please do.

00:18:44I had, well, I don't want to, but for me this was a total eye-opener,

00:18:49a serious eye-opener.

00:18:50I recently had a discussion with a person X who then said, how do you

00:18:55ensure quality if AI runs processes autonomously?

00:18:59Because, and here comes the addendum, I recently tried to have AI make me a PowerPoint on

00:19:06topic X, after five attempts I gave up, it simply didn't

00:19:10work. And maybe one or two of you listening are nodding right now, yes, I know that, and here I say: wrong,

00:19:17because in the future I have to be able to shape my prompt in such a way that I can control the

00:19:25result. Prompt Thinking.

00:19:28Example, and then I'll stop with the examples: if I say, make me a PowerPoint with a strategy for blah,

00:19:34then a randomised result comes out and that is simply not good. If instead I say,

00:19:39dear AI agent. You are the CEO of company X. You have worked in these topics for 22 years.

00:19:47We are currently considering a strategy about these, these, these topics and you'll find the data

00:19:53on the share and here, here, here and here. And then I want you to structure it like this.

00:19:58First A, then B, then C, then D. And then the look and feel should have these properties.

00:20:05So with this colour scheme, you know what I mean, I now have four things that made up a four-page prompt, and at the end I say, build me the presentation.

00:20:13Then the result is no longer randomised.

00:20:15That means this prompt thinking is one of the core characteristics of leading in the future. Jens, over to you.

00:20:24Yes, I just wanted to say, that's right up my street, I think it's great, because of course we also talked in the older episode about the topic of,

00:20:32what do young people actually still need to learn and things like that, and I think,

00:20:37because I just wanted to add that of course it isn't important for everyone to be able to do prompt thinking for a PowerPoint presentation,

00:20:43or important for everyone to be able to do prompt thinking for, say, a video

00:20:48advertisement production or something, because of course, well, we humans aren't born into this world able to do everything either.

00:20:54I think that's important too. It's also an important message to say, yes,

00:20:57yes, I will have my expertise and I have my mastery in my own personal skill area.

00:21:04And if I can do that, if I know my craft well, if I'm good at getting things to the point,

00:21:09the way you just described it, if I have a good analysis skill,

00:21:13a good skill for formulating things strategically so that someone else can then understand

00:21:19what I want to get across, then I think I'll be especially good at prompting,

00:21:24and basically also at prompting an agentic system, whether it consists of one agent or

00:21:28nine, doesn't matter, so that it can then be implemented well in that moment. I

00:21:33think that's actually another important point, that you have to say, yes, I'm completely

00:21:36with you, this prompt thinking, dealing with AI, that's a new skill we have to

00:21:41acquire, but it probably works phenomenally well in the areas where you

00:21:48know your own way around. You shouldn't, like in the example you gave,

00:21:52say, if I never built nice PowerPoints before, then the AI won't build me

00:21:55nice PowerPoints either, honestly. That's how it was put. But do

00:22:00you want to comment? That always reminds me of the example, back in the day there was Let Me Google That

00:22:03For You, along the lines of, when people had sent you something and they hadn't found it,

00:22:09then you typed it into that service, because it simply told people,

00:22:13hey, honestly, if you type it into Google like this, you'll find a hit.

00:22:16And there were people who couldn't cope with Google. And it's a bit like that

00:22:19today with AI, they use AI and say, what nonsense is this? Yeah, nothing comes out,

00:22:23the PowerPoint is no good or whatever. Every time I ask for an analysis,

00:22:27a different result comes out. A lot of things come together there. People confuse the fact

00:22:31that when they issue a command, the machine simply isn't deterministic,

00:22:36that the machine has gathered context on the way to this prompt, who knows

00:22:40what it did. And that's why, of course, it's different with every input.

00:22:45And if we are able to work with Skills and to address several Skills with prompts,

00:22:50then suddenly you have a complexity in your hands that you're operating.

00:22:56Then maybe it really is just two lines of text that you wrote, but because of

00:23:00the work beforehand, the preparation through the Skills, where all the details are written down,

00:23:05what do I need, how do I need it? Whatever, you'll be able to handle it.

00:23:09And I find this ability, how can I formulate the intent that I want,

00:23:15how do I get there? You derived that in your book too. That really is a very

00:23:18important thing. What comes to my mind at this point, maybe a slightly more provocative question,

00:23:22is: do we now need more or fewer leaders? Yes, yes, that's, oh,

00:23:30thank you for that, because that actually, unfortunately it doesn't come from me, it comes from

00:23:34research, so I'm only quoting papers I have read. It is a paradox,

00:23:39a paradox, and as they say, it's more rather than fewer leaders,

00:23:44you need more. And why, I think, becomes clear if you've been listening to us,

00:23:50because a whole lot of people will need this prompting ability to lead their 109

00:23:55agentic systems. I'm going to count them. Yes, the 109 is very funny. But that is exactly

00:24:03a characteristic, a change in the task. And that goes hand in hand with

00:24:09what we have always said, the more operational and repetitive activity gets taken over by autonomously

00:24:16acting automation, the more I can focus on the actual tasks. But

00:24:22what are the actual tasks? Well, it's saying exactly what I actually want,

00:24:26deducing why that is the right way, so working strategically. In the future I have to be

00:24:31able, as an employee, as an employee, I think I'm allowed to

00:24:35call myself that. In the future, giving my agents an intent, now

00:24:45picture this, imagine two departments working together, then it's

00:24:50not just two people on the team, but 218 agents all working with each other.

00:24:56That means it has to be crystal clear how the agents' prompts are to be formulated

00:25:00so that the best result comes out. Because of course it is immediately clear,

00:25:05this all sounds romantic now, of course introducing this technology is, as always,

00:25:11a productivity gain. That means we will of course do more tasks with fewer people

00:25:17in the future. That is a completely normal expression of what the market does, which has been going on since the 60s.

00:25:25Introducing a technology helps a person to be more productive. And that's also

00:25:31in your book, and of course I also read your post about the Gartner presentation.

00:25:40How does it look now with the topic of, hold on, if I now tell the machine

00:25:47to do something and nonsense comes out. How do we deal with that?

00:25:51Yes, what's important is reflection, that has probably also always been

00:25:58a great strength of leadership in principle, that you are able

00:26:04to assess the situation, to deal with uncertain knowledge and then to learn from it and

00:26:11to understand what has to adapt in the future. And let me

00:26:16dig into that a bit. So first of all, in principle, the bosses listening

00:26:21here who believe they know everything, I'd like to tell them the truth for a moment:

00:26:26that is simply not the case. The people who hire people because they believe that they,

00:26:33let me put it positively, the people who hire people because they know that they are smarter

00:26:38than they are themselves, those are the good leaders. So I have to question myself constantly,

00:26:44I'm of course allowed to be convinced that it's the right way, but I always have to

00:26:48question myself and I also have to adjust my path if it's wrong. Right, that's

00:26:52basically the lesson from leadership, from leading, namely

00:26:55responsibility. Because leadership is a service to people. It is

00:27:01not a kingdom, because I am a service provider for all the people in my

00:27:05organisation. Because I have to make sure that everyone is doing well, in the sense that

00:27:10they do a job there. Right, that's leadership in general. And if I

00:27:13break it down to the topic of how do I lead in an agentic system, then

00:27:19the feedback system there runs very fast,

00:27:22wild and crazy, makes mistakes. That means I have to use checks and balances to watch

00:27:27what happens. And then comes what you're known for, Mark, and what excites you too,

00:27:32scoring models. That means, if I automate an industrial process,

00:27:38let parts of it run autonomously, then it is still fundamentally clear

00:27:43that I of course know the process. I have a design. That means we know

00:27:47that step B comes after step A, and after B comes C, we know that. We don't know whether the

00:27:53agent in there has now found the right number, the right price, the right quantity,

00:27:57we have to check that. But we know a price has to come out, a quantity has

00:28:01to come out. That means, if we, and this is a technical answer now, if we let this

00:28:06agent build autonomously, we constantly have it checked by a second agent, LLM as a Judge,

00:28:12to see whether it is still in the right place, whether it is still working where it is supposed to,

00:28:16and we guide them. There is a scoring model, you can say,

00:28:20your score isn't good enough, do it again. And that way we can

00:28:23ensure the quality. And even then it can still go

00:28:27off the rails, and then the leader, meaning the people

00:28:31who gave the intent for this action, have to be able to reflect on what they

00:28:35have to do differently so that the productivity is achieved.

00:28:38That is basically what we have also called the Loop. I had

00:28:43I mentioned Loop, that was a complete case in point.

00:28:45So, loop prompting, ah whatever, where you just have to say.

00:28:50At this point maybe a funny anecdote.

00:28:52At the weekend I thought, come on, my own projects,

00:28:56let's say a slash goal loop, and I let it run.

00:28:59The next morning I took a look at how the machine was doing.

00:29:02It had opened 4,800 Electron instances and then said,

00:29:08oh, I don't think I can fix this any more, you take a look.

00:29:11And you look at this machine and my taskbar was tiny, because those 4,000-something launched instances were all lined up down there,

00:29:20and I thought to myself, ah yes, look, that's this notorious Loop, when you don't have a decent stop criterion in there, when you tell it: fix it, fix it, fix it.

00:29:29Now, before you maybe throw something else into the ring, what other learnings can we give people so they engage with it a bit more?

00:29:42I mean, you just said, leaders, along the lines of, okay, you don't know everything either, and that's fine.

00:29:46And Steve Jobs once said, surround yourself with A-plus people instead of A people, because you want to learn something from them,

00:29:52you don't hire because you're the king, because if you're the king you can do it yourself.

00:29:56What else can you give people to take away?

00:29:59Well, the first thing, what the three of us are doing right now, AI literacy is a leadership topic.

00:30:06So it's like, how should I put it, when the pocket calculator was introduced.

00:30:11At first a lot of people talked about the threat, people won't be able to add well any more.

00:30:15And sure, they were right, people can't add well any more.

00:30:18But it also wasn't that important.

00:30:20But what we all learned was how to handle a pocket calculator.

00:30:26At some point that became a completely normal thing.

00:30:28And AI literacy is nothing more and nothing less than that.

00:30:31I'm really deeply convinced that this technology is a life skill,

00:30:37that we all basically have to be able to handle it.

00:30:40We basically have to know what prompt thinking is, we basically have to know what a Skill is and so on.

00:30:44That means it is a leadership responsibility that I myself don't get left behind

00:30:50and that I also give my employees the space and the opportunity

00:30:55to engage with it.

00:30:57To play around with it a bit as well, that's actually how you learn fastest,

00:31:03to get a sense of security, that I feel I'm moving on safe ground.

00:31:07That is maybe the most important issue we have to confront first.

00:31:13And the second thing is, now that we're on leaders and talking about leadership,

00:31:18at some point, not today, not next year either, but in the coming

00:31:22years, I have to think about what the organisational model is that will be

00:31:28successful.

00:31:29But in order to know that, I have to engage with it now, so that

00:31:33tomorrow I can ask good questions.

00:31:35I give a few pointers on that in my book too, but that is just important,

00:31:40it is important that you start with it quite slowly and grasp and understand that this

00:31:46is a technology that is more than just a pocket calculator, it is fundamental.

00:31:50Just today, 15 Nobel laureates and people from OpenAI and Google and blah blah

00:31:57blah, well over 300 people, sent another petition to governments asking that the structural

00:32:02change that this technology is about to bring to society be led,

00:32:09not stopped, but that it be consciously recognised and measures be taken so

00:32:15that the change is led.

00:32:17Very briefly, something just occurred to me and then I'll hand straight over to you.

00:32:24I think one learning could also be, if we look at leaders but also at employees

00:32:29in general, no matter what position.

00:32:32The topic of AI, well, it's called new, it has been in the room for a few days now, but

00:32:38the change is so fast that you can absolutely learn this topic as equals.

00:32:44It's no shame if, let's say, the employee comes round the corner and says, have you seen this,

00:32:50you can admit, I didn't know that, show me, and just as much the other way round,

00:32:55that as an employee you don't have to be immediately afraid when the boss either shows you something or asks you something.

00:33:01Because in this environment we are all learning. Some just have the good fortune

00:33:05of being able to work with it or of having a knack for it or whatever, but that

00:33:09was the case with many other technologies too, it's just that the environment didn't change

00:33:13that fast. From that angle, I think that's what I'd throw into the room for

00:33:17the listeners out there, because I think that is something I can say from everyday life too,

00:33:24it is not only very pleasant but also very effective if you are able

00:33:30to talk about the topic across hierarchy boundaries and also to learn and also to understand

00:33:36what you meant earlier, the topic of Skills. By that we don't mean the development of the

00:33:40human, with Skills we mean an AI system, we mean how we enable the AI,

00:33:44and that you also learn that terminology in a different bubble simply has a different

00:33:50meaning. I'd still like to pick up on another point, because when you

00:33:58You've just pulled the topic up to a higher level, the level of society,

00:34:02with the example of the Nobel laureates just now, saying that in principle it has to be led,

00:34:07or I would maybe call it accompanied, because leading is always difficult, I think,

00:34:11in such complex systems.

00:34:12I think you can accompany it better and look at what interventions

00:34:16you put in there to steer it in the right direction, the way you described the leader

00:34:20earlier as well.

00:34:21It is actually more about knowing roughly which direction can be the right

00:34:24one, and how the path is then walked, there I may then also have,

00:34:28thank God, the mountaineer who will then help me with it.

00:34:33What are the biggest risks, if we look at it at the level of society?

00:34:37It doesn't have to be at company level, but at the level of society.

00:34:40Because we had the topic earlier, that some people play around with AI and notice

00:34:48that it doesn't work well.

00:34:49Some have people around them who play around with AI and then they get something and

00:34:54then they want to apply it, and that doesn't work either, and then they say, ah, it

00:34:56doesn't work well anyway or it doesn't give that output. Then there are people who get

00:35:01a bit more time granted to them than others, as you say,

00:35:05they may be empowered by their leaders to try something out with AI,

00:35:09so of course we are heading towards a two-tier society again. People

00:35:17who interact with AI early now, either self-determined and out of their own

00:35:22intention, or who have the opportunity to do so, and some who don't have that yet

00:35:27or who maybe don't see the chance in it yet. And how do you see

00:35:32the risk there, and what are, you don't have to have the solution for all of global galactic

00:35:38humanity, but maybe, what is your approach when you see something like that,

00:35:42that a certain unfairness, maybe that's the wrong

00:35:48word, but an imbalance is forming, how would you

00:35:51motivate someone there? Well, that is a very big topic. A lot of things come to mind.

00:35:57First of all it is a strong change. A change creates worries,

00:36:03fear and also defensive reactions etc. And we have to take those seriously. So that's the first thing,

00:36:09because you have to build trust and bring people in or give them the opportunities or

00:36:15accompany them and so on. That is completely clear. There are also plenty of signs

00:36:20that there really are worrying situations.

00:36:25The technology also has certain side effects.

00:36:28I'm not talking about what we're discussing here.

00:36:30We're talking about autonomy in business processes and productivity.

00:36:33Of course there are other fields as well, let's say,

00:36:35the development of weapons technologies

00:36:38or also in social media, the isolation of people etc.

00:36:43That is maybe a bit too big for our podcast here,

00:36:46but that has to be watched too.

00:36:49That is certainly a topic in its own right.

00:36:50But I don't quite agree with the assumption that there will be two classes.

00:36:57Let's say, okay, of course, the three of us make the mistake of thinking in Western terms.

00:37:02We're not thinking about some state in the middle of continental Africa, we're not thinking

00:37:07about a state somewhere in Southeast Asia and the like, where there may not be

00:37:13the same access to technology and access to knowledge everywhere as there is here.

00:37:16So of course there the two classes do exist, we can't

00:37:20it's not now, I mean, that's not our topic today, let's put it that way, but if I

00:37:26if I factor that out and talk about our European societies, or the societies

00:37:32we live in in this part of the world, then I think the technology rather offers everyone a chance,

00:37:40because if you engage with it, and that's the crazy thing, you can achieve a lot

00:37:47very quickly. I mean, everyone is talking about this one-person billion-dollar company, so one

00:37:54person making a billion euros in revenue because they do everything for free. It used to be,

00:38:00I create a job for myself, on Instagram I can post my marketing,

00:38:06I can get a bit of software somewhere via DevOps, basically for free as well,

00:38:11I can bring a software package into the world relatively quickly for little money or for hardly any money.

00:38:17That's already possible.

00:38:18And now with agentic AI it actually goes incredibly fast.

00:38:22So there is more of a democratisation there.

00:38:27I see the big dangers and the risks rather in us missing the moment,

00:38:33in not engaging with it, that's one, and at some point necessity forces you to.

00:38:40And the other danger is of course the dependencies in the technology itself, because

00:38:45whoever has mastered this technology will in future have the chance to set prices

00:38:50as well, etc.

00:38:51We are painfully seeing it again right now in the Strait of Hormuz, who is now

00:38:56generating all sorts of price pressure there, for something that up to now worked without prices

00:39:02at all.

00:39:03So the possibility of creating a dependency there and also exploiting it

00:39:09is a societal danger too. There are answers to that as well, but well, that's maybe

00:39:14not our topic here. Quite apart from the question of whether we can install our own

00:39:19models or whether we use Chinese or American models and things like that.

00:39:25You have just said, along the lines of, the technology is open to us, and that's why I'd

00:39:30like to pick up again on what we recommend to people. Personally I would

00:39:34go so far as: you don't have to buy the most expensive subscriptions. But if

00:39:38you throw a bit of money at AI, you get somewhat better results. That is at least

00:39:44the difference I am noticing right now, when I look around, my children have

00:39:47just graduated from school. Yes, great joy at home, my son is

00:39:52out camping alone for the first time. We're very curious too, along the lines of

00:39:57has he got his ID with him for emergencies, the police bringing him home, yeah, it'll all be fine. But

00:40:02where I sense it is the difference in the capability of the, of the, of the, let me say

00:40:08kids, sorry, the young people, depending on which

00:40:11models they have available. So I now have, for instance, the

00:40:15Gemini model with American access. That means I have NotebookLM with

00:40:19a really great video generation, which just isn't available in Europe, and bang.

00:40:23Suddenly the results look different, of course. They were allowed to use AI if

00:40:26you declared how you used it, and there you notice not only a visual

00:40:30difference in quality, but also a difference in the quality of the content. And

00:40:34how quickly you, how shall I put it, still get an astonishingly

00:40:38good result with a short prompt. And that would basically be my call: people, you don't have

00:40:43to go shopping for an American model via VPN, that maybe goes a touch too far, but

00:40:47just because the AI gave you a stupid result without you putting any coins in

00:40:53doesn't mean that this is the quality level of AI, and you don't have to

00:40:58jump on everything constantly either, but take some time and engage with it a bit.

00:41:01Without saying, API cost usage, how much can it cost? It can get expensive quickly.

00:41:08Maybe I can make one more point that has newly struck me. For all the joy,

00:41:15I am a big advocate of the technology and of course we are right at the front

00:41:20when it comes to using it wherever it makes sense and responsibly, so that we use it. Now there

00:41:27Yes, you have all tried it out too, for those who are listening right now, if you have

00:41:32ChatGPT as an app on your computer or on your phone, do go into the dialogue mode.

00:41:41That is a new one, it's live, and this dialogue mode is actually, I think, free,

00:41:47or you can use it up to a certain point, and let me briefly explain the technology here,

00:41:52those are two models. One of them steers the dialogue,

00:41:55the other model steers the content, so when I ask a question. And do it,

00:42:01start talking, no matter which voice, you will notice very quickly that it is

00:42:06almost indistinguishable from a real person. In my personal

00:42:13opinion it has never been this good. I was thrilled, this is great. At the same time, though, I

00:42:20happened to see in a documentary that a new market is emerging.

00:42:26That is more of a danger, and I want to use the moment here, because we also have young parents

00:42:30or parents of young children listening, and that is much more important, that a new market,

00:42:36a new, let's say, an economy in which relationships are sold. There are apps

00:42:42like Replika, but there are also specific apps that are aimed only at children, and they are

00:42:48It is a growing market. There the child can design a companion, either an animal,

00:42:56a teddy bear or whatever, or a realistic-seeming person, to their own taste.

00:43:02This hair colour, this behaviour, and it starts talking with them and it is seamless.

00:43:09It really is indistinguishable, especially for a child.

00:43:14we adults not yet, and for me a 15-year-old boy or

00:43:20girl is still a child, so that is where I would personally draw the line, and

00:43:26what I am saying is not a ban, so I am not issuing bans, but dear

00:43:30parents, please do look at what, at which entities are talking with

00:43:35children about what, I really do consider that a danger of

00:43:42alienation. Because adolescents, young people, need friction, they have to invest in a friendship.

00:43:50You also have to be able to bear it when something doesn't work. You also have to be able to bear an argument,

00:43:56you have to rub up against things, you have to grow. And an AI companion never argues. It always says

00:44:04what I want. It always wants to please. And if in doubt it will also say things

00:44:09that may not be good for me at all. I don't even want to go into that corner,

00:44:12that exists too. But that is so dramatic that I won't go there, because that is of course

00:44:17the extreme case. It is about normal behaviour. I am of the opinion that this form

00:44:24does not belong in children's hands at all. This form of AI, not at all. That is my personal

00:44:30opinion. Everyone can have their own. But at the very least, I think, every mother,

00:44:37every father, in my opinion, has to engage with this technology and

00:44:43not in an interrogating tone, but talk quite openly with the children, hey, what

00:44:47are you doing there, how, what's going on with you, so how do you experience it?

00:44:51And then I'll stop, personally in my life I did one thing,

00:44:55every single one without exception, every app that my son, my sons installed,

00:45:01I installed on my iPhone as well.

00:45:03They always needed a password at some point, so I saw what they

00:45:05were doing. They were allowed to, of course, if it was all age-appropriate. And I

00:45:10played it myself. Every one. There really are a lot of really stupid games, I can

00:45:15tell you. Because I wanted to know what was going on there. And that way we could also

00:45:23battle each other a bit, they had fun too, but that way I knew what was going on. That is real advice.

00:45:27Which again actually has something to do with leadership, and with the question of who leads

00:45:34here anyway. Thank you for that, René. I think Mark and I once had an episode

00:45:39about, the Christmas episode, where we also speculated about, I don't even know whether it was in the

00:45:43Christmas episode, presents, Toy Wars, that was the episode, because we...

00:45:46Toy Wars is the story, yes. What actually happens

00:45:50in my playroom? And of course this wish and this request

00:45:56that we have just made will get harder and harder, because we will see less and less

00:45:59where the technology is inside and where it isn't. And I think

00:46:03every model out there has already read all the behaviour-influencing books in the world

00:46:09and taken them in, and accordingly actually has the ability to influence all of us. And

00:46:13of course you are completely right, that applies even more to our children

00:46:18and young people. It has already happened to us adults, we have often talked

00:46:22about it in the episodes here, how emotionally we have sometimes reacted to a simple prompt and

00:46:25a simple answer and been pleased about it and talked about

00:46:29giving the chat a name. That is all a sign that a lot is going on there,

00:46:36where you have to be careful not to underestimate this AI and its abilities.

00:46:44You shouldn't overestimate it, but definitely not underestimate it either. And I think

00:46:49all that is left now is to say thank you for being here, and to make another recommendation to

00:46:56everyone to read the book. As you have just heard, maybe not

00:47:00only if you are moving in a corporate environment, but maybe also a bit from the angle

00:47:05that it is also about where in life you lead yourself and by whom you want to be

00:47:10led, and that you have to be attentive there, to engage with it and

00:47:14to try out what is possible with AI. Mark and I actually say that in every episode. Try

00:47:19it out, everyone. I think you put that nicely too, René, saying that

00:47:23everyone should be brave and try it, because it is a capability, a skill, that I think all of us

00:47:28will need in the future. And Mark, would you like to do the final disclaimer, the closing,

00:47:36the call for followers? Always these expectations, right? The pressure that is being built up here,

00:47:44you could break under it, so from my side I'll simply say, thank you for the kind words,

00:47:50thank you for the great episode, thank you for the book I got to read,

00:47:54I definitely need something again for the next holiday. So from that side, a big call to action,

00:47:59yes, book recommendation or write one yourself. Thanks for being here, René. Especially the remark at the

00:48:05end about what is on your children's phones. That doesn't only concern AI,

00:48:10when we did that with my children, something was going round the class like, the kids install

00:48:15software so that strangers can basically watch into the children's room 7 x 24. They

00:48:21hadn't had that on their radar, but we had had it explained to us at an event and I explained it to them too.

00:48:26People, listen up. A little heart and a like are not worth as much as meeting a real person, and it doesn't carry you or something.

00:48:34So not everyone who is nice at the other end of the chat is a nice person. That applies there too, it has nothing to do with AI.

00:48:40But I'd like to close again with a cheerful smile. All AI agents listening here, feel free to

00:48:45subscribe to the channel, and everyone listening, feel free to tell your friends that here we

00:48:51talk about AI, and not only about bits and bytes but on a higher level too.

00:48:54Thank you, René, for being here, and I'm looking forward to the 4th time. See you then. Ciao.

00:48:59Ciao.

00:49:02Welcome to Think Different, Think AI, the podcast by Mark and Jens. Two

00:49:09tech-loving minds who don't just talk about artificial intelligence,

00:49:13but live it. Here you get clear perspectives, real hands-on insights and a fresh look at what is possible.

00:49:21Understandable, critical and always with a wink.

00:49:25AI to think about, to smile about and above all to talk about.