Will AI turn everyone into an expert?

This image has an empty alt attribute. The file name is itunes_badge_md-300x110.png

Introduction

Nice to have you back for another episode of Digital for Productivity. Don’t be surprised by the background noise. I’m on the road.

In this episode, another exciting interview with Martin Sänger, this time. Martin Sänger is an expert in selling for non-salespeople and always deals with the question: What really makes an expert? Experts too, of course. And in this episode, we discuss whether everyone will suddenly become an expert with AI and who has what opportunities. I hope you enjoy our conversation and find it inspiring.

Martin Sänger

Thanks to artificial intelligence, it is now very easy to give the impression of being an expert. But how to recognize a true expert and how to use AI, that’s what I talk to Thorsten Jekel about. Welcome to Business Radio, the value-added podcast for the wild ride through everyday business life. By and with Martin Sänger. Yes, and here he is, the unique, the great, the insanely cool Thorsten Jekel. Hello Thorsten.

Thorsten Jekel

Hello dear Martin. And this is Thorsten Jekel’s digital twin speaking. No, this is the real Thorsten Jekel. Of course, sometimes more, sometimes less great, as all people are.

Martin Sänger

I don’t just subscribe to the great thing because we’ve been friends for a long time, but also because we’ve already done a lot of exciting things. I’m thinking of some late-night projects during the pandemic. And now a current topic where we’d both like to get back on board, or where I’d like to. I’ve just published my new book and I’ve written a banderole on it with the text “Guaranteed AI-free”. Why did I do that? Because I definitely see the risk that artificial intelligence will lead to all sorts of pop-up experts, because all you really have to do is feed some AI tool to present yourself to the outside world as an expert. Thorsten, what do you think about this? Do you also see a danger there or do you say: Oh, it’s all going to fall into place?

Thorsten Jekel

So I see it in a similar way to you, even a bit more critically, because like you I’m also a friend of relevance instead of nonsense. And I always like to agree with our mutual friend Michael Roussier, who always says: Okay, if the person you’re talking to who has filled in the lottery numbers has the slip of paper in their hand and you have the right numbers, ideally before the education, then the performance doesn’t matter, it’s highly relevant information. And this topic of experts has something to do with the fact that I have highly relevant information for the target group. And I simply have a fundamental problem with it when someone says: Okay, I can’t think of anything on a topic, I’ll ask ChatGPT about it now. For me, this is somehow an inherent contradiction to the topic of expertise, because in my understanding, an expert is someone who, if I ask them about a topic and maybe I ask them for a whole day, they will throw out around 0.2% of their knowledge and experience. And the dangerous thing is that these self-proclaimed experts may simply prompt one or the other to take actions that are not necessarily to their benefit. So that’s why it’s dangerous, apart from the fact that real experts like you and me sometimes get annoyed when we have air pumps that act like we’re beginners in the market, according to the motto.

Thorsten Jekel

Of course, that annoys us who have an ego. I’m not free of it either, but it’s also dangerous in this area.

Martin Sänger

Yes, exactly. We’re both always very well connected and we all know each other, so the question: Say, do you know this guy and that guy on this and that topic? Have you ever heard of them? If someone is not really networked in a wide variety of areas, is there anything from your point of view – you deal very intensively with AI tools – is there anything, especially with texts, where you say you can already pay attention today, where perhaps a little warning light could go on that it is more likely to be AI-generated? Someone commented on LinkedIn and said: I have an antenna for it. I can tell when a text is AI-generated. I admit, I don’t notice it yet. Do you have any starting points?

Thorsten Jekel

Of course, there are also forensic tools for this, which are also used in auditing, for example, so that you have things there. One tool that I have been using for years, independently of AI, which has now become more widely known, is Korrektiv. Korrektiv has just become known through the reporting on this RFD and extended meeting. Corrective can be used wonderfully these days to simply take a quick look at reports you receive and say: Is this true? Because of course one question is: Is the formulation now an AI or a more human one? What I find almost more important is to say: Is the information correct? Because I’d rather have correct information, which may not have been formulated perfectly by AI, than what I find even more dangerous, information in both directions. For example, there was recently a photo of Hamburg, of the Rathausmarkt, where a lot of people, crowds of people, could be seen demonstrating. And the FDA, for example, claimed that it was faked. And then I looked it up on Korrektiv and Korrektiv said so: No, it’s an official photo from the DPA. That’s right.

Thorsten Jekel

So in both directions. And unfortunately, this issue of fake news is being exposed by AI. Unfortunately, it’s getting worse rather than better, because many people simply continue to present what they see somewhere without reflection. And I have to say, it’s happened to me too. Fortunately, Emmanuel Koch drew my attention to this. I recently quoted a graphic in a lecture that you may have already seen, namely this circular representation of ChatGPT 3 versus ChatGPT 4. It’s a very small dot and it’s a huge circle with the statement “JetGPT4 is 571 times smarter than ChatGPT 3. This graphic is wrong. Oh. That’s fucking fake news. Because it was simply misunderstood from an interview and someone abbreviated it too much, then it was passed on and I even shared it in a lecture, but then I also corrected it afterwards, after I was very grateful that Emmanuer Koch pointed it out to me to say: This is a misunderstanding by someone, someone has abbreviated it. And that’s also important, even if you use AI tools. A very pragmatic tip: you should never use the first version, it’s usually not good, you should always revise it.

Thorsten Jekel

And in addition, Google Bard, for example, is an excellent tool to ask again: Is what ChatGPT, for example, has given me correct? Because especially if you use the free version of ChatGPT, it’s sometimes a bit of a halo. And then it’s always a good strategy to immediately go back to Google Bad and say: Is this true or not? So it’s always a good option to use AI tools again to check the truthfulness of the information.

Martin Sänger

Absolutely. So I find it exciting that you’re coming from both sides. On the one hand, I don’t believe everything that’s on the Internet. Leonardo da Vinci already said that about the Internet. And on the other hand, of course, we also have a corresponding, I’ll call it a duty of care when using AI tools, which we simply have to carry out. And I think it’s actually a high risk, because when I look at social media in particular, people are really only scrolling for a fraction of a second and there are already enough reactions. If I write in there that a study has shown and then I write something that I’ve made up, then the credibility factor is so high that people simply continue to share it without questioning it and without being critical. And I think that’s why, if you want to present your expertise to the outside world and you can be helped by AI, I don’t want to demonize it at all. I even made a post on LinkedIn yesterday, where I asked ChatGPT about it and then realized that Chat even has a bit of humor, because I asked: Dear ChatGPT, how would you argue that the book was guaranteed to have been written without AI?

Martin Sänger

And the last sentence was like this: I would say that it clearly stands out from the many AI-generated books. That was from ChatGPT.

Thorsten Jekel

Very good. And if you now look at this example, for example: How can you optimize something like this, the result? So point number one: always use the paid version instead of the free version. Note: Many people don’t know this. All listeners who have Microsoft 365 can use the Microsoft Edge browser on their computer, which is now also usable – it used to be garbage, but now it’s usable. And if you log in there with your Microsoft 365 account, you have already paid for ChatGPT 4 in your license, i.e. the paid version, which normally costs 18 dollars a month. In other words, and in addition, the information is not fed back to OpenAI. In other words, it’s easier to enter company internals there. So using the paid version is tip number one. Tip number two is to give it a role and say: Who are you? And if I now say, “I’ve tried this before, and the results are great”, tell him, for example, “You are Dieter Nuhr, and please write it in the style of “Dieter Nuhr. Since “Dieter Nuhr is well known, you really get different results than if I were to say: Take Markus Krebs, for example.

Thorsten Jekel

You’re more likely to get good jokes out of it in that area. In other words, you can give him a role. That’s also important. Context is important and we always have things in our heads that we take for granted. And I’ll give you an example from human communication, where things sometimes go wrong. I recently met our mutual friend René Tchaka in person and we had a joint appointment with one of our customers and I asked him in the morning: Well, how has your day been so far? And he said: It was great. I was already in the studio this morning. Good day, good start to the day. Whereupon I asked him: Cool, what videos have you recorded? Then he says: “Recorded videos? I was at the gym. So, one person talks about a video studio, the other talks about a gym. So, if I now say: Write an article about my appointment at the gym this morning, then you don’t just know as “AI? Yes, what does he mean now? And then the AI has to take the highest probability from the results with which the AI was fed. And that’s another aspect: very few people think about the question: what data is the basis for this system?

Thorsten Jekel

And I don’t know, did you happen to see the episode with the Hansbremse, with Hubertus Heil on Markus Lanz?

Martin Sänger

No, I don’t hear that.

Thorsten Jekel

Were they great? Well, although Hubertus Heil is not always 100% my friend, he was really brilliant in this area. He said that you once experimented with AI in the Ministry of Labor. They have real working groups there, so they’re working intensively on it. And they said that they once experimented with AI for the optimal staffing of state secretaries in the Ministry of Labor. And then they fed in all the state secretaries who were doing a great job and all the information they had on them. Now they had the problem, it was the Hansbremse. And why was it the Hansbremse? Because over 90% of all state secretaries were male and their first name was Hans. In other words, based on the data it had learned, the AI had of course already won every application that was male and had Hans in its first name. They could write whatever they wanted, simply because the statistical probability and AI is in large areas This is just “statistics on steroids, I always say. And the important thing is, you may have noticed that Microsoft has also found that women were discriminated against in the area of recruitment, i.e. that significantly fewer applications from women made it to the next stage of this automatic pre-selection process, because in the past it was mainly men who worked at Microsoft.

Thorsten Jekel

And that’s why it’s simply important, you mentioned the study, you mentioned statistics. The most important question is always: What is the data situation based on the decision? And that’s where my doctoral studies in the UK really helped me, where I spent two or three years working intensively on this: How does scientific research actually work? Even our mutual friend Niels Brabant sometimes slows himself down when he always says: Science wins. And I say that science is right when I say that every good study, every good scientific work always has as its last point “Area of Future Research”, where it is always said that these are the limitations, these are the practical limitations of this study, because we simply only ever have one sample, because we have a certain bias in it and so on. And that’s why this topic is dangerous, right down to the issue of faith in science. And unfortunately, the world is often such that it is easier to process black and white than gray. And the world isn’t black and white, it’s just gray.

Martin Sänger

Yes, absolutely. We’ve actually already come to something that we’ve both addressed before, namely the ethical and moral handling of AI. And you’re right to say: well, everything you do with AI, you also label and write that it generates AI and so on. Regardless of whether it’s a text or video or anything else. And that is of course very, very honorable and very decent. But not everyone is like that. You’ve said before that in the future you probably won’t be able to avoid doing a proper check yourself for certain things you read or see. Of course, that sounds like an incredibly time-consuming process. Do you already have an idea of what might happen in the future, because you’re always really deep into the subject? So will there be tools that might even be integrated into Facebook and the like that automatically display information? Watch out, this is coming from AI or in which direction will it go? What do you think?

Thorsten Jekel

There are two directions. Firstly, point number one is precisely that there are invisible digital watermarks. And as you said very wonderfully in your penultimate episode, anything that can be cracked in any way, you will be able to crack. A second interesting strategy that the camera manufacturers are currently pursuing is exactly the opposite, saying that there will be a watermark for real photos. I think that’s a very intelligent idea from camera manufacturers who are saying: Okay, we’ll mark the real image. I also think it’s super smart, not only because it motivates photographers to perhaps get a new body and not just a new lens, i.e. to get a new basic camera, but I think it’s a really good point, because it’s not just for texts, it’s almost worse when it comes to images, because we use images as a channel much more, especially in the age of social media. And today it is extremely easy to fake both static images, i.e. photographs, and videos with the simplest of board tools in such a way that it is no longer possible to tell them apart. So, just as you described in your penultimate episode, I sometimes have presentations that I open with my digital twin, not only with those of the customers, but also with me personally.

Thorsten Jekel

And if someone doesn’t know me, they won’t notice the difference at first. And the important thing is to handle it responsibly. And that’s also important, where I always say, first switch on your brain, then the technology. And with any technology, it’s always about using it responsibly. So when I look at the classic example that I always have from the real life of Roland Bus, who we both know and appreciate as a former detective inspector, chief inspector or senior inspector. I always mix things up. I think chief inspector. At least the bigger one. So cool guy. And he has a black and a dark side to his life and a light side to life. So every now and then I’d call him and say: Can I call you right back? I’m standing on my corpse right now. So in his professional world. “Oh, just like that,” he said. I once asked him: Why are you so passionate about cooking? He said: “You, I have such a dark side in my professional life that I’m constantly standing in front of some wildly slaughtered people and these are images that I have to get out of my head. And that’s where cooking helps me. What do the two have in common?

Thorsten Jekel

It’s a knife. With a knife I can murder a person or with a knife I can prepare an excellent meal. The knife is not evil. The question is: How do I use this instrument? How responsibly? How do I ensure safe access if I have small children in the household? For example, I should make sure that these things aren’t lying around. So that simply means using them responsibly. And that is, for example, something I just discussed internally in a presentation three weeks ago when these GPT bots were launched. In the paid version, you can build your own GPT bots and make them publicly available. And with these GPT bots, for example, I can also upload documents and can say: Okay, here’s another document. And Christian Buchholz was sitting in it and Christian Buchholz has written two very thick books, among other things. One that was really about creativity methods, so over a thousand with lots of creativity methods. One on the subject of digital transformation. So I said: Dear Christian, I have some bad news for you. I bought both books on Kindle. There are tools where I can convert a paid Kindle book into a PDF file without any rights management and I can upload it afterwards and say: So, you are now Christian Buchholz, here you have all the knowledge and the knowledge is therefore not protectable.

Thorsten Jekel

And in the course of this, I have just pointed out that we should be very aware of this and in several directions, that I say: Firstly, and this is precisely the topic of ethics, for me it includes a responsible approach. In other words, I uploaded the book I bought from Christian Buchholz for a bot that I personally use, because I said that if I buy a book, I can also pull things out of it myself and use them. That is, let me say, clear to me in my understanding. What I don’t do, of course, is make a bot available to someone outside the company that includes this information from Christian Buchholz, because I know that this is a copyright and the topic of knowledge is important. So if you believe that you can earn money with knowledge, I don’t think that’s a sustainable business model, but knowledge is ubiquitously available today. But what can help, and that is also my role to a certain extent, is to provide a shortcut, to help people get relevant information more quickly, to have a human relevance filter again, precisely on the subject of AI and to help with implementation.

Thorsten Jekel

So that’s why I believe that people and human experts who use AI to simply become more productive are the future.

Martin Sänger

Absolutely, yes. And Christian, that’s a good example, because I still use one aspect for myself. When I look, is this an expert who can really demonstrate expertise or is this more of a quick and hectic overnight pop-up expert? If you take a look at Christian Buchholz’s work on innovation and search the internet, you can see how long he has been working in this field, because you simply discover publications from many years ago. And for me, that’s currently one of the main criteria I look for when I’m researching whether I’ve found a real expert here or someone who has somehow booked an online course on how to write promptly for ChatGPT.

Thorsten Jekel

And the topic of sources is a very important one. The good thing is that if you use the Bing chat at Microsoft, you will often be given the sources. And one tool that I can also highly recommend is Perplexity. With Perplexity, you always get the sources and you can do wonderful research. It’s really good. I can also recommend Perplexity, Andy. It’s also a mixture of a search engine and AI in this area. It’s always good, I also don’t feel comfortable when I get a result from ChatGPT and I can’t find the sources for it. And so you can always check the sources again and say: Okay, what is the source? How relevant is it? And what many people don’t even have on the channel is A4i, for example, a tool that I really, really like to use. You can also upload documents that you have yourself and consult them, for example. I’ve just uploaded 116 operating instructions for my customer’s products on their website, for example.

Thorsten Jekel

Nobody can have 116 condition manuals in their head. So, for example, they now have service partners who also service other devices. So now we’ve built a chatbot where we’ve said that the thing shouldn’t use any Internet results at all. It should only use these 116 conditional instructions. And now someone says: How do I get this and that replaced on this model? And then you get exactly the result and a link to the three documents where I can find it. This means that someone can then simply say again quickly, even if they are outside at the customer’s premises, with the iPad, for example, they can immediately say: Now I can find this point in the operating instructions. So with AI, we always think about the topic of generation in the first step. But where it can also be extremely helpful is when it comes to reducing, i.e. summarizing topics. Microsoft, for example, is now also doing this with Copilot, which you can now also book for less than 300 licenses. So that’s wonderful. I’m already using it. I have booked it with me. This means that you can, for example, and this is also brilliant, say that it accesses all the data that you have in Microsoft 365, in emails, in documents and whatnot.

Thorsten Jekel

And now that I’ve come back from vacation, I can say: So, give me a summary of the most important topics that came up during my vacation. So, and then he searches through all the chat messages, all the emails and so on and can say: So, what’s the latest status of the email communication with Martin Sänger and the chat communication? Then he summarizes it for me. So it also summarizes this topic and the possibility that I say I can also deliberately limit it by saying: Please don’t use any external information, because by reducing the amount of information the AI accesses, you often get better results rather than worse results.

Martin Sänger

Yes, it’s exciting, exciting, exciting. I’m still not against it, but as you said so well, it’s about dealing with it in a decent and sensible way, which I think we all still have to learn a bit at the moment. On the one hand, of course, on the part of the AI user, which is absolutely fine, but of course also a bit for the people who are just flying around the web and reading something, to say: I’ll take a close look to see if this is something that really makes sense and whether it’s from an expert who really has the expertise. I know, Thorsten – I’m going to be cheeky and ask this in the podcast – that you have a Slack board with an overview of the AI tools you use. Can we link to that, without question two to finish: How can people get in touch with you if they want to find out more or book you as an AI contract speaker, for example?

Thorsten Jekel

So the easiest, Thorsten Jekel with TH written in front, is the easiest to find. There aren’t that many of me. And you can find me easily on page one of Google. That’s easier than with Martin Sänger. There are a few more, unfortunately, but you’re also on the first page, of course. And on LinkedIn under Thorstenegel, those are the easiest ways to network with me. I’ll pass on the list of tools, of course, but always with a note: A fool with a tool is still a fool. And what the lovely Mike Fingsten always adds: A fool with the wrong tool makes the disaster faster. I learned that from Mike. I’m very happy to quote him on that. So for those who don’t know him, Productized Service, that really is Mister Mike Finston, I can highly recommend it. And the target group of your podcast is, well, we are also entrepreneurs who say: What can I use this stuff for? And that’s why, perhaps to bring it all together, AI can be used to automate things, to let the machine do them, so that I have more time for people, so I can be more human.

Thorsten Jekel

I can use it to hyper-personalize. In other words, I can hyper-personalize videos the way you’ve seen me do in newsletters: Dear Martin, dear Thomas, dear Sabine. In other words, I can send out a thousand videos and have a personalized message with several variables in each video. That wasn’t possible before with AI. With a significantly higher response rate. And thirdly, I can compress. For example, I can have a one-and-a-half-hour video from YouTube summarized with the “ATIFY” extension application in such a way that I say: What are the ten main ideas? What is the summary? So that’s always important to say: automate, personalize and compress. Then the topic of AI makes sense. And then also think about what completely new business models are possible with it. We mentioned Christian Buchholz earlier, for example. He brought in investors at the beginning of last year because he said he wanted to set up an online MBA and assumed that he would need a lot of resources. He has since said that he is happy about the money, but he doesn’t need it for the resources because he is using AI for this. So it becomes much more profitable and can grow much faster. AI then makes sense in combination with human intelligence.

Martin Sänger

That is a wonderful final word, namely that human intelligence should still be the controlling component. Dear Thorsten, thank you very much for being there and for your many, many insights, which, as always, came across very compactly here. We’ll put all the links in the show notes accordingly and then I’ll be happy, because we always meet somewhere online and in a meeting almost every week. I’m looking forward to our next contact. Thank you for being there, dear Thorsten.

Thorsten Jekel

Thank you very much for your invitation, dear Martin. Good luck. Thanks for listening and see you here at Business Probeo. This was another episode of Digital for Productivity, the podcast for productive digitalization. And always remember: Switch on your brain first, then your technology.

Yours, Thorsten Jekel.

Also available in: Deutsch

Antworten

Your email address will not be published. Required fields are marked

{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}