Introduction
Nice to be part of it. In another episode of Digital4Productivity. This time on the topic of artificial intelligence or common sense. Preferably both. Just listen to the presentation I gave last week as part of the Digital Days at the German Institute for Internal Auditing. I took a look here: What is AI? How can AI be used? With a focus on auditing, of course, but perhaps there are one or two ideas there that might also inspire you on how you can use AI. And finally, I’ll give you a few tools. Of course, as always, you are welcome to have the list of all the tools I have presented. Just send me an email to t.jekel@jekelteam.de or simply click on the links in the show notes. And now I hope you enjoy the talk “AI or common sense. Preferably both.”
Yes, wonderful. And a big thank you to you, Mr. Bühne, is also a big compliment to the organizers of today’s congress. It really is great cinema. I travel a lot as a speaker and unfortunately I experience a lot of events where I see a lot of hot air. And I thought today, just looking at the introduction after your professional moderation, dear Mr. Bühne, for example by Dr. Diestel, who said in a very differentiated way that on the one hand we are perhaps not as bad as we think when it comes to digitization. On the other hand, there are a few challenges and I am very pleased to be working with you to address them. And when I look at the last session, I don’t know who was there with you from Professor Eulerich. So the topic of AI is also very exciting here. And the good thing is that these presentations really build on and complement each other. And that’s why I’m looking forward to going through the question of “Artificial intelligence or common sense?” with you today.
Yes, and on the subject of artificial intelligence, an example that one or the other of you may have already experienced. In 2019, I was in Rockefeller Center just before the pandemic in New York, AmazonGo was opened just before, so very professional, no employees, completely with AI, with cameras and so on. I only had two challenges. The first challenge was that the shopping experience was similarly boring to what I see in computer-only stores today. So the question is, is that effective? And the second challenge was usability: it took me 20 minutes to get my Amazon app to the point where I was allowed to go in. Shortly after the start of the pandemic, the double whammy here, you see me live and once here in Tyrol, I don’t know who you know. There are little huts in Tyrol for those, if you like. By the way, I always like to keep an eye on the chat, so please feel free to comment in the chat. And of course I also keep an eye on the questions. Here in Tyrol, and AmazonGo has been solved very pragmatically, namely there are these little hats you may know, there is also a camera so that you don’t grumble(?). Then there’s a pad of paper and a cash register that you can use to pay. I actually liked the system better there. And that brings me to the topic I’ve brought you today. And I have brought three things for you today and you will notice that they fit in very well with the topics that we have already seen and heard together today. Firstly, the question: What is AI? I will give you an overview of what AI can do. And on the third page, I will also show you a few AI tools, and I will slide them to the side so that you can read them. And I’d be delighted if you could take part in the survey, which you should already be able to see, and tell me what your level of knowledge on the subject of AI is. Because it’s important to me that I don’t bore you on the one hand when you say that you’re all already super professionals and on the other hand that I don’t overwhelm you. That’s why it’s really good for me if you kindly give me that. And I can see that Mr. Neugebauer, whom I would also like to expressly praise as a very professional sparring partner here from the technical side, has already shown us this. So we are already much further ahead here than the German average. In other words, our level of knowledge on the subject of AI is already in the middle range and that is also in line with my assessment and we will be happy to let it run for a moment, we will be happy to see that I give you the things that are relevant to you on the one hand, of course, but on the other hand I don’t get too technical. Please feel free to give me feedback in the chat if you think it’s too fast or too deep. Do you have any questions? And Q&A, as I said, also at the end.
This is AI
Yes, let’s start by looking at “This is AI”. And when it comes to AI, I like to have a few pictures that make it easy for you to understand. And for me, classic data analysis is always a bit like a library. In other words, I have structured data and I have structured rules, like a university catalog, where I have very clear rules and structures according to which I have program-controlled algorithms and can evaluate this data.
The basic idea of AI is somewhat different, namely that you often have structured data and that you feed the system with, let’s put it simply, sample questions and sample solutions. And the AI generates the rules from these sample questions and sample solutions, as we saw earlier with Professor Uhlig on the subject of exam questions. In other words, a big difference is that with classic programs you always say program flow chart, these are the rules, this is exactly how it has to be done. And the idea of AI is that you say no, I can do the whole thing in such a way that the whole thing learns by itself.
And how does such an AI work? And here’s a simple example to make it a little more plausible and tangible for you. I can well imagine that most of you can continue one of my sentences. The sentence: blackbird, thrush, finch. And what comes next? And the whole bird? And most of you will certainly say blackbird, thrush, finch and starling. And the whole flock of birds. And this is exactly how AI works in principle. In other words, it calculates the probability of the most suitable next term. The nasty thing is that there is often not just one version. I don’t know if you’ve heard the alternative that we used to sing as children, which was blackbird, thrush, finch and titmouse. And the whole bird, I’ll leave you to complete it. In other words, there is also another version. Which one is right or which one is wrong? Neither of them is right or neither of them is wrong. The first probability is likely to be the one that most AI systems will output because it is more common, it is more likely. And it’s important to understand exactly that, because with any tool, it’s always important to understand how a tool works.
So, for example, if you were to hand me a super marble chisel that Michelangelo would lick his fingers with, I still wouldn’t be able to carve David out of the stone. If you were to equip Michelangelo with today’s tools, he would be able to realize his idea much faster. Because he is quoted as saying, when he was asked, “How did you somehow manage to do Michelangelo? And then he says, I just had the picture in my head and I removed everything that wasn’t part of it. In other words, he had the image in his head, he knew what he wanted and then achieved better results even with poorer tools. If he had today’s tools, he would be able to produce Davids non-stop. And that is important from the point of view of understanding that you simply say, first switch on the brain, then the technology. What can the systems do and what can’t they do? And here again a distinction. Because what we usually understand by AI is what you may have seen in Terminator, namely to say, yes, it’s super intelligent, it’s improving all the time. That’s strong AI. That means that it transfers from one area to another, that it continues to develop and that it really has intelligence and even consciousness. The bad news is that it doesn’t exist yet. What already exists is weak AI. And weak AI, we’ve also heard the term machine learning today. And perhaps you know this little ecodose from home, and evil tongues say Alexa is like Siri only with a high school diploma and, I don’t know about you, we’re already desperate for these systems at home. We now have a little button where we can press a button again if the digital lady doesn’t understand us. So these things aren’t really that sophisticated. Does this mean that we should now rest on our laurels and say ah, we can do everything much better. Let me show you something else. I took this picture at an exhibition that my wife and I visited last year. It was about a technical vision of the past. And that was Berlin 1929, the radio exhibition. And it wasn’t a mock-up, it was Zoom around 1929, which means video telephony. Sound and image were transmitted here in a quality that is sometimes almost superior to that of today. So to say we have two tendencies, we have a tendency to completely overestimate technology. And we have a tendency to completely underestimate technology. If we take a look at this picture in this exhibition, for example, where these two ladies were making video calls here with the precursors of the iPhone, we say, wow! And when I look, for example, we also had pictures of General Motors in videos, where the whole family was sitting in the car, the car was driving autonomously and they were drinking coffee, and that was perhaps a bit premature. And the keyword that you may have seen before, which is always worth looking at, is the Gartner Hype Cycle. You can see it for many of these technologies. Something new comes out. At first we say wow, super exciting, it’s great. Then we realize, ah, it doesn’t work, then it goes down and then many people drop out. But the interesting thing is that some technologies make it and then slowly go into productivity. We usually only know which technologies these are afterwards. And the great danger we face is that we tend to either underestimate or overestimate technologies. We rarely have the tendency to assess this very realistically. And quite honestly, it’s not possible to assess technology 100% realistically, but it’s a good idea to try things out.
The next challenge we have. As I said, AI learns on the basis of data. And what I often see out there are piles of data that are sometimes stacked on top of each other, but often just don’t really fit together. And you are no doubt familiar with the term garbage in, garbage out. I’d like to add the topic of Nino to this, namely Nothing In, Nothing Out. This means that if you don’t have any data, if you can’t access this data, then it’s simply difficult.
The good news is that it’s no longer like it used to be, where you had to say, okay, I can’t do this somehow, but the systems are increasingly able to cope with unstructured data. Nevertheless, it is important when you analyze data, and this applies in the analogue world as well as in the digital world with normal classic IT systems and also in AI, that you first switch on your brain and then the technology. So think about what kind of data is it? What limitations does the data have in order to be able to evaluate the quality of the data and, above all, the results. And that is always very, very important. And I believe that people can still complement artificial intelligence very, very well with common sense. And the second topic is, and I found this very, very nice today in Dr. Diestel’s presentation, who said that we also have this topic, I found it very nice. Incidentally, I have a wonderful digital notepad here. I wrote things down by hand during the day today. And when I looked at uncertainty avoidance, I found a very nice quote from Dr. Diestel today, where she says that in Germany we tend to come from a culture where we are more on the preservative side. And if I take a look at the topic of employees, I have a picture for you here that I like to use with two axes. One is the question of “how digital are employees” and “how productive are employees”. And when I look at this, there are question marks on the one side that are neither productive nor digital. A clarifying discussion is certainly in order here.
On the other hand, when I see the digital dinosaurs, as I always like to affectionately refer to them, they are usually the older employees who are not so digitally savvy but have a lot of experience. I prefer them a thousand times over the digital game kids you see here, who always have the latest Tamagotchi but aren’t productive. Ideally, of course, all employees should be stars. Can you bake your employees? Nope. And now, of course, one or the other of them is saying, of course. The digital dinosaurs are the older ones and the digital natives, the stars are the younger ones. I don’t know what your experience is like. Feel free to write to me if you like about your experience in the chat, which I also keep an eye on here. What your opinion is, what your experience is. Let me give you an example. I was involved in an iPad rollout at Rewe Food Service as project manager. We trained a branch and switched from PCs to iPads. And there was a group of field staff who said that it wouldn’t work and was too complicated. And then the 62-year-old chairman of the works council stood up and said, “Guys, I’m happy to support you, but you’ll have to get off your butts yourselves. Let me show you how it works. He demonstrated. Sit down. Six. So here it was an experienced employee, also a works council member, where one or the other says, works council is always against it. No, I haven’t always had that experience. And on the other hand, my daughter, 26, now in her first job, when she had her first working student job at Siemens during her master’s degree, she came to me one evening and said, Dad, can you help me? My boss said I had to put together an Excel list and I’d never worked with Excel before. She thinks I can do it all because I’m a math major. In math, we don’t work with Excel. So she didn’t even dare to ask. And of course the younger ones, and my wife is a professor at the Beuth University of Applied Sciences, are all ten times faster at typing WhatsApp messages. But younger employees aren’t necessarily any more productive. That’s a management task. And the trick is for the digital dinosaurs to learn from each other’s digital kids. Exactly. And thank you very much for the addition. So when I look again here at Mr. Lipinski, he says that, to my great surprise, it’s the older employees who are more open to AI. Very exciting. Thank you for the addition. That’s also my experience, simply because they understand it, to be honest. And the most important thing is simply to have respect for this topic in particular. Because respect is the glue that enables these two areas to learn from each other. So very, very importantly, it is a management task and AI projects are also change management projects. They are management projects during the introduction and also afterwards during the respective implementation. And that’s why Mr. Bühne said it so well earlier, first switch on the brain, then the technology. And exactly as you say, dear Mr. Besser. Besser, you already have the right name for it, it’s always just a question of mindset, more mindset than age. Sometimes it correlates, but often we also have a self-fulfilling prophecy. Because if I approach an employee in such a way that I say they can’t do it and they’re kind of stupid, then they will fulfill it. And if I ask employees and then make use of their experience, then the experience is that it works wonderfully. Yes, very, very gladly.
Like this. I should be back. I hope you can hear me. Right. Wonderful. So, you should see me on air again. Hear me on air. The god or goddess of the internet wasn’t quite so kind to us just now. So, but as you can see, we have professional technology here that makes the whole thing work. Yes, and a lack of digitalization prevents AI. But the important thing is simply to have the basics in place. And that’s why it’s very, very important to talk about AI. Very, very important. And before we get into the topic of what AI can do, we have a second survey. And with regard to this survey, which Mr. Neugebauer will be showing again, I would be interested to know where you say that human AI is more of a gimmick from your point of view? Can it already be used productively in parts or is it a productivity booster from your point of view? And I’m also going to go big again here, so that you can of course also see me here.
And we’ll take a look at this together, in other words, and as I said, I’m always happy to hear your additions in the chat. So much for the topic of AI. I’m also pleased about your comment, Ms. Hirsch, that we don’t just have an all-male group here, but that we also have the ladies on board. And if I don’t always use gender consistently, please understand that as the father of a daughter I am very sensitive about this.
What can AI do?
Yes, I can see that most of them will have been at it all day. That’s why you say it can already be used productively in parts. And that’s exactly my recommendation: neither total hype nor denial makes sense, but a good idea is to have a healthy realism and use the tools that you can already use. And there’s always more to it. Because I am also asked again whether AI is replacing humans. And my clear statement on this topic is that AI will not replace humans in many areas for a long time to come. But, people who use AI as a tool will replace people in their professional lives, others who do not use AI. It’s a bit like if you were to write a letter, a business letter – not a personal letter, but a business letter – using a typewriter and correcting it with Tipp-Ex, it would take you five times as long as if you were to do it with text modules in Word. So nobody would think of writing it on a typewriter today. Does that make you less creative? No, quite the opposite. Instead, you don’t have to struggle with the technology; the technology supports you. And that’s exactly why I always say, let’s take a look at what AI can do. And above all, AI can, and we’ve already heard a keyword today for those who were in Professor Ulrich’s session. He coined a really great and important phrase for me. He said that we have the opportunity for full population testing. This means that AI has the opportunity to simply extract a mass of data from a wide variety of sources, including unstructured ones. In other words, you can extract data from transaction systems such as your ERP system. They can extract data from media, texts and images. We have already heard the keyword OCR, for example, and of course also from the topic of websites and social media. So if you look at the whole topic of social media reputation risks, for example, you can get a feel for it, but you can evaluate it correctly and even in such a way that you are not just in a random sample, but that you can permanently view a wide variety of data as a whole. AI can do that.
And why all this? So that you can reduce the risks that ultimately lead to you making money and ultimately also to be able to make sensible predictions for the business. For example, I was at Tschibo Cafe Service from 1995 to 2000. I was responsible for the operational business. So we filled these drinks machines. And back then it was a bit dull, just going there two or three times a week, depending on the location, depending on the customer, and whether the crate was empty or full, the service employee went there. In other words, sometimes we were there too often, sometimes we were there too rarely. We’ve also heard the topic of predictive maintenance today. These are exactly the kinds of things where this topic is simply becoming much more intelligent and where technology that used to be very expensive and very complex is now very affordable and practicable. Mobile radio modules, where the machine itself transmits when it is expected to do so, were once unaffordable. Today, they are standard items. When you say AI is also brand new stuff. I suspect every one of you has ordered from Amazon before. And it’s not the most beautiful site in the world. So web designers always say, boah, how can you build such an ugly site? But it converts easily and it’s very intelligent. This means that when you order a product there, you always get a recommendation as to what product B is suitable for it, based on experience. And AI can help with this, even more simply by looking at historical data on the one hand and providing central evaluations on the other.
We have already heard today that stakeholders simply use such systems much more if they are smarter, if they are more intelligent, if the results are much better. So this is where internal audit has the opportunity to become a much more proactive sparring partner, to really become a sparring partner for management. For me, this is really this extreme role where you have a chance. And if I look again and say, yes, do we need sales controllers? I always like to compare it to a pocket calculator. If you enter a calculation into a calculator, the calculator will, with a few exceptions, always calculate faster than you. Only if you don’t know that dot before dash is a rule that should also be used in a calculator will you get the wrong result. And therefore, I like to repeat myself – A fool with a tool is still the fool. So very, very important here. And that you use these tools wisely. Just like I can’t get a David out of a block of marble with a chisel, even if I were to take a hammer to the roof, I wouldn’t be able to cover a roof properly either. So even if you go from the artistic to the practical area here, first switch on your brain, then the technology.
And the next step is what you can simply use it for. We have already heard several times today about the challenge of the shortage of skilled workers. This means that automation is no longer a question of, yes, will I lose my job? Instead, the question in many organizations is, how can I manage to do as much or even more work with the same number of employees? So that’s why, yes, and thank you very much Mr. Kilian David, calculators have nothing to do with AI, I’m 100% with you on that. Thank you very much for your comment. And sorry if the picture wasn’t clear. What I was trying to point out is that if you put garbage into an IT system because you don’t understand how it works, the result will be wrong and that’s the case with calculators and it’s the same with AI. And we always have such a blind, often believe that we say what I put in is right. And that’s the case with the calculator and it’s the same with AI. And it’s even stronger. So, I hope that has made it a little clearer, Mr. David Kilian, that I was able to clarify that again. So it’s also important here that you can simply use it to automate things. The topic of internal and external communication, i.e. the creation of reports and things like that, are topics that simply tie up your capacity today. And you can simply do smarter things and thus become faster. And, of course, you can also look to the future and recognize new risks.
In other words, if I can simply analyze more data in less time and, above all, and this is the great advantage of AI, simply also unstructured data where you don’t yet have any rules, where you don’t even know how to evaluate it? And that’s simply the idea. Put simply, you just throw in a bunch of data and say, this is what came out of it in a different context and this system learns from that.
And we saw it earlier with Professor Ulrich. We always say yes, AI, they’re still practising. Yes, but if you just look at the difference between ChatGPT 3.5 and 4.0. The systems are getting smarter and smarter in this area. And that’s why it’s brain first, then technology. But it also continues on the basis of simply using technology. And that’s why the next thing I’d like to know, and my dear colleague Neugebauer will be showing another survey with the question Do you already use AI in your company? It would be interesting for me to know which of you are already using AI? And I can see that you are higher than I would have expected to be honest. I wouldn’t have thought that you were already using it so much. So we’re at just under 60% of you who are already using AI in your company. So I really have to say, chapeau! I wouldn’t have thought so, to be honest. So please feel free to add to this, if you like, write in the chat which solution you are already using. So it’s always interesting for me and of course for the other colleagues too, if you can share it, then feel free to share which systems you work with.
I have brought along a few systems for you, where we can simply take a look at what systems are available today, some of which we have already heard today and perhaps one or two systems that you have not yet heard?
ChatGPT
Yes, ChatGPT, the classic. If you use the paid version, which is also reasonably priced, then you not only have better availability, but you also have the option of using the 4.0 model, which is significantly better in terms of results. Importantly, you can also supplement it again here, not only with voice modules, such as the Talk to ChatGPT already shown here, but also with tools where you can also display web results. And when you say what Jekel is saying, how can I remember it all? You also have a handout with you, so you have things with you. Wonderful Mr. (…), we use chatbots for customer inquiries on the Internet. Classic topic, wonderful.
CopyCockpit
There is a system for this, by the way, and we haven’t discussed it before, which works excellently. CopyCockpit is a Swiss company and for me, it’s based on the same engine as ChatGPT, but it has a few advantages for me. The first advantage is that you can use it completely in German and of course you can also use ChatGPT in German, but the results are usually slightly better if you stay in English. It does this very well. You get very good support in terms of what prompts you enter. And you can use it to build excellent chatbots, for example, which you can embed and pass on without others needing access to the system. So this topic of chatbots for customer inquiries is another area where I see a great use case.
Chatbase
We’ve also heard about the use of documents today. The idea here, for example, from Chatbase. What you see here is that you can upload your own PDF documents here and then you can question these documents. For example, you can test the history of humanity by Johanna Harari, which is uploaded here, and you can also simply question this book. And I am always amazed at how good these results are. Tools that you can use today.
DeepL
Another tool that you may also be familiar with in the field of translation is DeepL. At DeepL you also have the option of having texts transcribed. And now you say, okay, text is one thing, image is another. When you present, you often have the challenge that the images are a bit muddy. Then you have another tool here with which you can upscale, with which you can have images upscaled again with AI. Because iCandy, if you have board members, if you have supervisory board members, if you have your users and you have muddy images, it looks different than if you present them professionally.
Forensic AI tools
However, I have also brought a few forensic tools for you, and it is very, very important to say that the topic of AI is making it increasingly important to look at tools that can identify whether this content was created by humans or by artificial intelligence. This means, for example, the AI VoiceDetector, where you can upload sound files. And it says, with what probability is this a human voice or an AI voice? The same applies to written text, which you have here. There are various tools that you can use here, right up to the issue of copyright infringements, so that you can check again to what extent there are copyright infringements that you may need to keep an eye on in internal auditing. And if you look at the fact that the Queen would certainly not allow herself to be photographed in this way, you should surely be aware of that as a human being. But if you take a look at the quality of fake photos and fake videos, you’ll see that there’s a lot going on today. And that’s why it’s important to have systems that you can use today. Because ideally, of course, AI tools should be integrated into your systems. And Microsoft, for example, is currently integrating the entire technology from Open AI, the provider of ChatGPT, into the Office Suite, into the systems, into the Bing search engine, for example. And the auditing systems are also integrating such things step by step. I don’t think I need to say much here. BullshitDetector. So to say, here I take the question of how high is the probability that this information is true? Because you remember AI works with probabilities, and that doesn’t mean it’s true. And ChatGPT 4, by the way, fantasizes much less than ChatGPT 3, which is much better. And, on top of that, fact checkers simply help even more. Yes, copy leaks too. Here’s another tool. And if you say, how can I keep track of all these tools that are out there, then I’d be happy to give you two tips. Firstly, the newsletter from AI Nauten – recommended by my wife Professor Dr. Nicole Jekel, who is also a regular speaker here at (…). Very good newsletter, where you regularly receive good information. I can also highly recommend the newsletter from Jens Polomski, who has been working on the topic of IT tools for a long time and is now also working on the topic of AI on LinkedIn. And then, last but not least, a database. This site from Advanced Innovation, where you can search for tools and take a look at them. So you can read on the Internet. If you want something to listen to, there are also good podcasts, including one of mine if you like, where I talk about Digital4Productivity once a week, because the idea is that technology doesn’t make us less productive, but more productive. And my experience is that technology is often the second largest cost factor in companies after personnel, but does not necessarily always increase productivity.
Conclusion
So my conclusion When we look at what Call can do, what AI is, what AI can do, what tools are available. The original question we had was artificial intelligence or common sense? My conclusion is best both. And with that, I look forward to your questions and hand over again to the highly professional moderation of dear Mr. Böhnis.
Yes, I hope the presentation has inspired you a little to not only think about the topic of AI, but also to use AI. Because you know, I firmly believe that when it comes to digitalization in Germany, we don’t have a knowledge problem, but an implementation problem. And I also see this very strongly when it comes to AI. Because when I ask you, who among you has ever heard of ChatGPT, a lot of hands go up. But when I ask who uses ChatGPT regularly, few hands go up. And that’s exactly what I typically see. We know a lot of things, but we try far too little. And other cultures, such as the Americans, are clearly ahead of us in this respect, because they simply try out new things.
With this in mind, I would like to cordially invite you not only to understand AI, but also to grasp it, to try things out. And as always, the links to the presentation can be accessed via the show notes. And if you would like to have a presentation on AI for one of your next events, just get in touch with me!
I look forward to seeing you and your next event.
Yours, Thorsten Jekel
Also available in: Deutsch
