INKOVEMA Podcast „Well through time“
#212 GddZ
A look behind the scenes at INKOVEMA
Small series:
New technologies for mediation and conflict counselling. Part 8
In conversation with Gregor Nobis
Gregor Nobis: Computer scientist and responsible for artificial intelligence at INKOVEMA. He advises on the use and benefits of new technologies for the consulting industry.
Well through time.
The podcast about mediation, conflict coaching and organisational consulting.
Contents
Chapter:
0:03 – Introduction to the AI Compass
2:57 – Determining the status quo of AI development
4:23 – Development of the AI assistant
7:42 – The MediationSimulator in detail
12:15 – Challenges and learning spaces with AI
15:49 – Practical applications of AI in everyday life
17:18 – The state of AI in the consulting industry
23:55 – Outlook for future developments
28:58 – Conclusion and outlook for the next episode
Summary of content
In this episode of the "Gut durch die Zeit" podcast, we discuss the fascinating world of new technologies and artificial intelligence with Gregor Nobis. As a technology fan and computer scientist, Gregor talks about his experiences and how he has started to develop new products that offer real added value for people in connection with developments in the field of AI. We reflect on the rapid development of AI technology, which has changed dramatically in recent months and years, and therefore the challenges that come with it.
We start with a Positioning, how this technology has developed at INKOVEMA since then and what innovative tools we have introduced within our institute, INKOVEMA. Among other things, we talk about the AI assistant that we have created, which is specifically tailored to the needs of mediation and is based on extensive content from our training resources and additional knowledge. This assistant has enabled us to significantly improve both internal use within the institute and support for our clients.
During the interview, we also take a look at the Mediation Simulator, another AI application that allows users to play through various mediation scenarios and gain valuable experience in the process. We provide insights into the challenges of text-based input and the desire for more natural voice output to make these tools even more user-friendly.
A key issue we are addressing is the need to create learning spaces where consultants and users can test and explore the applications of AI. This led to the creation of the AI Compass initiative, through which we support a community of experts who deal with the practical application of AI tools. The idea is to familiarise the consulting world with the possibilities of artificial intelligence and to use technologies that not only make work easier, but also enrich it.
We emphasise the importance of not only moving in the fast-moving technological landscape that surrounds us, but also asking the right questions: What are the challenges in our consulting work that we can tackle with AI? We also discuss the added value that we can achieve through the targeted use of AI in our day-to-day work as consultants and how these technologies, despite their proven effectiveness, are not yet reflected in existing tools.
We also dare to look ahead to possible future developments in the field of AI and ask ourselves how far we are technically and what this could mean for our industry. We are convinced that we are in a dynamic phase of change and want to continue to observe and understand the application of these technologies in order to be able to offer relevant solutions for consultancies.
To summarise, we find ourselves in a very vibrant and constantly evolving artificial intelligence landscape. We look forward to delving further into these topics and exploring new applications that may be useful for our consultancy work.
Complete transcription
[0:03]
Introduction to the AI Compass
[0:00]We get an overview for you and pass it on to you. And at the same time, you can try things out with like-minded people and people with the same goals and problems, because there is always a practical component to the AI compass. Hello and welcome to the podcast Gut durch die Zeit. The podcast about mediation, conflict coaching and organisational consulting. A podcast from INKOVEMA. I'm Sascha Weigel and I'd like to welcome you to a new episode. And today's episode is all about new technologies and artificial intelligence. I would like to welcome Gregor Nobis for the first time here in my podcast studio and also in the real podcast studio in the Spinnerei right next to me. Hello Gregor. Hi Sascha, nice to be here. You're here today under this motto because we've been working together a lot in recent weeks and months on the topic of new technologies and artificial intelligence. Today I'd like to take stock of the situation.
[0:58]But first, who are you and why are you doing something with artificial intelligence? Wow, yes, that's an interesting question. I've been fascinated by technology for a long time and have always done so in my studies and professional life to date. I was a computer scientist and have worked in many things related to it. And with the current developments in the field of AI, it was completely natural to try things out, to see what's new, how you can use it to build products that really help people. Exactly, that's what drives me to do things. That's also what brought us together, the idea that this AI is now here, what can we do with it and how can we use it, for example, for INKOVEMA and how can we stay on the ball technologically so that the development that is taking place doesn't bypass the institute, I'll say that now. And I mean, I can, I have to be frank about that, because I…
[1:53]I am endlessly grateful that this collaboration didn't just start with AI and that I've always had professional advice, but I have you to thank for the fact that I'm involved with things like smartphones, the internet and computers at all. And so it was quite logical that when these new technologies, i.e. this turning point, came about at the end of 2022, beginning of 2023 with the release of ChatGPT, it was quite logical that I approached you again and got you more involved. And then, in order to determine where we stand, both with the institute, but also as a consulting and training institute with these topics in a representative and, at best, exemplary manner for the industry. In order to position ourselves in this area, we then published the first AI, tailored to mediation. That was a little over a year ago.
[2:57]
Determining the status quo of AI development
[2:58]Where would you say we are today in this development if we jump right in?
[3:05]Let's start with the review. Really quite exciting. When we worked together for the first time, I still had to convince you to get rid of the fax machine. I don't know if I've managed it yet. Maybe there's still one hiding around here somewhere. What being a lawyer in 2010 means first of all is getting into the world of the industry if you had a fax machine. That's just the way it was. There was no getting around the culture.
[3:28]Yes, exactly. And in any case, it's always difficult to determine where we stand in the field of AI. The technology landscape has…
[3:36]In fact, it has always had the ability to change very quickly and we are currently experiencing this at double to quadruple speed with AI technology. A lot changes from week to week, from day to day. You never really have the feeling that you've landed in front of the wave, but are actually always a bit behind the developments. And that's great on the one hand, because you can learn something new every day and try out new things. On the other hand, it's also a challenge when it comes to declaring things ready or getting out of a test operation and putting products into production and handing them over to people and saying, this is good enough now, you can work with it. Because you always have to worry that you'll be disproved the next day.
[4:23]
Development of the AI assistant
[4:21]and things could work much better. So it's kind of difficult to determine where we stand. Perhaps we can talk about the topics that we have already tackled for Inco Firma. The tools that we've developed, a little bit of a long haul. Exactly, we can start with that. In addition to using ChatGPT for our very own research area, translation area, etc., we then started by putting the IncoFema AI assistant online, initially preparing it with the help of a CustomerGPT. And that was really a test operation for us, because we now saw ourselves as a training institute, not just purely as a consulting institute.
[5:05]In the meantime, I would say that the technology we set up 12 or 14 months ago is outdated or, I think, no longer up to date, or what do you think? Well, I've just had another look at our notes from back then. The first attempts we made were actually still with GPT-3.5. Hardly anyone talks about that now. It's perhaps still used for a few special cases, but generally not any more. That shows how much has changed. And I remember that output didn't work at all with 3.5. We always had to make sure that 4.0 was switched on because the requirements were already too weak for 3.5 back then. He was just talking rubbish. Exactly. There was a lot of hallucination and that is of course disastrous for an assistant who is supposed to appear as competent as an institute or someone who is employed by an institute. And it was important to make sure that we always used GPT-4. But then we realised, well, it's a lot more expensive if we do that, but we've since switched to it because it simply works so much better than it did before. And it hasn't got any cheaper.
[6:15]No, not that. Perhaps some background information. The assistant is not a pure LLM application, but is trained with additional materials, such as the content from the IncoFema website, as well as various literature sources, training materials, exactly, they are all included. In this way, we have created a system that goes a little beyond what GPT can do out of the box and brings more knowledge and is also closer in tone to how it is taught and spoken here at Kofema. Exactly, that was the AI assistant who really helped us, as an institute, as a training institute and further education institute, to understand how these AI assistants are used. By ourselves, but also by our customers. That has already had a big impact. But we weren't so much involved in counselling. In the sense that it has helped us as consultants. However, we have seen that workshops have been designed, that topics have been looked up, i.e. requested, consulting topics, concepts. And, as everyone else does with other GPTs, with Cloud or JetGPT, they were able to familiarise themselves with topics really quickly. The second thing we built was the mediation simulator. From what I've learnt, it was more complex.
[7:42]
The MediationSimulator in detail
[7:42]What do you say, configuring or prompting? Exactly, perhaps that's also quite exciting in terms of the process, when we report on it now from hindsight, it all somehow sounds as if we had thought about it beforehand. In principle, however, we started with the idea that we now have AI and what we are actually going to do with it and we have these steps, we are now making an assistant and we are now making the simulator, we actually got to grips with them through the first steps that we simply took and didn't really know beforehand.
[8:09]What will this lead to and at what point can we really say that we now have something here that will help people. There was actually only one question where I always thought that technology was still lagging behind and that we were ahead in our demands. That was the issue of input, i.e. the fact that entering text in the form of chat is already very time-consuming and that it has long been possible to use voice input in other applications and hear the voice output and ultimately have a conversation. Perhaps for all those who are not yet familiar with the simulator, let me briefly explain what it's all about. The first project, the assistant, was more of a tool that you could use to quickly ask a question that would then be answered, either in terms of content on the subject of mediation, transactional analysis or counselling somewhere in this area. But also organisationally, for example when it comes to dates or something like that, when does the next training course start or what is the content? From a certain module or something like that. So, the simulator is a completely different type of AI application. We have tried to find a playful approach to how I learn mediation or how I can learn it. And the simulator works in such a way that you feed it with an imaginary case study at the beginning.
[9:28]Mediation and then how a role play starts, in which you can try yourself out as a mediator, hold preliminary talks with the parties to the conflict, all in the simulation. It's all still text-based, i.e. typed text. Just as you just said, we quickly realised that this is very tedious as an input method. And it also limits this playful approach to a certain extent, because the game itself is perhaps exciting and fun, but this process of typing in text doesn't feel quite so natural and has often led, as we have seen when people use it, to people looking for shortcuts or choosing the route where the simulator provides the content. Sometimes you let yourself be sprinkled. Exactly. People were interested in what it was saying and writing and there was less of an invitation to enter something themselves and get feedback. Yes.
[10:31]Especially since, and this is still due to the fact that feedback was always very friendly and favourable and it was very difficult to really find improvements. So that was my impression when I used it from time to time and realised, well, as a real learning tool, you learn more about AI at the moment than about mediation. Yes, we definitely saw that, it was the same with the assistant and similarly in the mediation simulator, that this topic of AI, LLMs, what can they actually do, is already on many people's minds and yet these tools were also tested to see if they could, for example, alienate the actual purpose and use them for other things. The assistant was used in the same way as a normal GPT assistant would be used to answer any question and not in the area of Incofema. The aim was to push the boundaries, to break out of this simulation game and get the simulator to step out of its role and reveal things that were not intended. We have definitely seen that. But we also saw that if you get involved in the game and accept the work that it means to type in something meaningful, you can get meaningful results and this process, this playful process that is intended.
[11:59]You can also play through with pretty good results. So if it works well, then it's still quite impressive, but you also have to play along and it's relatively easy,
[12:15]
Challenges and learning spaces with AI
[12:09]to get out of these guard rails for a good simulation and drift off. That was also a reason. At least, I don't know if I communicated this much, but I also realised that I needed learning spaces to use and apply AI tools and try them out. And we… then founded and launched this AI Compass initiative, where we get together with colleagues and try out the whole thing with an AI application in a very low-threshold, very concrete way using a specific use case. We have now held several meetings and it has been well received, more than we expected, so that we now have a small community of several hundred people who watch the videos live with us, so to speak, and then watch the use cases as well as the videos and are actually really willing to use and consider AIs.
[13:11]Not only for the consulting business in the administration, so to speak, but also to apply this in the context of concrete consulting work, which is enormously high. To be honest, I would never have guessed that before.
[13:29]But you were also involved in the AI Compass. What is your impression of where we stand as a consulting industry of process consultants? We are not management consultants who also develop AIs ourselves, but really process consultants who actually focus on interpersonal communication as the centrepiece of consulting. So let's start with the success. I was there for the first few days and noticed how more and more people signed up and at some point it was well above what we had originally expected. And I believe that this is also due to the multitude of information and options for using AI, which we mentioned earlier, that it seems so difficult to get an overview, to get in front of the wave, so to speak, and to feel well informed. And that's what this AI Compass guide for advisors was all about. We have.
[14:27]We get an overview for you and pass it on to you and at the same time you can also try things out with like-minded people and people with the same goals and problems, because there is always a practical component to the AI compass, where you can try things out together with colleagues or distant colleagues. Yes, I think that's what makes this format so valuable, that you get something concretely demonstrated from this huge range and can look at it together with others to see how I can use it for myself and simply see through this huge jungle a little better, because I would also say that the main difficulty in using AI today lies in this. Which tool do I actually use now? This is directly followed by the fact that most of them are chargeable and it's not so easy to say I'll just try everything out, because you can very quickly reach your budget limits. Yes, but before that, which tasks do I actually want to use with an AI tool? So what are my actual tasks as a consultant? That's where I'm focusing, so to speak, because I think that's where we have the greatest opportunity to achieve our own digitalisation and our own application or facilitation through AI. What are the tasks that actually challenge me?
[15:49]
Practical applications of AI in everyday life
[15:46]And I remember that it was very simple.
[15:49]This issue with the photos, which is a huge effort to have good photos of yourself for the business, for the website, for profiles, for posts and you don't always want to take the same one. And it actually works like this, you somehow get one that's good and then you take it for ten years. And how much easier it is now with an AI application to have professional photos. It has nothing to do with counselling at all, but it's a development.
[16:17]Which makes the business as such absolutely easier. So at least by a day's work for photographer and then pick the thing. Hey, you who listen to this podcast, don't forget to rate it and give feedback. Thank you so much and now we’re moving on. Yes, exactly, and I think that we'll be looking at many more examples in future editions of the AI Compass, which are not primarily pure consulting problems, but which can help consultants because they are simply part of everyday working life. Things like optimising calendars, how do I manage my time, we are still a long way away from the content-related issues. Where do we actually stand as an industry?
[17:12]Not in terms of subjective abilities, but in terms of how far is
[17:18]
The state of AI in the consulting industry
[17:15]AI that it really helps and we just haven't used it yet? Or are the things themselves still in their infancy and is it perhaps even better if we don't use them at all because they are not yet fully developed? Yes, that's a very interesting question. I believe that we are already further along technologically than we are in terms of actual application. With the technological advances that have now been made, many things are already possible that have not yet reached people. And I think this is largely due to the fact that the tools we use on a daily basis are relatively slow to incorporate AI capabilities. For example, I notice that the application I use to process emails doesn't yet support AI. That's not a technological problem, it's all there, such as summarising and writing emails, separating important from unimportant ones, etc. AI can do that very well technologically. AI can do that very well technologically, but it's just not in the applications yet and I think that applies to many applications.
[18:15]So because you haven't added it yet or because it's not yet in your software? Because it's not yet included in my software. Email isn't the only tool I use either. I have a bunch of probably 100 different apps that I use every day. The effort involved in looking for solutions myself is simply high and the penetration of the existing tools with AI is still relatively low. I remember a year ago, or maybe even two years ago, I told a forum that it wouldn't be long before, for example, I would be travelling to Frankfurt next week, I would make an entry in my calendar and say Frankfurt Main, I'll be there in one day. And then an AI tool will automatically book me the train journey and put the ticket on my mobile phone, in my Deutsche Bahn app, book a hotel and… that it is known, so to speak, that I like to go there and not somewhere in Wiesbaden. And if I don't keep the appointment and don't cancel, then these tickets and these things are automatically cancelled. This is not yet the case. Why not? I've seen that it works elsewhere, but it's not the case here yet. Are we slow or are we not demanding this enough as users?
[19:33]I think it's just too difficult to get there. And that brings us back to the question of which software it is built into and how. If at the moment you make your calendar entry, drive to Frankfurt or maybe you tell Siri, I want to go next Tuesday, I have an appointment in Frankfurt, please enter that. If you were asked at that moment, dear Sascha, you don't have a train ticket yet, I can see that because I have access to your train account or something. Should I book one for you, would you like to travel second or first class? If this happened automatically, if these connection points were there, then I think it would be much easier to use such things. At the moment, the landscape of these tools is still very fragmented and it's not integrated into the tools that we already use anyway, but I can automate travel bookings, but I have to take care of it very specifically. And that might be worthwhile for people who now work in a larger company, for example, and do nothing else all day but book trips for the various employees.
[20:32]But if I only book something for myself from time to time, then I don't yet have the pressure of suffering or the great benefit of looking for a specific tool that can do this. And I think that's true in many areas. And that's why we see that the main use of AI today is really still chat tools, which are stand-alone applications and are not linked or are poorly linked to everything that happens outside. With my private knowledge and also with the knowledge of what else is happening in the world. But that's how it would work in practice, because that's what I realised when you said that. It's introduced, so to speak, because my technical devices ask me what follow-up actions they can still carry out and make me realise that they can. Which I normally just do as a matter of routine.
[21:27]So I realised when you said that you don't have to enter the same thing in the calendar, you just have to tell Siri. And I hadn't even thought of that, although it's probably been going on for two years, three years, five years, ten years. Yes, I think that's the thing, that people just stick to the tools they've got used to and leave the tasks to them. And if you think back to the invention of Google, that was so long ago that it's almost hard to imagine, but before Google there were other search engines. And the moment Google came onto the market, the search issue was, let's say, actually solved.
[22:04]Nevertheless, it still took a very long time for everyone to use it, simply because nobody knew that this solution existed and because everyone continued to type in their AltaVista or Fireball or whatever the other search engines were called that had bookmarks. And of course they weren't talking about Google, but simply continued to spit out the bad search results that they had before. And then this tool slowly became more and more popular. And so tools are simply sluggish or software is sluggish because once it has made it to the user, it stays there for a while. And either it changes or is replaced by better things. In the case of Google, it was the case that Google replaced other search engines because they couldn't keep up with these improvements. And I think that will also be the case with a lot of software that comes from the non-AI era if it doesn't manage to make this transition well and incorporate AI in a meaningful way. And there are exciting developments going on right now where I would have thought it would happen faster. For example, I'm still amazed at how little AI is used in Google search, because we were just talking about Google. It's still very traditional in what it spits out.
[23:13]It will be exciting to see how this finds its way there, but also in all other areas. And the first steps are already being taken. With Apple Intelligence, I think a large number of people will have more access to AI tools. Users will then decide how good they are. AI tools are now being consistently integrated into a relatively closed and harmonised ecosystem. And therefore, the corresponding users would then also be tempted to use AI. I've heard that's coming soon.
[23:50]If something doesn't happen now in the big political weather situation,
[23:55]
Outlook for future developments
[23:53]then that will also happen for us. But that's already the point, so to speak, and that's why I think it's quite consistent and right that we are increasingly emphasising the importance of creating our own routines, our own consulting work with the AI Compass.
[24:09]During the counselling session, but above all before and after, to question and look at what we can actually improve there. What is annoying us?
[24:17]So I'll come back to this example. We just go to the photographer as normal and take a photo there. And we might feel new when we do this with a machine where we go. But the fact that we can do this with photos that we can take on our mobile phone and then get professional photos on the computer alone is something that is not just a technical question, so to speak, but it is a question of whether we want to interrupt and at least test out our own cherished routines. Okay, Gregor, in view of the time that has passed and the many things that I would like to enquire about now, keyword Titans, Google is Titans. That's right, we haven't even got round to it yet. That would be a small outlook, but it seems to me that we could just poke around in the fog a bit. If we were to touch on that. Yes, well, if we were to touch on it, we would probably come across an exciting topic. Sascha, what you mean is Google Titans. So what I would mean. What I would mean in theory, exactly. It's a paper published by Google, which is seen by many as the next stage of AI development and is supposed to solve some of the problems that we currently have with the so-called LLMs, the language models that we typically use.
[25:43]May I try it, how I imagine it, so that the whole suffering becomes clear? In just a few years, we have learnt to build chats on certain topics at ChatGPT on the side, if necessary to build our own special GPTs so that they can remember certain topics and queries that we have already made in inverted commas. And this separate libraryisation of chats would no longer be necessary? Exactly, that's what it would mean in practical terms. The background to the project or this way, this AI architecture that is being proposed, the background is to be inspired more by how the human brain works. That you have a short-term memory, a long-term memory and then somehow a kind of meta-memory that recognises the memory itself. The disadvantage of the current LM implementations is that the context in which I talk is new every time. And that's why we build these chats that you mentioned, which then live in the sidebar, which we dig out again when we need the knowledge again, but we have to manage it ourselves, so to speak. And we don't have this long-term memory into which things are transferred and a once-trained LRM model as we currently know it.
[27:12]Can stop not learn more, but it can that, what it white, always again give and in one certain Context also Things customise. But the Underlying lying Model, the Knowledge, the there inside is stuck, leaves itself not without new Training change. And these Problems should with this Titans architecture solved become. Now must one times watch, who the then actually in a functioning Product transformed, when one times ins Year 2017 rewinds. Also there has Google a pioneering Paper published, there went it at the so-called Transformer. What we then seen have, is, that Google itself from other simple overtake leave in the Implementation. I believe partly from the So a little Safety concerns and perhaps also, because them as Group simple Carrier were as other Player. And OpenAI has quasi on Basis the scientific Findings from this and other Papers even the GPT family developed and is then clear on Google passed by. And I believe, one can also now still say, that Google in the Range LLM always still lags behind, as now the Number one Player to be, but one plays there.
[28:15]All play somehow Ketchup, but Google creates it there never so right on the Top to come, but is there also rather always so a little back on it. And the means, also when this Paper now again from Google is, the means still long not, that we at some point a Product See would from Google, what on it based and great is, but the can also from whole someone different come or perhaps have other Researchers in the In the meantime also still better Ideas. I think, that there long not only Google on it researches. And exactly, so it remains exciting, so can one it believe I summarise, also the entire, the was Yes Beginning the, so would we it summarise, set the Case, we would have the now addressed and the would then Yes also one whole good Conclusion again to the Beginning allow,
[28:58]
Conclusion and outlook for the next episode
[28:55]it went Yes so at the Topic Positioning, where stand we actually in the Range AI. We stand in one very lively Area, the from Day to Day always again what New to offer has. Had I on Beginning already said, applies now always still and the becomes also, believe me, still one While so continue. The is good. I happy me also on it and we become the then also really more concrete tackle in the AI Compass. Since become we the to follow, which Products actually then help, also in the Counselling work and for Consultant then also Helpful to be.
[29:25]For the Moment for the first time, Gregor, many Thanks to, that you for the first time, I hope but natural not the last Paint here with not virtual, but whole in the real Podcast studio are. Yes, has Fun made. I come with pleasure again. The is good. Is good to know. And I fear also not, that us the Themes go out. Nope, the Impression have I today also not won, that we there somehow on one End are. Handled you probably. You must now go. Likewise, yes. Best Thanks to for the Invitation. Was beautiful, the here also in live times as Studio to see. I knew it although as Office, but not with more current Recording so to speak. Has Fun made, times thereby to be, also the Technology here to see. May I the here say, that not everything on First attempt one hundred per cent works has? Nothing works on Right away.
[30:09]Yes, beautiful also in the Share yours Labour times live looked in to have and with pleasure again. Likewise, Gregor, until soon. Until soon, comes good through the Time. The was my Conversation with Gregory Nobis to the Topic artificial Intelligences And where stand we actually with this Tools in the Counselling work and for the Consultancy business? When you these Episode please has, then not only like and the Podcast subscribe, but on the Website incofema.de and the AI compass subscribe and then there with thereby be. The AI Compass, one Initiative from us here INKOVEMA, at the Consultant world with more artificial Intelligence familiar to make, Tools to utilise, the the Working facilitate. We talk not about Technology, but we talk about the Counselling work. For the Moment thank you I me, that you here with thereby were, say goodbye me with the best Wishes. Until to the next Times. Come good through the Time. I am Sascha Weigel, yours Host from INKOVEMA, the Institute for Conflict and Negotiation management in Leipzig and Partner for professional Mediation and Coaching training programmes.
Leave A Comment