Use artificial intelligence strategically

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

Introduction

Nice to have you with us again. Another episode of Digital 4 Productivity and the 10th and final chapter of the book “Using digital tools effectively”, which I published with Gabal Verlag.

Using artificial intelligence strategically

Yes, today’s topic is “Are you using artificial intelligence strategically?” AI instead of MI? Do you know that? You have a deadline for an article and the day before you sit in front of a blank sheet of paper or a blank document on your PC. The deadline is approaching, but the ideas just won’t come. A Google search on your topic returns over 2 million hits, but that doesn’t really help. Fortunately, this is a thing of the past since the introduction of ChatGPT. Simply enter something into Audi and a fully written article comes up. But the result is not really convincing. You wanted to approach the topic differently. And until you’ve finally found the right input trees, you’d rather write the article yourself. Computers are stupid after all, aren’t they?

Main problems of AI

Yes, when I look at it, computers are becoming more and more powerful through pure computing power. And you can see that especially in the field of artificial intelligence. The results of artificial intelligence, of ChatGPT and similar AI services are getting better and better and the manageability is also getting better and better. So if I take a look now to say that artificial intelligence is nothing new, it has been around since a conference in the 1950s. There were two scientists who once said, what do we call this whole topic? And artificial intelligence is a bit like statistics and steroids(?). In other words, in many areas it’s just number crunching. There are about 10% of areas where you say, well, it’s more than statistics, it’s really independent learning. And AI is a bit like all those protein quarks and yogurts you see today. There has always been an incredible amount of protein in supermarket quark and skyr. Now it’s just written on them. So to a certain extent, it’s also a bit of marketing. And I always like to compare this topic of ChatGPT with a new employee. In other words, if you tell a new employee to write an article on the topic, you might not get the results you want. But if you say to the trainee, please read this book, take a look at the website here, take another look at the three or four articles I’ve written. Please follow the style, talk to this colleague and that colleague again to prepare yourself. And then please ask me again if you have understood everything correctly. Get the result. And that’s exactly how you should work with the prompts. That is, the idea that you simply enter these input requirements as well as possible. It’s the same way that you should ideally brief an employee as smartly as possible. So, if I take a look at the three main problems with AI.

Yes, first of all, try it once, it doesn’t work, get rid of it. So we often find ourselves saying that the big hype doesn’t work instead of really getting to grips with it. It’s really because we don’t get to grips with the new technology, because we use the wrong tools or apply the right ones incorrectly. So I like to say fool with the tool is still a fool and Mike Pfingsten recently added …and fool with a wrong tool makes the disaster faster. Yes, and even if we use the right tools, we forget that an application is only as good as what we enter. So garbage in garbage out. You know, this is even more potent when it comes to artificial intelligence. Yes, problem 1 is this issue of hype. Just like with the topic of metaverse, people said, boah, metaverse, great. Then you have the first attempts at it, realize that it’s somehow not yet so mature and throw it to the side instead of really sitting down and thinking about how to take it further, and of course the press always needs a new sow to drive through the village. And it’s precisely this topic that is being discussed, that AI has just displaced the topic of metaverse in the media competition for attention. Of course, if you look at what Apple is doing in this area, what Meta is doing in this area, it’s far from dead; in fact, AI is making these things even smarter in combination.

So that’s why we always have this, this culture of convenience, of course it’s always easier to get a new gadget than to deal with the topic yourself. I’d rather use the co-pilot in Word than deal with how to build a table of contents and a clean structure with this topic. So maybe it makes sense to really sit down here and deal with these topics.

Yes, then it’s simply that we often use the wrong tools or the right ones incorrectly. So I say, even with Michelangelo’s marble chisel, give me one of the great marble chisels that are available today. I only get a Mickey Mouse out of it, if that, and Michelangelo got better results with the worst tool than I did with the best. And it’s the same with AI tools. To be honest, I’m not one to say that advertising agencies are dead. I think advertising agencies that don’t use AI are dead, but those that use AI to, for example, develop their ideas even more professionally and quickly can simply generate much more output. And then the idea is that invoicing should not be structured according to the hours worked, but according to the results, in the spirit of “product means service”. So when you buy a BMW, you don’t say how many working hours do I pay for the thing? Instead, you pay a product price for this product, and that is precisely the idea behind “product means services”.

Get to know the limitations of AI too

Yes, and when I look at it, when we put waste into systems, waste also comes out again. So it’s very, very important to say, what are the solutions? So number 1, be really prepared to understand AI in practice, i.e. to say that the first question should always be on what basis was this AI trained, what data is in it? This also allows you to recognize the limitations. For example, if I have an AI for selecting employees and I have entered all successful managers in the past and they were only men, then we shouldn’t be surprised that female applications are not taken into account. Microsoft had this problem, for example. So that also means saying that if you are dealing with tools, you should also have the idea of asking the right questions. For example, which AI tools are available for which requirements, which tool is best suited to my requirements, what data is this AI tool based on? Are current events also included, results included, can you also enter your own content into the tool so that it learns from it? What is the usability of this tool like? Can I integrate the tool into my IT landscape? A very important topic. Can the tool process content in my language well? What is the operator’s business model and how future-proof is the provider? So here’s just the idea. And if you also say, “Gosh, this topic of prompt engineering is just a new tool, a new job description, I don’t necessarily believe that, but you can also simply ask about the optimal prompt structure. And with Prompt Perfect, for example in the paid version of GBT in the 4-zero version, there is also a tool that optimizes the prompts once again. The book also contains ten tips for using AI tools as productively as possible in business. I had ChatGPT write it down for me and I have to say, it’s not bad at all. So to say, do you understand the basics of AI? Define clear goals, identify relevant application areas. Invest in the right infrastructure. Do you create a data-driven culture? Yes, of course. The less data and structured data I have, the less I can evaluate. Promote collaboration between specialist departments. Invest in training and further education. Check compliance with legal regulations. Evaluate the success of the implementation on a regular basis. Stay up to date. So can this be adopted one-to-one? You can read it again in detail in the book. Not necessarily, but you can get wonderful additions, ideas and you can also use methods such as de Bono’s hat method, personas etc. and things like that in services such as ChatGPT. And there are further development services such as Perplexity, with which you can simply go much deeper into things, or next step multimodal tools such as GoCharlie, where instead of taking a picture with Midjourney, you simply take a text with PPT and then generate a post from it with a machine, you can say, generate 52 LinkedIn posts for me based on these files, these links, there’s a lot going on in this area.

Yes, solution number 3, implement it properly in your company. So inform yourself, deal with it, say what is AI? What kind of things are there? Identify use cases. Choose a suitable AI tool or platform. Less is more. Also look at where you may even have systems that integrate AI. Microsoft 365, for example, now integrates AI with Copilot, and they even have the advantage that this data is not returned to train the AI systems. Yes, start a pilot project. Always a good idea. Collect and prepare data. Train and test the AI model. So it’s important, for example, if you also have chatbots, it’s a good idea to use a chatbot where you can also see which questions were entered and which answers came out. So that’s always very, very important. Implement the tool, monitor its performance, collect feedback, adapt it and then scale it up accordingly. So the classic rules for intelligent IT things also apply to AI projects. So against this background, AI is not a completely new world, but simply another extension where you have even more computing power, more possibilities, more resources. Especially when I look at the topic of translation, for example. I was with a customer recently. They had videos produced very professionally in German, but didn’t have them in other languages. And with a tool like HeyGen, for example, you can have it translated wonderfully into all the languages in which this company operates and the quality is now outstanding. Incidentally, there are also a lot of IT tools that work wonderfully even without AI. For example, I manage my browser bookmarks with raindrop.io. If you have Microsoft 365, you can easily synchronize them across all systems using the Edge browser. I also use the Safari browser on the iPad or Chrome from time to time. So with that in mind, I like to have it in all browsers. I like to use Raindrop for this and I also have the bookmarks in a very nice, easy-to-use app, especially on smartphones.

Yes, I’ve been using Camtasia to edit my videos on Windows on Mac for years. If you need handouts, name tags, episodes, images for podcasts and bloggers, then I use Canva. It’s also very good, by the way, and has now been enhanced with AI functions.

Yes, I record this podcast episode with the Ferrite app and optimize it with Auphonic. And if you want subtitles on your PC, I use Happyscribe, web-based or with Captions you can even do this on your smartphone. However, you can always make automated appointment bookings on one page; I use Acuity Scheduling for this. There is also Bookings with Microsoft 365. It is always important that you have a system that works with your CRM system as smoothly and automatically as possible. Yes, then what are my favorite AI tools? So ChatGPT in the paid version, because you can now also create your own bots there. There are more advanced ones like Perplexity, for example Copilot. I really like (…) because you can summarize YouTube videos there. So there is a whole lot more Midjourney. In my opinion, it’s currently the best tool for futuristic images. Style Raw and V6.0 are the two most important parameters. And you have a, so my list of AI tools is getting longer every day. Every now and then I throw things out because there are better ones. You can always find the link to my current AI tools in the show notes for this episode.

Yes, there are still good newsletters, such as the AInauten newsletter. There is also the KI-Tool newsletter from Jens Kolomski. A highly esteemed colleague in this field.

Yes, the important thing is that you can also ask what are the best tips for prompts for ChatGPT. Just ask the question to ChatGPT itself again. Yes, of course, here’s another addition from my experience with AI tools: brief the tool as if you were briefing someone from your team. You would also provide more information in such a briefing than simply writing an article about this and that. Think about what questions a person might have about your task and enter the answers promptly. Or say ChatGPT, ask me questions if you don’t understand. Specify the purpose and target group for which text is to be written. Always define the role from which the AI should respond, just as you would address the appropriate employee. So I met with a colleague of mine the other day and he asked, how was your day? But he said great. I was in the studio earlier. I said cool, what kind of video did you shoot? He said why? I was at the gym. So context definitely makes a difference. In other words, if you say write an article about my gym, then ideally you need to know gym or video studio? Specify what form the input should take. So bullet points, formulated, ten chapters. Tell the AI whether the text should be written more formally or informally. Specify longer texts. This does not work perfectly in all systems, but it is at least a guide. And always work top down. In other words, always start with a title, subtitle, outline and then go down further and further.

Yes, the results are still better in English than in German. In principle, however, these systems are always so in the language in which you ask the question, you get the answer back again With Deepl, you can also translate back into German afterwards.

Top ten tips from chapter 10

Yes, here are the ten best tips from chapter 10.

  1. Deal with the concept of AI in order to understand what AI cannot yet do.
  2. Don’t succumb to the news hype, because afterwards comes the disillusionment.
  3. Hang in there after the disillusionment. The systems can do more than you think.
  4. IT tools without AI are also very useful.
  5. With Auphonic you get great sound.
  6. Create and translate subtitles with HappyScribe.
  7. Calendar(…) tools simplify the scheduling process.
  8. With Raindrop.io you always have your browser bookmarks synchronized.
  9. Deal with the input of the AI systems. You will be rewarded with better results.
  10. You can also use ChatGPT to generate prompts for other AI systems, for example Midjourney.

Conclusion

Yes, that brings us to the end of this book and next week I’ll be giving you the top ten tips from each of the ten chapters in a condensed episode. Then you’ll have the 100 best tips for digital productivity.

Until then, I wish you every success, both analog and digital.

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"}