New Year's wishes

Transcript

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

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

00:00:14Here you will find clear classifications, real practical insights, and a fresh perspective on what is possible.

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

00:00:24Food for thought, something to smile about, and above all, something to talk about.

00:00:29A warm welcome to Think Different, Think AI.

00:00:37The Christmas season is over.

00:00:40What rhymes is good.

00:00:42I hope Christmas was a reflective, contemplative feast for you all

00:00:47in the company of family or wherever you were.

00:00:52Let's not forget those who worked during this time,

00:01:00as one should certainly recognize, because especially in places like

00:01:05power supply, healthcare, police, or others that I surely

00:01:11can't list all here, I just want to take a moment to say thank you.

00:01:16With me is Jens, who seems to have made it through the Christmas season well too, right Jens?

00:01:23Is everything okay?

00:01:24Yes, yes, all good. As always, had great food. Had a lot of fun.

00:01:28Drove around a bit, but everything manageable.

00:01:31And I'm just enjoying the time off between the years.

00:01:36Then I can tinker around a bit.

00:01:38Tinkering with some things, trying out one or another AI.

00:01:42I'm looking forward a little to what will come next year because since

00:01:49we started the podcast, it kind of began in summer '25, the pace

00:01:56at which new AI tools, innovations in models, and so on are popping up has picked up even more.

00:02:01So if everyone wants us to have even more content to report on,

00:02:06Mark, I sometimes feel that way, and that's why I'm talking about it a lot.

00:02:09Yeah, yeah, I think they're doing everything just so that we both have enough things to play with and

00:02:15to report and tell about. I'm really looking forward to the year 2026,

00:02:21which starts in a few days. Yes, we had considered using this

00:02:25current episode to think about what do we wish for 2026?

00:02:31And I'll say at this point, I hope I'm not surprising anyone and

00:02:35I hope I'm not disappointing you. So I won't wish for a GPT-6 or a Nano

00:02:41Banana 5 because I assume that will come anyway. Today, I'll put something in my wishbox

00:02:48that I'm pretty sure we'll get because it's something closer at hand,

00:02:54namely the whole topic of Voice First, the new form factors, the Smart Classes. So

00:03:03why do I say closer to the box? I took the liberty of ordering this

00:03:08Avengy 2 glasses, even if they will not come out until 2025, so that

00:03:15probably in 2026 more will be delivered. I'm pretty sure

00:03:19that we'll experience quite a bit there next year,

00:03:24regarding the always-present assessment of your environment, that is to say. The

00:03:31glasses don't have cameras, the always-present availability of AI in the form of wearables

00:03:38like glasses. I believe that very much and I hope so too, that

00:03:44something like that will come because we are both also glasses wearers, so some kind of add-on,

00:03:48that is not as bulky as an Oculus or Apple glasses, which I somehow have to

00:03:53wear to have a VR feeling. I find that a bit nice,

00:03:58that I can say we are moving away from only being able to interact with the world by

00:04:04somehow interacting with a kind of screen, so with the digital world, by having a

00:04:08kind of screen in front of me on the desk or holding it in my hand in the form of a

00:04:12telephone, I would really wish for that. And with AI, I believe a lot is

00:04:17possible because AI can take away a part of the interface. So we don't need

00:04:23to click a certain number of buttons and something else. AI can prepare the information for

00:04:28me in my context situation as I need it. Because we will also be

00:04:34entering a multimodal world. It won't just be text, it will be video, it will be sound,

00:04:37I can talk well with the AI, it can understand me well, we have notebook.lm,

00:04:43has often been mentioned, she can make great podcasts, she can tell stories in an exciting way

00:04:46for me as an adult, for me as a child something different. So I believe,

00:04:51there will be so many different interaction possibilities that go far beyond the normal, I have

00:04:59a two-dimensional screen in my hand and constantly have to navigate the surroundings and

00:05:04somehow pull myself up if I want to have any information. I would very much

00:05:08hope that AI will provide other input possibilities, other output possibilities really a

00:05:14Boost will provide in the next year. There will be some devices,

00:05:18that will probably be smart glasses, but maybe something completely different.

00:05:22Maybe it could be something like, I probably use headphones all the time,

00:05:27sometimes it's actually quite nice to have some AI on my head or in my ear

00:05:32and have a little chat with it, not while I'm out for a walk or something.

00:05:36That's definitely nicer than having to walk around and constantly look at my

00:05:39I look at my phone and then trip over a branch in the park. That doesn't happen often now.

00:05:44And if you have a variable that calls for help, so that you can be provided accordingly

00:05:49will have AI at least somewhat included.

00:05:52But I also believe that text is increasingly just one of many channels.

00:05:59with which we will interact and that this whole visual aspect, be it video, is

00:06:03that it can audio or whatever in all directions, so that it can be combined, interact

00:06:10can, can be obtained, so that it becomes increasingly present and then for example plays a role in glasses or

00:06:17somewhere else.

00:06:19What there is, if I am now in terms of wishing, then I am going in that direction

00:06:25it goes to the engineers and designers out there, I think I've mentioned this before.

00:06:30I expect that we will gradually, it's already shaping up a bit

00:06:33and new interaction patterns are emerging again, so this is something that we used to know

00:06:37because everything was slow back then, because the modem only had a 56k connection and you had to

00:06:43essentially bridge a little waiting time until the website loaded. We don't know that

00:06:47anymore. Our world has been shaped in the last years, the last decades

00:06:52by the fact that I click this, the video loads immediately, the website loads directly.

00:06:56We have optimized everything down to the microsecond so that the conversion rate is high. One

00:07:01knew that if the customer came from Google to their website and the website

00:07:06did not load immediately, then they would quickly bounce back. This is the

00:07:12performance-optimized world we live in. Now, with the topic of AI, because it's no longer this

00:07:17click that I make with AI, but rather I engage in a conversation, whether that’s through

00:07:23clicking, real speech, or text input at that moment. To this

00:07:29In a conversation, there is a certain kind of response that can sometimes simply take a bit longer,

00:07:35because I can also simply tell the AI, think about it first,

00:07:38search first, or the AI figures it out itself,

00:07:41that it takes a little while, right?

00:07:43You also described Workforce, which I would love to use,

00:07:45in other episodes already, where it’s about,

00:07:47that maybe girl-AIs interact with each other

00:07:49and come up with the optimal outcome later.

00:07:51Now to my long-winded speech

00:07:53to the designers and engineers out there.

00:07:56I think we need something,

00:07:58and we need to experiment with it,

00:08:00how this pattern can work again,

00:08:02Now imagine this hypothetical walk in the park, while I’m talking to my AI that

00:08:09I assigned to fulfill a certain request now, and then it takes hypothetically maybe

00:08:1540 seconds. Meanwhile, the person Mark is coming towards me, and I naturally don’t want,

00:08:20the AI to just start chatting while Mark is in front of me,

00:08:24but we also need to develop things like, when it’s done,

00:08:28I have to ask again. I have put the result together now, want to hear it now,

00:08:34want to send it to you, want to somehow read it to you on your phone.

00:08:39It shouldn't always be like that, it shouldn’t be as it has been so far, that I say I have now

00:08:44made a click, I had an interaction or I have taken an action,

00:08:48set a trigger, and then at some point it comes, well, not immediately in the next

00:08:52a second, but suddenly half a day later, it feels like as if it just

00:08:56works, as if I had expected it in the next second.

00:08:59You know, that’s something different. So this waiting period needs to be

00:09:03consciously shaped with its end, in my opinion.

00:09:05I have this thing, it’s probably happened to everyone and probably

00:09:09everyone has experienced it. You’re sitting in a meeting, in a conversation, in a nice

00:09:13gathering and somehow it’s a bit boring or I don’t know

00:09:16and you quickly look at your phone and at that moment you open your

00:09:20browser and you realize that you had left your browser open

00:09:23while a video was playing. Your brain is really quick at that second, because you’ve

00:09:28already clicked, it opens up and starts blaring at full volume and

00:09:33everyone is turning around. Yes, something like that shouldn't happen and I don't think that

00:09:37AI will ask you in the future. I would wish that AI is indeed

00:09:41aware of whether it is a good idea to tell you what it has found

00:09:47out now, because suddenly a woman, a colleague, whatever, came towards you

00:09:53or the phone rings or the postman rings and now you don’t want to

00:09:56continue listening to the poem or anything, and that it behaves context-aware there as well.

00:10:02Definitely for this context-awareness, we need to be clear, it requires

00:10:05an always-on setup, because it must theoretically

00:10:08perceive what is happening around us. From a usability perspective,

00:10:11it would be perfect if it, of course, because I am talking to you,

00:10:15like saying hello Mark, gives you a wink or something else,

00:10:19that I, that she realizes through this that I am now in a different interaction pattern,

00:10:24that I am no longer engaging with her at a 200% level, and then she filters that out and waits

00:10:28until the moment, okay, I can now tell my user, here,

00:10:33I have the result, because apparently it's, you know, past the market or I

00:10:36get a little vibration in my palms, so that she really reacts contextually

00:10:40and that the interaction is designed in such a way that I know, okay, the

00:10:44working phase is over. So I can continue the conversation with my beloved afterwards after Mark has

00:10:49talked my ear off in the park. I can then continue the conversation with my dear Kai because she is already

00:10:57ready with the result as soon as the conversation with Mark ends. And I believe such things,

00:11:02now, of course, there is also a data protection perspective and everything we are thinking about here

00:11:05is not so simple and difficult at the moment, but I think such things

00:11:11need to be broken out of. With all the surrounding factors that need to come in, that is, of course, the conversation,

00:11:15that I have with Mark, definitely cannot be recorded while I basically

00:11:19have my microphone open so that the AI can interact with me. That must be directly

00:11:22deleted. All such things need to be designed in from the outset, and I am again

00:11:27glad that we live in the European Union, from the outset it should be included.

00:11:31And that is the advantage of the EU Act as well, which next year, that is

00:11:36also a thing for 2026, that all these things established in the EU AI Act,

00:11:43they also come into effect in 2026. I believe that grace period will also be over.

00:11:46From mid-2026, everything must also be traceable.

00:11:49I must be able to explain as a company how agents actually operate.

00:11:53And accordingly, is it a good thing that we have this part?

00:11:56And I believe if companies, designers, engineers do this right here in Europe,

00:12:01then you can also design the application so that it has a high usability.

00:12:05Nevertheless, the topic of privacy is also important.

00:12:08I believe that in school, what we had mentioned before really comes into play,

00:12:12the whole topic of small models comes into play again.

00:12:15Not the one large artificial intelligence that basically consolidates all sensors.

00:12:21And then, depending on how the patterns are, somewhere in the data center it decides,

00:12:25what it displays on your glasses, on the speaker, and so on.

00:12:28So there is also a very good interplay between what happens On-Device,

00:12:33what happens in the cloud? Where is it a collaboration of systems to manage information flows,

00:12:39displays, and so on?

00:12:42Before we now get to the UI act, which didn’t even appear on my wish list, it’s nice,

00:12:46that it is there, and I’m also very happy to be in the EU.

00:12:49There are plenty of countries that somehow, I know,

00:12:53I wouldn’t want to go, whether I’ll come to the States next year,

00:12:56because I’m not at all sure about my DNA status, but that’s another topic.

00:13:00We are not a political podcast.

00:13:02I would like to return to the wishes and also something that feels more tangible,

00:13:06where the USA and China, but also Germany play a role, namely robotics.

00:13:10Yes, robotics, to say that now with the Neo, which was presented, yes with remote

00:13:17control operates, so teleportation of a person who with glasses and a controller then

00:13:22does the services for you, I found it also quite funny, the story,

00:13:26if you saw it, Tesla also had such an event, the humanoid robot

00:13:30Optimus was seen, and it performed a movement and you can see in this video

00:13:35how it executed a movement and what it knocked over, how the hands go up to its head,

00:13:39as if it was taking off a pair of glasses it is wearing, and after it took them off,

00:13:44which it wasn’t wearing, it falls. One really has the feeling that an operator has

00:13:49controlled the thing from a distance, took off the glasses while operating, and at

00:13:54that moment the Optimus fell at this event. I’ll put the video in the

00:13:59show notes. But if I could wish for something, not that the Optimus would fall over,

00:14:05I would really love to have an Optimus. But it’s another topic, that we really want to see more robotics,

00:14:10whether in the form of those Pixar lamps, which we already mentioned in a previous episode

00:14:15that Apple had in this lab, or really in the form of something more than just that,

00:14:22what today is just the robotic vacuum cleaner of your choice cruising around the house

00:14:27and sucking up dust or nowadays taking your socks away with tiny grabbing arms.

00:14:31Maybe we would really see something humanoid. I find that,

00:14:36I find that, I find that really cool. I think, I think that's cool. I would now just

00:14:39predict that in Germany, you won’t see humanoid robots yet,

00:14:44humanoid robots in 2026. So I think in toy form maybe, yes,

00:14:50maybe a little bit, as I said, I’m always waiting for the cool, the cool R2D2 robot,

00:14:56things that drive around here and with which I could converse, I think that’s great.

00:14:59I believe such things will simply come. You might already have something like the

00:15:07vacuum cleaner robot, which will then also interact with something else via voice in your

00:15:12household. Such topics, when we see humanoid robots, I believe primarily more in the

00:15:17industrial setting, where we really say, you see, whether it’s in figures or

00:15:24other robotics companies that are out there, you can already see very, very good approaches. I believe,

00:15:28in such controlled environments, where sometimes, in quotes, one hopes,

00:15:35but can also go wrong. We will see that. I believe that in German households,

00:15:42we definitely won't see that many Hume next year. Can one wish for something?

00:15:46Can one wish for something? Is that allowed? But I think I would first say,

00:15:49normally I am also very, very optimistic about such topics. I do believe

00:15:53that we won’t see them yet. So, here at this point, greetings to a

00:15:57Reiner. You don’t know him. I go back to school days. He always says to me, chubby people

00:16:01rub their bellies. I do it to myself. It works wonders. Maybe you have

00:16:06the wonder here, and then it was the rubbing's fault. Many, many. May I still

00:16:11express a wish? Yes, please. I would have wanted to say something,

00:16:15no now, but I’ll try it anyway. You already said and

00:16:18some people have heard it in past episodes, that things change here, end

00:16:21and so on, so workflow automation is very much in demand for me. We have

00:16:25also occasionally talked about Lovable and about Gemini and how to build yourself

00:16:30little helpers. I believe the second to last episode was when we reflected

00:16:33on that. There were also a few examples presented. So I wish for 26,

00:16:39that everything becomes much easier. Even though I enjoy making it fun,

00:16:43building workflows, I don’t actually want to build workflows, below the frontend so differently

00:16:49coding, not coding, vibe prompting, doing vibe coding that hopefully also just a bit

00:16:55punches together the backend. I actually want to describe my problem, maybe discuss it with the

00:16:59machine and implement complex things. So as an example, I would like

00:17:06to also get a bit more concrete here, you might remember especially

00:17:09I had mentioned my advisory board before. That was a bit of work

00:17:14to say, okay, which agents do I need and where do the prompts come from? And how do I build a

00:17:19agent that then constructs the conversation between the agents he selected? And

00:17:25I have also seen one or two demos where discussions can be done,

00:17:32Can get systems to produce this completely autonomously for me, so that I no longer

00:17:36have to do it myself. If this continues with the variety, then that is already...

00:17:41And it will definitely continue there.

00:17:44I think so too. So that's one of those things where I would say I'm totally with you.

00:17:48I believe that 2026 will be the year where we really see this shift from

00:17:53human interacting only with one AI

00:17:57to human interacting with multiple AIs.

00:18:01You have already seen this about these workflow issues this year as well.

00:18:05Towards AIs, interacting with other AIs almost autonomously based on a task I previously set here.

00:18:14So it will hopefully not be a matter of this anymore, and that's also a hope, but I think we will see this as well.

00:18:21I absolutely need to know how to set up such a workflow.

00:18:25I need to know that workflows even exist. I don’t know that at all.

00:18:28Maybe it’s just not necessary at all, and the things should do that in the background.

00:18:32This is the difference between IT 2.1.0 money and IT 2.0 money, in which we are currently engaging, this AI age.

00:18:40These things should not work automatically. I believe we will really see this shift into these agentic networks we have often talked about, where I say,

00:18:49I push,

00:18:52or I call out or blow into the horn my wish for what output I would like to have.

00:18:57And in the background, a magical network of agents takes care of it, who then communicate with MCP servers, use standard APIs, autonomously use a browser to visit web pages that aren’t there, maybe generate their own APIs if they see, oh, this is a repetitive task I’m doing again.

00:19:21So I’m trying to essentially build something standardized, so I don’t have to start over with each case every time.

00:19:28All these mechanics, which modern IT provides, allowing complex systems in digital to be available, should now be made very human again.

00:19:41That this human component is also being pushed into the background and can be replaced by AI.

00:19:47Because these are things that democratize the power that IT brings for

00:19:53many, many more people out there who can build meaningful things for themselves.

00:19:57And it shouldn't feel like this year anymore, but I'd also like to

00:20:00give an example.

00:20:01I recently built a daily news agent again, I think it was in ChatGPT,

00:20:06that cost me again somehow three

00:20:12prompts.

00:20:13I find that too much.

00:20:14The last prompt, and that's why I'm saying this, asked, should I do that,

00:20:20you haven't even told me that you want this daily, and honestly,

00:20:24I did that at the beginning, in my first prompt, that I wanted the daily news

00:20:28again, and then I thought, did I forget? No, you haven't given me

00:20:33the command to compile it daily again. Those are things,

00:20:35oh, Mark, I really have to say that annoys me and I don't want that in the next

00:20:41year anymore. We're complaining at a very high level. A year ago, we had completely different worries.

00:20:48Then, when we started out and thought about where the journey is headed. I say

00:20:53not that you turned it inside out, but the possibilities of the systems have

00:20:57changed quite a bit in 20, 25. We wish for something for the next year

00:21:04and I'm really curious about what we will discuss in some episodes next year.

00:21:10We have already thought about one or another thing.

00:21:12I'm curious if we will still do an EU AI episode.

00:21:15Let's see.

00:21:16Definitely.

00:21:17Definitely.

00:21:18Yes, okay.

00:21:19Good, then let's see, let's see about inviting.

00:21:20We had already thought about it two episodes ago, right?, next

00:21:22year maybe on a stage, next year then.

00:21:24Where, yes, the project that you are currently working on in your free days

00:21:28to hear something more about it.

00:21:29But I would say, Jens, the Christmas time is over, the turn of the year is

00:21:35approaching.

00:21:36I wouldn't promise too much when I say, according to me, we will

00:21:39see each other again next year. I'm sure you'll hear us next year again. And so

00:21:45that many others can hear us too, I would recommend you, let them know, that

00:21:51sings different. Sing AI. Your companion will also be in 2026 in a perhaps casual form to

00:21:59learn the news that will certainly affect us all. As I said, we were

00:22:06surprised. If 2026 allows, it will surely surprise us again.

00:22:11And with that, I would say, Jens, I wish you a nice turn of the year, come healthy

00:22:17through, don't do anything stupid, yes, don't do anything you wouldn't do.

00:22:22Never.

00:22:23And for God's sake.

00:22:24No, we can't do that.

00:22:26All the best to you out there, have a good start into the new year, and we would be happy about every

00:22:33person who remains with us and everyone who joins. And with that, I am now

00:22:38finished for today. We give the last word to Jens at four to one and say goodbye until

00:22:42next year.

00:22:44Thank you very much, Mark. I can only echo that. It has been great fun, this

00:22:48first joint year of our podcast history. I'm looking forward to our second year

00:22:52in 26, and despite the stormy times we might sometimes experience

00:22:58Not only the issues playing out on the political world stage, look positively into the new

00:23:06year.

00:23:07We both believe that AI will help us with many, many things.

00:23:10Remember, we are soon in 2026 already.

00:23:13Some of your parents might still remember that the world was supposed to end in the year

00:23:182000.

00:23:19That didn't happen.

00:23:20So I'm also optimistic that it won't end in 2026.

00:23:24We're looking forward to it.

00:23:25It's going to be neither a cool year, cool AI tools will be released that will

00:23:30soften things up for us in many areas.

00:23:33Critical topics will also come up again, and we've already looked at those this year

00:23:37as well.

00:23:38That is a promise.

00:23:39So we won't always just look at everything through rose-colored AI glasses, but

00:23:43we will also critically examine another topic next year.

00:23:46to question one or the other AI-Teo.

00:23:48I'm curious whether we will actually see real AI Christmas Tady-Bern next year

00:23:53under the Christmas trees in Germany, talking with our children. I'm being

00:23:57positively optimistic that there will be nice little teddy bears and R2-D2s and

00:24:06C3-POs, which might come in toy form, at least then you have a humanoid that can

00:24:12walk around a bit and will give us the feeling of the future. So, on to the New Year.

00:24:19Exactly. Look good again. See you then. Bye. See you then. Ciao.

00:24:49Thoughts for a smile and especially for joining the conversation.