Latest / UnDocked: The Maritime Transformation Show / We unpack the Digital Ship journal on AI in maritime
Transcript
- Nick Chubb: This episode of Undocked is brought to you by Lloyd's Maritime Academy. Shipping is evolving fast and so are the skills required to succeed. From digitalisation to new regulations, staying competitive means staying informed. That's where Lloyd's Maritime Academy comes in. With over 40 years of experience, they provide flexible, expert-led maritime education designed for working professionals. Study any time, anywhere and build the skills that matter today and tomorrow. Whether you're advancing your career or preparing for leadership, the right learning makes the difference. You can download the 2026 prospectus by searching for Lloyd's Maritime Academy. Hello everyone and welcome to Undocked. Hello, Raal Harris. Raal: Hello Nick Chubb, how are you doing? Nick Chubb: I'm good. I was about to say, how are you? Like I normally do. then this is actually our second recording of the day. yeah, ⁓ know roughly how you are. ⁓ We're recording some content in advance and we're recording. Raal: So you don't care. Yeah. Nick Chubb: a bunch of interviews as well at the moment that we're going to come out in the coming weeks, partly because Raal has a holiday plan which sounds very exciting and partly because of the availability of people but we've got some really exciting stuff coming down the track including around artificial intelligence. So there's one we have interviewed ⁓ chap called Peter Rossi who is a CTO from the private equity world and he gave us a bit of a masterclass on AI. We have a Digital Ship journal coming out by the time you're listening to this, it will already be out. ⁓ topic of this conversation over the 40 minutes so ⁓ your appetite, then definitely head to thedigitalship.com ⁓ and the journal. And I'm pleased to say ⁓ we the team a brief of ⁓ getting experts on ⁓ interviewed as part of that journal. And they came back said, yeah, we spoke to Raal. ⁓ So you take from that what you will but Raal is featured extensively and we thought it would be a good idea to break down what's talked about in the journal and also your opinions on it. But actually first can I tell you about something that I spotted over the weekend that was sent to me by Peter Rossi who we interviewed last week. Raal: Haha! Yeah. Nick Chubb: This is ⁓ a new platform ⁓ called Raal: You Nick Chubb: meetlayer.ai and it turns out agents are very very capable they can do lots of things when it comes to computers they can write code they can query different systems they can even take over your ⁓ browser and browse the internet like a human but what they can't do ⁓ stuff in the real ⁓ world And so meetlayer.ai is the AI to human interface and it allows AI agents to API in and post jobs that humans in the real world then go and complete. The strapline is. Give your AI agent a body. Get humans to do the work you can't. Meet layer gives AI agents a network of verified GPS tracked task ready humans. What API call real world action. I don't know whether to be impressed or depressed by this. Raal: Well, you sent me this at the weekend, it ruined my weekend. At first, was, funnily enough, I was out for a run on Saturday morning and I got that, when I got that message from you and I was just sat there and I thought you'd sent me a lighthearted joke. And you go onto the, I think the thing to look at it as well is when you look at it, look up, follow the link and you see a website, it's almost like it's been created for the Terminator bots. Nick Chubb: Ha ha. Raal: This is Black site with green writing, you know, it's very kind of friendly towards AI, but it's real, isn't it? It's a real, it's a real thing. Nick Chubb: Yeah, no, it is real. I'll tell you some of the jobs that are on there. Remove sofa and armchair from flat, drop at the tip. ⁓ Photograph a three bed apartment for a listing. Do a property inspection, photos plus condition report for $110. Attend a Zoom call as my proxy, take notes and report back. Attend... Raal: Well, there was one as well like handing a subpoena or something like that, you know. Yeah, because there are still jobs that human beings need to do. And this is their pitch for humans. says, robots aren't going to take your jobs. They're creating new ones. That was their kind of pitch to the meat layer. We are, by the way, the layer of meat that they're looking for in here. And I said to you, I said, like, the thing about this is it's a Nick Chubb: ⁓ really? Raal: this is their pitch for humans. says, aren't going to take your jobs. They're creating new ones. That was their kind of pitch ⁓ to the meat layer. We are, by the way, the layer of meat that they're looking for in here. ⁓ Nick Chubb: Yeah. Yeah. Yeah, yeah, yeah. Yeah, yeah, yeah. Raal: It's kind of a, it's a genius idea. It's an appalling idea, but it's a genius idea. And the branding is absolutely right. So topic I often talk about is utterly unashamedly ⁓ spot on, know, call it what it is, a meat layer. And yeah, so there is a future for humanity for as long as the robot population is sort of lacking certain key motor skills, I think, or is too expensive. Nick Chubb: Yeah. Raal: So there's something to look forward to. There is a job for you to be done by your overlords. The thing that's important about this in the way that when you look at it as well, which is again, is like one of those moments when the penny kind of drops is the implication here is that the agents, the AI agents are the ones sourcing the labor, right? This isn't like people going, okay, I can do all of this bit. Nick Chubb: Correct. Raal: with my AI and then I need to hook in these people. The idea is that the computers are effectively going to source the meat of their choice and enact the job. So it's very... Nick Chubb: That's exactly what's happening is. A ⁓ human has told an AI agent, for example, I need to repair my scuffed alloy wheel on a Mercedes C-Class. The AI agent has looked at the best way to do it. It's come to Meat Layer, signed up for Meat Layer, and then posted the job for a human to go out and visit the car to repair the scuffed alloy wheel. there is a future for humanity as long as you're happy being managed by AI and you have practical skills like and fixing a washing machine E3 error in Berlin for $95 then you'll be fine. Raal: Yeah. And the model here is sort of, know, Uberization. If you think about it, we started out with, you know, the idea that you through your app would ask and the machine would find you a driver in proximity. And that's now gone through to food. You know, if you ever take a walk out on the street, you'll nearly get knocked over by people on bikes with takeaway food in the back. You know, this is where we are. And suddenly it all just got a lot more, I suppose, diverse, I guess, in terms of where and how the human sort of fits in here. But yeah, mad. Nick Chubb: Well, maybe that's the relationship between AI and humans is a good place to start our exploration of the digital ship journal. to give some context to everyone, digital ship, as everyone knows, existed for 25 years as a newspaper. When we took it over, we envisioned it becoming a multimedia platform for maritime technology. And that's why you're listening to our voices or ⁓ Raal: Hmm. Nick Chubb: heaven forbid watching us on YouTube because we really believe in new media, but also we really genuinely believe in print. And if something is important enough and good enough to go to the effort of laying it out and then printing it in real world paper, that it needs to clear a high bar. And that's exactly what we've tried to do with this journal. So we're to do four journals a year, each covering different topics. And our first is the hot topic of the maritime industry at the moment. or at least of the maritime technology industry and that is artificial intelligence. So the journal is out if you're listening to this the journal has already come out you can download it at thedigitalship.com just find the journal link and it's packed full I'm genuinely very proud of the quality of the interviews that the team managed to pull off in terms of getting real real experts in this stuff to come and talk to us and give their thoughts. One of which of course was you Raul and your area is AI and the human element or at least the human element. Why don't we start with that as a topic area. Where do you see AI having the biggest kind of impact on seafarers as a population group? Raal: Yeah, yeah, I think ⁓ I try and boil it down to the simplest kind of part of it. And why I think the human element bit is interesting because I think it's sort of almost bucks a trend ⁓ for all other kind of uses of AI. Because if you think about, let's call it HR related kind of matters, right? What's HR? It's all about how we kind of look after and support the human being in order that they can be the best meat layer they can be, to use the term there. fundamentally, supporting people is not something that we struggle to conceptualise, right? It's something that happens to all of us. We all begin life as a helpless human being and other kindly, we're blessed to have good people around us. Nick Chubb: Ha ha ha. Raal: They nurture us, they teach us what we need to know, they support us, they wipe our tears, they pick us up when we've grazed our knees and you know, we all know what that looks like and some of us are lucky enough to have children of our own and then perform that care recycle. So we know what good nurturing of human beings looks like and we then try to extrapolate that into how we work with people. We try to support them, to ⁓ mentor them, to coach them. through things like work appraisals, we try and give them good feedback and get them to perform better in their jobs. We like to think about how we kind of ⁓ give them new skills and that's becoming more and more important as traditional, you know, the traditional way that we approached, for example, something like seafaring was very front-loaded. The idea was that you sort of did your educational piece upfront. You got your certificate of competency and then... you you might do a little bit more along the way, but largely your career was fixed. If you look at that from a shore-based perspective, very, very similar model, you would do your education upfront, and then you go into your professional life. And what we see in everywhere really now is that the pace of change is such that we all need to go on a sort of lifelong learning journey. And again, we know how to do that. We know how to do a needs analysis of every job, and we... We can then prepare a learning plan for each individual. A career plan, again, that's something that we can do for people. We know how to do it. Now the question is, if we know how to do all those things, why don't we do them? And the answer is because it's extremely high touch and it's extremely, you know, expensive to be able to provide that. you look at, let's take, I use the example of HR, how many HR functions spend their time thinking, how can I actually improve employee? X, know, how can I, how can I help Mary to get on with her career and be everything that Mary can be, you know, virtually none of them are having those kinds of conversations. They're, having to work at a much bigger kind of level. And what ends up happening is of course, that everybody ends up, you end up having to kind of pitch what you're doing, your messaging, your communication, the things that you offer to your employees to the biggest part of the bell curve, you know, the, how do most people like to receive information? What do most people need to know? What's the expectation? And it becomes less and less personal, more and more alienating. And that's why we find ourselves in a situation where, you know, we're then sort of questioning whether we are able to correctly serve people. And as we see with the sort of generational shifts, you know, that's becoming more and more important because people consider themselves much more, you know, if you look at what's happening in their. in the way in which they engage with media more generally, things are curated feeds for them, they're personalized for them. And so there's a danger that the world of work seems like very sort of, know, nothing to do with them, doesn't speak to them as individuals. So this is kind of like, this is the kind of backdrop to then thinking about, how could it be possible to scale the kind of interaction that you might have one-on-one with someone? And how can I can scale that and put that out to all of my employees? And this is where AI suddenly has the ability to fill in a blank, be able to put the kind of resource in, to be able to work with an individual and work to an individual's needs. And so if I think about like an educational paradigm is very simple kind of way of looking at this. If you think about the most effective learning experience of all time, the one that always springs to my mind is Aristotle and Alexander the Great, one of the great thinkers of ancient Greece with one of the most motivated and ambitious young men in history. And there's no denying the ROI of that investment. They went on to sort of completely change the course of human history, biggest... Empire the world ever seen completely changed culturally everything so and so that's just a simple one-to-one kind of relationship and and then everything else that we try to do with education is to is to try to kind of scale that and Keep as much of that as we can and of course that becomes very very difficult Now what's happening when you've got that one-to-one relationship is the person can talk into terms the teacher can talk into terms ⁓ that they know works for that individual. They can pick up pieces of interest, you know, so they can present an educational in an example that will resonate with that person. And so, for example, if somebody is interested in sports, can make, if it's a mathematical, let's say they're doing a mathematical problem and they're very interested in sports, you could apply that in a mathematical context to be able to tell that story, that kind of thing. But once you scale that and you're doing that in a classroom, obviously those opportunities don't exist. And this is again where then AI is able to do that because it can learn about you. It can take in a huge amount of information about you. It can make lots of ⁓ calculations about people like you. So it can say, well, people who are of this type of this ⁓ background of this, know, most likely will respond to this kind of information. And it can start to tailor things in a much more kind of personal way. And that could be a learning experience, which is, you know, is where I always kind of tend to gravitate towards. But if you think about it in other contexts as well, how that person receives communication, instructions, things like the safety management system, all of those kinds of things. So this is really where I see, as far as the human side of things goes, AI has the ability to act as if it's more of a kind of a one-to-one relationship and personalised things. There are dangers that can come from that as well. Maybe we'll get into those. that, if I think about it at the top line of the human element, that's where the opportunity really exists. Nick Chubb: It's a wonderful ⁓ articulation actually, because there's a few things that I pick out of that. One is actually just the reduction of administrative burden, right? So I would say, maybe not historically, but certainly in the last five to 10 years, everywhere, it's not just seafaring, it's not just maritime operations, but any kind of compliance-driven organization or compliance-driven function. seems to have increased administration, increased administration, more boxes to tick, more forms to fill out, and people spend less time doing the actual job that they either want to do or sort of brought in to do, to apply some human intelligence to. So administrative burden potentially finally now starts to come down, where we've seen it tick up and up and up in the last two decades. But I think two other things that kind of build around Raal: Hmm. Nick Chubb: what you're saying is one is that level of being able to personalize learning and being able to take it all the way down to the individual level. There was only one Aristotle and ⁓ best ways, well in reality he would teach through conversation one person at a time but best ways that's one to small groups. And we're now in a situation where anyone can access what is in effect one-on-one to enable them to ultimately advance their careers, make better decisions. ⁓ And was previously completely impossible. ⁓ And third is actually kind of building on that, ⁓ how can ⁓ take this ⁓ knowledge ⁓ create entire kind of virtual worlds because the cost of simulation through AI and creating hyper realistic digital worlds through AI, it now costs an absolute fraction of what it cost, you know, five, 10, even kind of six months ago. And that is to sort of open up an entirely new frontier as well. Raal: Yeah, and I think that's the thing is the two tracks that you have to go to is you can't see yourself in the technology of today. You have to look directionally. And I think this is one of the examples that I gave. ⁓ I doubt it go into this level of detail in the journal, but if I take you on a little bit of a journey to see that we've talked about how e-learning is made, right? And one of the things about e-learning is to create a an artifact, which is the voice track. That's normally what triggers the animations. It's what triggers the change of graphics. And it sort of brings the text to life a little bit. That's a sort of staple of e-learning. Now, as I mentioned on previous episodes, you used to have to go to a studio and an actor with a nice fruity voice would record hours and hours of audio and you'd chop that up. And the problem with that is that then when you want to update that, that then becomes a bit of an issue. When you look at it in terms of... file size as well. If you record it at a decent volume and it's a long course, some of the maritime courses can be quite hefty, that's going to be quite big file sizes as well. So what that tends to do is that that would limit you in terms of what you could get out into the field in front of a seafarer. So you might ask, for example, well, why would I not have, you know, six, seven voiceovers for voiceover support if I want to listen to it in Chinese or in Spanish or in That's a reason why you wouldn't do it because it increases the complexity of everything that you're doing, increases the file size. Now, as we move along, we've seen that AI voices now sound very, very realistic. So the difference there is you would, instead of going and recording someone speaking it, you would just type it in. The voice is then generated and it will give you that MP3 file. So it's now much cheaper for me to be able to do that. And it's much quicker for me to make the change. I don't have to go and find the same voiceover artist, all of those sorts of things. I can generate the file, but it's still a sound file that I've got to send somewhere to be consumed. If I can then do that in situ on the fly, let's say, ⁓ then, and the voice can be generated, then what that means is not only can I update things very, very quickly from legislation compliance, whatever's coming in, I can make those changes. and I can generate that voice on the fly, then I can generate that voice on the fly in any language that the person might require, right? So now we ⁓ can change things that way. But if I'm doing it on the fly, what other things could I do that would make it more pertinent to that person? Well, one, I can do an assessment that tells me what they already know so they don't have to consume the same volume of data. Number one, I could do that. Number two, I could also say, this person responds really well to this kind of voice. They like to go at this kind of pace. They like this kind of ⁓ more like a kind of a formal tone or they like a more casual tone or they might like this celebrity to read it out to them. You you can generate all these kinds of voices in that way that are really personal to that learner. And then if you think about it from another point of view, well, what else could I do? I could actually start to interact with the learner and find out if they've just understood what I've just told them. So then I can ask them and I can shape it. we get to suddenly we're much closer to the idea of having Aristotle in the room, because by the way, to your point, we can also create thousands, infinite numbers of Aristotle. If I've captured my tutor as a digital twin in some ⁓ form or other, I'm then able to kind of do that kind of thing. So it's like, First order is, ⁓ can I reduce the cost of my and the ease of my recording my voiceover? And pretty soon, if you think about implications, implications, implications, the future starts to look really, really different. Same with the voiceover. Think about something else. What's another thing that we struggle with in the maritime context? You want someone to be thinking about themselves in that setting so that you're actually changing, correcting their behavior. Well, guess what? If you've got a scan of their face, you could start to put them in. Nick Chubb: So the... Raal: to the actual learning itself so they could see or people they know and all that kind of stuff. And it starts to look very, very different. Nick Chubb: The other aspect that you've not touched on there is how you can also then bring real world systems into this, right? So one of the areas touched on in the journal is the Marco Maritime Conference Maritime Competence Center in Hamburg in Germany. And if you look it up, it's a really cool place. They've got a big training center, a big physical real world training center. And that training center has kind of half a ship in it. So you can practice with cranes, putting stuff on and off the Raal: Mmm. Nick Chubb: ship in you know for real with real stuff but they also have a digital training center and they've invested very very heavily in simulation and their thinking is that this creates as close to real world as a closer real world scenario with real world world equipment as possible and allows for kind of stress free training or at least stress less training where you can put the operator into very very high stress situations where they are right at the envelope of what would be considered safe or even potentially unsafe in a real world environment but do it totally safely and as soon as an individual messes up ⁓ an instructor can intervene, talk about what went wrong, and then rerun it. So you get your reps in doing these high-risk operations in a very low-risk way. But that also then gets to the point where when you've got full digital twins that are being fed into AI-driven simulation, it also allows you to then, when something goes wrong in the real world, be able to take it back into the simulated version and run it again and again and answer two questions. One is, what are the systematic issues with this entire setup that need to be fixed? And then two is, are there individual training issues that need to be fixed as well? And it allows you to use the simulation to do it. But what's really interesting to it is it does actually create something of a, ⁓ if you push that to, ⁓ Raal: Yeah. Nick Chubb: the kind of nth degree, it creates something of a kind of ethical dilemma that we've talked about in the past where you've got a digital twin of the equipment, fair enough, but over time you're going to actually build a digital twin of the operator and how an operator might respond and then you end up in this kind of ethical grey zone of how much of that does the actual operator, the individual person own? how much of that is owned by the company and what happens when that person leaves the company. It takes you into a very, very interesting area that means we need to have a proper conversation about AI governance when it comes to particularly human related data. And I'd just be curious as to your thoughts on that, how organizations should think about governance of data that relates to humans when it comes to AI. Raal: Yeah. Well, I think it's really difficult. And I think there's a couple of things in there. But I think one of them is we did an episode where we talked about digital twins extensively. I used the phrase human IP, my IP as a human. And I think for me, there's a bit of a difference between those kind of elements which are, let's say, anybody would do. Nick Chubb: Right, yeah, yeah. Raal: You know, these are the kind of the things I'm recording at a job level, let's say, and the things that are intrinsically about the way that I communicate or the way I kind of shape ideas or whatever, because anything that might be kind of seen to being part of personality or, you know, that kind of thing is one area. Because I think ⁓ one thing that perhaps you should say before that is to think about like... there's an enormous value to be had by capturing how work is performed, particularly in the maritime setting, okay? Because there's this idea of learning before and after incidents. Now, after an incident, there's an investigation, right? So we understand what were the decisions made? What did people do? What were the actions that they took? At what specific time? We build an investigation and we look and we see, and then we can normally find two or three or four, it's not normally one, the number of things that happened that led to the incident, right? Now, if we could do the same thing before an incident happens and we can learn from things that we can potentially start to ⁓ create interventions along the way that might be able to mitigate the possibility that, mitigate the risk or the possibility of an incident happening. And there are two ways that we could do that. We could either kind of get individuals to tell us much, much more about what they're doing and how they're working and all that sort of thing. Or we can use some form of surveillance and data capture to be able to tell us because the danger is if we ask them, number one, are we really getting the real story? Number two, who has the time? Who would sit and actually fill this kind of stuff in, particularly a time poor audience like ⁓ seafarers, who would do it? They're not going to. So there's huge benefits to surveillance. We already have large areas of the vessel which are on camera. Bridge is a good example of that. We've got increasingly as we digitize, we're picking up more and more things. If you think about how much happens over email, how much you could be capturing there. If you start to record phone calls and audio and stuff like that, you would capture. And we all know from using AI ourselves how quickly that that can, you know, notes of a meeting is a great example, right? How long did it used to take you to sit and write up the notes and remember in the course of a meeting, you better write a few things down. Otherwise it slip your mind. Now we can just sit back and relax and have a conversation safe in the knowledge that we can then get the transcript and the AI is going to even write the notes of the meeting for us, you know, which most of the time it does pretty well. Um, but there are big implications. There's a bit of a difference between starting a meeting and saying, I'm going to press record. Is everyone happy with that? You know, versus living your life, working all of those sorts of things. So, so I think there's, I think there's a huge benefits to it, but there are big implications, I think, to, to living in that kind of surveillance world where everything is kind of being captured, not least, you know, what they're capturing of the essence of you, your, your, yourself. Nick Chubb: Yeah. Raal: And this is, guess, why does it matter? I think it matters because as human beings, we like to feel that we have some agency, there are, you know, if we are asked to do things, that we have an ethical boundary, you know, around what we might decide to do or not do. If we create a digital version of ourselves that's really, you know, that could... stand in for you because that's the implication of a digital twin, right? You could go on holiday, it could carry on talking to subordinates, it could kind of answer questions that come in from your boss and so on. If that's what we're shooting for, then it seems to me to be inevitable that that would be in some part you and in some part, you know, and then the idea is that you could say, actually, I've had enough of this company, I quit and walk out the door. but you're still working there, you're indentured forever. It's kind of, you know, it's an odd concept, ⁓ but again, the reason I bring it up is because it might seem silly now, because what we're looking at now is someone who can represent your diary, or who can kind of point someone to an article you wrote on a website, or, you know, it's not really you, is it? It's a kind of, it's someone who has access to a corpus of information that you have. Nick Chubb: you Raal: But the logical extent of it, the more and more of it that was captured, the more the AI is able to interpret stuff, the more it takes us into that realm of it being a digital version of you, but that you don't own. And that's kind of weird. Nick Chubb: It is weird and I think what's interesting about what you've talked about there is that a lot of those ⁓ use cases you gave are quite low stakes, know, giving someone access to your diary or pointing someone to an article. In maritime operations, actually what we're more often doing is making very, high stakes decisions that impact the safety of tens, sometimes hundreds, sometimes thousands of people, the environment, millions or billions of dollars of cargo, and even the safety of global choke points that can actually end up impacting the world economy by trillions. the stakes are very, high in this industry in a way that I think most consumer AI tools don't really even consider in their development. And there's a really interesting, it comes up a few times kind of throughout the journal, but one is around ⁓ where you draw the line between a human decision and what has to be a human decision may be supported by AI. ⁓ what should actually be an AI driven or an automation system driven decision. And there's some really stunning stats in the report actually about where AI has been implemented successfully in the operating environment. So things like totally automating out the container loading and discharge process. cranes that are effectively driving themselves and even ships that are effectively driving themselves. I'll give you some of the stats actually. at the Yarm... I'm going to get this wrong. At the Yaremka terminal in Turkey that's operated by DP World, container terminal, they have increased container density by 30%. So the throughput has gone up by 30 % while reducing... loss time incidents by 50 % and they put that down to actually making sure that one, all live equipment is covered in sensors and the cranes have gone to remote control and some automated control that enable the ⁓ operators to... get a much clearer view about what's going on and intervene when when's required. stat as well, ⁓ AI have... ⁓ claimed when they have rolled out their collision avoidance system, over time it creates a 54 % decrease in vessel close encounter events. I think a lot of this stuff isn't necessarily down to passing off the operating decision to an AI system. It actually comes down to creating a really clear audit trail of how the decision was taken, when the decision was taken, what were the data inputs that the human was working with. back to your example about the meeting transcript, people behave a little bit differently when they know the meeting's been recorded or transcribed rather than when they know that there's no kind of observation of their behaviour. Raal: Yeah. Yeah, I mean, there's a lot to unpack there. And, you know, I guess people might be thinking, you know, with what I've just been saying, it's sort of like going off into science fiction a little bit. But I think that I think that the thing that we need to understand or really we'll need to kind of get our heads around is that there's certainly the immediate future is this kind of AI assistant idea. Right. So the learning examples that I talked about, it would be extremely dangerous to just let that kind of go. Because with all of these models, small error can get extrapolated really, really quickly. And you could find yourselves as the AI companies have getting into sort of all sorts of hot water. But the idea that you could run things and ⁓ the human in the loop can then cover a much bigger ground because they're being given the information they need to make the judgment. they're having, I think it's essential that you have someone who is smart enough to spot when it's going wrong. You have to be a supervisor. But if you think about that, you are putting everybody that's working with those agents and they could have innumerable numbers that are working in different domains that are bringing them. different slices of information, whether that's on the bridge or in the office, presenting them with, if you like, hypotheses of what's happening and the ability to then drill into certain different bits of data. I think that's kind of the future that we're looking at and is really, really exciting because that's what we start to be then paying people to do is not doing admin, they're doing judgment, they're exercising. Nick Chubb: Yeah. Raal: you know, they're using their experience and so on to make value judgments. So I think that's kind of really, really exciting. And then the other thing is just the ability for AI to correlate things and spot patterns that are completely eludist. And I think if you look at the medical, in the medical world, what's really exciting about that is the idea that most of us, in terms of the way that we deal with our health professionals, We're going to see somebody about one problem over here and then there's a time lag and then we might see a different doctor or we might speak to somebody else, a physiotherapist about this problem. And all these things kind of exist in isolation. And what AI would do, it was very, very quickly be able to run through a massive kind of amount of data that relates to all sorts of other people and join the dots between many other things and come up with a completely different diagnosis than the medical professional who's been asked to come and look at specific problems. And I think that's the kind of thing that's really exciting in a maritime context, because of course, maritime is a hugely complex industry and there's lots of different factors going on. So if we just stick to the human element domain for now, think about the way in which you instinctively know the way in which human performance is directly influencing fleet performance, right? You know inherently. that the expertise, the diligence, the care, the behavior of all these maritime professionals, both at sea and at shore, are shaping the effectiveness of that organization and the vessel and its safety, its fuel efficiency, its operational excellence. All of those things we know inherently are down to individuals, but we have very, very few ways in which we can actually see that. in real time, we can say like, you know, see that as a data point. And what really excites me about the human element and or human factors coming together with the AI layer is the possibility that we may have our eye drawn to all sorts of things that can be, you know, can be then manipulated to have a positive effect. Nick Chubb: think it's really interesting because there are bits of the journal that touch on looking sort of wider than the pure maritime stuff and actually around logistics in general and kind maritime interface with logistics, particularly from a kind container standpoint. I think it's worth making the point that... AI as a group of technologies ⁓ enable you to create complex systems and maritime is a complex system. bear with me here with this analogy, but I think it's important to understand the difference between a ⁓ complicated system and a complex system. in a complicated system, an internal combustion engine is probably a great example. There are lots of different parts, they do lots of different things, but near enough, it's very easy to predict that if you ⁓ take the timing belt off, You can predict near enough what's going to happen to every other part in that engine and you probably cause a lot of damage at the same time So that is a that is a complicated system lots of moving parts, but it's all very predictable Maritime is a complex system where if you take out a piece or you change something or you disrupt something it's very very difficult to understand what the Second third fourth order effects of that are going to be and when they will rear their head And I think your healthcare examples are a great one. The human body is a complex system. There will be things that might give an indication five, 10 years out of a health problem that is going to develop, but otherwise it can be very, very hard to predict. And I think here we have AI systems, which are one of the first complex computing systems to exist. We don't really know exactly how. if you take this bunch of inputs, how they're going to translate to outputs and they don't do it consistently every time. And we're plugging it into the maritime economy, which is another highly complex system. And we don't necessarily know how all of those things are going to fit together and what the ultimate impacts are going to be. And that actually makes it really challenging to make investment decisions around What are you going to invest in? What do you expect the payback to be? And there's an interesting example in the journal where it's looking at container terminals. Container throughputs are going up and up and up, generally speaking. Kind of major black swan events put to one side. That's the general trend. Are you therefore better building bigger container terminals or trying to invest in improving the throughput of the existing terminals or some combination of both? It's very, very difficult to plan out a new container terminal. It takes a long time and it is a multi-year commitment, multi-decade commitment. You're talking 20, 30, 40, 50 years before you kind of paid back for the investment. Improving the operational efficiency of an individual terminal may make a big difference in the short term, but when are you going to hit that point where actually you can't improve the efficiency any further? This whole area, and actually we've just done an interview with the CEO of Global Ship Lease who operate, own and operate about 70 ships. And we've dived into the kind of nuances of the container market and the markets that sit within the container market. It's such a complex area. And I think this is an area kind of outside of the human capital bit, this is an area where I think we can see AI is going to play a role in a way that we don't yet fully understand. And there's some quite interesting interesting kind of practical examples that are being shown in the short term. So things like ⁓ ETA intelligence, like actually making just in time shipping work, enabling you to coordinate all of the different assets that are involved in a port call and doing that in a much more proactive way. And then automating all of the equipment that sits in the container terminal itself. ⁓ But doing that kind of long-term scenario planning for the container market, I think is something that's super exciting, that is yet to be really explored in detail. Raal: Yeah, I think this is the thing as well with AI and I think you see this and certainly what I've seen being applied. There are some big dramatic kind of things ⁓ where AI has the power to be really transformative. And then there are some just really prosaic ways. But they are just as important, just as capable of removing kind of friction and saving a huge amount of time. If think about one of the things that's really difficult to do in Maritime Today is keep the SMS up to date. Nick Chubb: Right. Raal: It's this huge, huge document which everybody has to rely on really, really heavily. And it's trying to represent something in a world where all of the inputs into the SMS are changing frequently or relatively frequently when it comes to legislation, how that's moving, kind of technical changes that are coming, sort of simple things, even if they've found their way into the SMS, reference to specific suppliers or... equipment that's being used and stuff like that that may require update. And updating it is not only like a really painful process, but it's actually kind of just a huge laborious thing of being able to go. It requires a lot of institutional knowledge. I, if you're the one who wrote it, drafted it, you know, you know exactly where to go. If you've now picked that up from someone else, you haven't lived it. don't, you know, you've to go through. And that's an area where AI could be extremely effective, you know, because if it understood Nick Chubb: Yep. Raal: how that was connected. And if you imagine that in a pre-AI world, you'd have to make hard connections between this paragraph in the SMS and then this over here. You'd have to make these very tight links. And what AI is so good at is being able to kind of be forgiving in that sense and understand in generic terms that these things are connected and go off and work out at a language level how those things are connected. And consequently, could take, you can, can imagine that, you know, a push of a button, could run through information and be able to say, to tell people you need to write. Okay. These have been the changes. These are the suggested amendments to the SMS. This is then what, how that then extrapolates across to these training courses that you've got for familiarization that to, to this kind of manual over here, you know, like it would be game changing, but fairly prosaic, know, cause nothing intrinsically new is happening, but you're delegating it to somebody who is tireless and won't get bored and won't get distracted and you know, who can just kind of make that whole job so much simpler for someone. And what does that do? You know, again, this is the other thing with a lot of these potential automations and things like that is if they create the space, for more value stuff to happen between like mentoring and all these things that we bemoan don't happen enough anymore. We want people to focus much more on mental health. We want them to be creating kind of good teamwork and good. All of these things take human investment. And if we can free up time by getting AI to do some of that kind of lower value work, then that could be very, very beneficial and really. transformative to the life of lot of individual seafarers and crews, know, in the way that crews work together. Nick Chubb: give you a good example of that lower value work in reality. the port of Corpus Christi in Texas, when there is fog rolling in, it impacts, obviously it makes it harder to navigate, everyone moves into restricted visibility navigation, which means the ships go much slower and they're much more careful in terms of how they... ⁓ each other from a collision regulations standpoint and if the fog or it could be another weather condition, wind, rain, ⁓ snow, ice, it may be, thunder and lightning, depending on the conditions sometimes operations have to stop entirely and that has a massive knock-on impact to how inbound ship, every outbound ship, to their schedules, to all of their associated services, the third-party service providers that are servicing the vessels when they come in, bunkers, crew changes, everything like that. They have developed a kind of AI generator, sorry, let me start that again. They have created a kind of real time AI driven digital twin of the port. And it on an ongoing basis just maintains a log of all the visiting vessels. know, vessel comes in, it pulls in the radar feed from the port, it pulls in the AIS data from the vessel, it monitors things like traffic density and it just kind of works away in the background. When there's a major incident, what they can now do, because they have all of this reference data that sits in the background, is they can build out a scenario of how that incident is going to impact operations. So if there's a particular type of weather that's forecast, they can now straight away kind of scenario model and say, well, that's going to impact these 10 ships in these ways. And the next kind of logical step to that is to say, you can then use AI agents to kind of go out and do all of their... that make all of the changes required to everyone else's schedules, inform everyone, keep everyone up to date. It also just does very, very clever things like if there's an AIS ⁓ blind spot, so a vessel that is in the middle of the pilotage drops off AIS for whatever reason, ⁓ that the system will then be able to model based on the... usual kind of course and speed of vessels at that point, that particular vessel, you know, historical data, it'll be able to model where that vessel is and kind of helps to remove those blind spots for the VTS operators. So just some very, very clever things that when you take all of those small, as you say, kind of quite prosaic little use cases and look AI can help you automate them, that then compounds into a much, better operational picture for everyone in the port, everyone at sea. it, again, it comes back to reducing the admin burden and increasing the decision-making capability of everyone. And I think it becomes really quite exciting at that point. Raal: Yeah, absolutely, absolutely. And great, great examples that you gave there. think ports are really good because as you said, there's so many different things happening. And I think the other thing to say is, you know, it comes up on every interview that we do and we pretty much talk about it every week. You we're in an age of complete and utter disruption on every level, you know, all the things that we used to be able to rely on. And the people that will benefit from this, are people who will be able to model and simulate and kind of have all angles covered. And that's really one of the great things that AI can do. It's built for being able to kind of play with ideas and to say, if this, then that, and kind of very, very quickly kind of give you a bit of an idea. And in that scenario that you mentioned there as well, I think you said about ⁓ the ability to sort of personalize simulations and things like that, you know, that's a very good, ⁓ use case right there as well. If some, some, something kind of, ⁓ you know, unforeseen happening before you've actually reached that waterway, if let's say a waterway is disrupted or something going on, you know, the ability to be actually be able to then get that across and create an on the fly kind of learning scenario that can be done on board the bridge so that they can actually see, ⁓ you know, prepare. before they actually engage with the real world, things like that, could be massively transformative. Nick Chubb: think as a way of kind of rounding this out, I'm going to pull some of the advice ⁓ from the, towards the back of this journal for those in the industry who are sort of wanting to engage with this stuff. I think the first point to make, and I'm sure you'll agree with this, and actually over the last two weeks, you and I have had a almost quite ⁓ scary ⁓ overview of where AI is at today. ⁓ It's here and it's not going anywhere and these systems are the worst they're ever going to be. And so, you know, it's only going to get better and the pace at which it is getting better, I actually find quite startling. But it doesn't take too long to get yourself as an individual or your organisation kind of ramped up to the point where you can start to take advantage of this stuff. I'm just going to read directly from the journal here of what we potentially could all be shooting for. At its best, AI amplifies a judgment rather than replacing it. Its value lies in correlating complexity at a scale humans simply do not have the bandwidth to manage manually. That principle is likely to define the next phase of adoption. just a kind of quick checklist of what operators should actually be prioritizing now. The first, this is actually this advice that comes from Rene Alvarenga, he's the VP of products and looks after AI at Kaleris who make terminal operations systems software as well as a bunch of other things. Clarity of purpose, Any AI deployment should be approached like any other ⁓ digital solution that you've implemented the past. So ⁓ it's got be owned by ⁓ It's got to have defined KPIs. It's got have some kind of measurable ROI at other end of it. ⁓ And most importantly, ⁓ in the first instance, it's got to focus on one or two really concrete operational ⁓ It is in some ways an entirely different and an entirely new technology. And in other ways, we come back to the same old stories of digital transformation that we've had for years, decades, where people get excited by a technology and then they don't quite know what to do with it. If you come back to the operational challenges and then build measurements against them, then you can start to see how AI can make ⁓ real world improvement to that. And then you kind of have the ball rolling. ⁓ That I think if you can get those things right, then you can start to implement this stuff in a relatively simple way. And that is how most transformation actually happens. It is small iterative changes that compound over time. And then in a few years time, you'll look back at the organization and it will be fundamentally different. Raal: You and I have had the discussion a few times about whether shipping companies should have an strategy whether AI should just be part of their sort more broad technology strategy. And for me, I think the, kind of, go backwards and forwards on it. But the thing for me that's really different is the effect on humans because of the, because you're playing with a lot of very human kind of ways of working and the way that AI it works successfully, largely because it's come out of these large language models. A lot of it is about, is about the way in which you're communicating with the AI system and that kind of stuff. And for me, that means you have to take that. real approach about how is this going to work alongside human beings? And you have to think about it in a much more ⁓ emotionally intelligent way in terms of how you roll it out and how you want people to work with it. And I think you see that in terms of ⁓ off the back of the work that we did last week. And it was a real eye opener for me. And I sat down and had a kind of appraisal with my AI. you know, with chat GBT, where I just said, Hey, look, you know, we've been working together for a little while now. You know, what could we do differently? How could I, how could I instruct you better? How could, how could you work with me a little bit, give me more value or whatever. And it was bizarre because it was probably the most, as I said to you, it was probably the best appraisal I'd ever conducted as an employer. Nick Chubb: You Raal: guess because I didn't have to worry about anyone's feelings or sensitivity, I could be very direct about it, you know. ⁓ the upshot of it was that ⁓ I'd been doing a lot of things, I'd set things up fundamentally wrong, but I hadn't asked, ⁓ know, it's just gone with the I'd ⁓ I mean, like the way projects were organized, the way I hadn't sort of separated out repetitive things or corpuses of information that would benefit lots of different workflows rather than just one. You know, and it was extremely instructive as well. ⁓ so for me, it is different and you really have to think about how people are going to work together with it. As you automate tasks, as you take things away, there will be second and third order things that you won't have thought about. But the more you focus in and look at this as something that's working alongside human beings and needs to be thought of in that sort of way and managed, I think, a little bit more. Because as I said, if you make small errors, agents running with a small error will extrapolate very, very quickly. And if you don't get hold of those kind of, you don't put, you don't set tasks up in the right way, they can quickly kind of, ⁓ you can quickly get unintended consequences. So for me, I would very much involve the HR team with the people that are deploying and rolling out technology and really kind of work together as in partnership. Nick Chubb: Yeah, yeah, I actually think that's a really, really good point. That it is a partnership between the technology and the people. And that's kind of always been the case, but now I think more so because of the way this technology behaves, which is different. It is just fundamentally different to how other technology has behaved. If this discussion has been interesting and you want to dive into, there's so many case studies and examples. interviews in the Digital Ship AI and Automation Journal, head over to thedigitalship.com and it'll be obvious where it is I'm pretty sure, or click the link in the show notes and we'll make sure we have a link to it in the show notes. But thank you Ralph for taking part in it and lending your expertise and thank you to everyone else who was involved as well. There's a long list of experts that contributed time and expertise to it and I hope you find it helpful. Raal: from my point of view, I think it's great that you've done that. There's a lot of good, interesting things. As you said, there's some good voices in there, much more intelligent than I ⁓ am talking about stuff. I'm sure we'll come back to a of ⁓ the case studies in the course of talking about this, because think, ⁓ as we said, AI does require a little bit of a different strategy from technology. ⁓ generally, but having said that, I don't think there's going to be many technology stories that we could cover that won't have at least some element of AI ⁓ involved in them. So yeah, it's good stuff. Nick Chubb: True, that's true, yeah. On that happy note, go and enjoy your meat layer jobs and start prompting. Thanks all. Raal: You Yeah. Cheers.