Latest / Machine Dreams / AI Is Learning to Predict Which Bridges Need Fixing Before It’s Too Late
Transcript
- speaker-1: Bridges have been inspected and photographed for decades. And so what that means is that there are years and years of pictures and reports that show just how the structures have changed. But almost nobody has had the time to put all of that history together. What if AI could do that? Welcome to Machine Dreams, the show where we talk to the people that are actually behind the scenes building the future of AI. And helping us to understand where this technology that is just racing to the top could take us. My name is Kalawale Samuel Adiabayu. I'm a technology analyst and I am always covering cybersecurity, AI, enterprise technology. As always, I'm joined by the remarkable, delectable, one and only speaker-2: Hi everyone, I'm Leah Stern, global communications strategist and former venture capitalist. Hey Sam, how are you? speaker-1: Very well. I got my leather jacket. speaker-2: You look very cool. I seem really geeky next to your cool leather jacket. speaker-1: I na now I sort of feel like I'm Janssen Huang. speaker-2: No, it it suits the vibes. It's great. Today we're joined by Saar Dickman, CEO of Dynamic Infrastructure. It's an AI company that works with transportation agencies across fifteen US states, as well as in the UK and Australia, and is currently testing with New York's MTA Import Authority through the Transit Tech Club. His team has analyzed more than 7,000 large structures and tens of thousands of smaller ones. So how is AI being used to keep our roads and bridges safe? Saar is here to answer that and more. Saar, welcome to Machine Dream. speaker-0: thank you Simon Lea for having me here. Happy to share with you what we're doing out there. speaker-1: Great. So sor, tell us what does AI have to do with roads and bridges? speaker-0: Okay. So a c a quick a quick quiz. do you know that Bridge was bridge are dynamic infrastructure and it was planned to last seventy-five two hundred years. most people don't not aware of that. So when we're looking on a structure that lives and react to the environment for a hundred years, we need some kind of sophisticated something to monitor it and to to to watch its performance. That's pretty much the role of AR in this war. speaker-2: So for people that don't understand how it works and how AI has gotten involved, how was these large structures originally looked at, inspected, kept safe, before AI came along? speaker-0: Yeah, so let's put aside the AI word for for a second. For decades, as a result of good regulations, the owners of the structure, the bridge, the culverts, the panels are paying are paying inspection engineers and going out there to capture the actual condition of of the infrastructure. The reason for that that you continuously need to monitor the infrastructure to Eventually manage risk and and moving dollars to the right places. So the information is out there, a lot of information. Think about those seventy-five to a hundred years, cameras, people repelling from bridge, crawling under under tunnel just to take the right picture of what's the actual condition of the infrastructure. So this information is piled. Is piled in the servers, in the folders of people. Back then it was piled in papers. And the main headaches is to really understand what's the actual condition. Because things usually start fine. We have a nice ribbon cutting with the governors out there or the chiefs out there, and then we drive on the bridge, we we commute in tunnels, and they react to the environment, to the weight, to the land, to the natural disasters, and we need to track that. So Being humans, we have some or limited capacity tracking every single every single change in those a big infrastructure. That's actually the role of the software and more than that the AI. speaker-1: J just so our viewers and listeners can understand, can you frame to us what exactly is the risk or are the risks involved if this kind of technology were not in place? Why is this important? speaker-0: Yeah, the the one risk that we are all aware of is the such a structure fail. We hear this in on the news. It's dam that was a collapse, bridge which falls on the president, go there in the in the next day or tunnel, which is a close those are the catast catastrophic events that usually usually bring the those inf the infrastructure to the news. But on the day today, it's just a closed bridge. Or a limited lane, which for and force us to detour the entire city, drive half an hour, more to our work, more traffic jam. So that's the actual consequences of of not understanding what the the what's the actual condition of the infrastructure, because the information is always there. if we've just mentioned that every piece of pretty much every piece of infrastructure is documented. So the documentation is there. I'll give you an example. The the front bridge in in in Maryland in which President Biden just went on Sunday morning to visit was documented for twenty two years that something happened there. It was fully documented, but no one has the attention or the capacity or the budget to do something with that. So that is actually what's happening now. speaker-2: It's very interesting to note, you know, bridges and other structures that you monitor. I'm also curious to understand how the influx of data centers across the US may be affected or benefit from the work that you're doing. speaker-0: Yeah, it's always follow the money. Data centers data centers drive our w you know, modern world right now, the United States and globally, everybody is heavily invested in data in data centers. it's usually when you prepare the the press out of the ground to build the new one, and you wanna rate it for get the the loans or the right financing from from the bank, you present the the electrical capacity, the the facility strengths, but the civil infrastructure factor, the sensitivity to floods, to ground changes, to things that will happen around the data center and and and for sure will affect that's pretty much something new because it's eventually it's an asset, a traded asset. the the rating of this asset along the years is one of the key factors in pricing it. And we see more and more interest from the data center developers and rating rating agencies who are curious about what he'll have what will happen in fifteen and twenty years. And in this place, what we do is very what we are doing is very meaningful because again, collecting the data, combining to this world information from different sources. We mentioned the inspection report, it can also be satellite information, monitoring from above in a very high accuracy what's happening to the ground, combining anything to a current picture understanding will result a very good assessment of what's the condition of this data center civil infrastructure wise will be in ten, fifteen or twenty years. speaker-1: So y you've sort of talked about this already, but I think it's one of the things that I f find most fascinating about what you do. You're saying that AI can actually look at ordinary photographs and actually measure a crack or a damage in three D. How is that possible? speaker-0: First of all, possible. Not so easy, but possible. I'll take you to the to the technological advances because this is the right place. building a 3D of a three-dimensional model of anything is is is is not news. you can do it with a LiDAR point client base, you can do it with multiple photos, but building kind of 3D the depths in our world with a few photos is definitely a challenge, a challenge. We took this challenge a couple of good years ago, and we know how to take a single number of photos, capturing, let's imagine together crack or have a spell on the bridge, creating these depths to in a very high accuracy and understand what it means. It means in terms of measurement. When you have measurements, you pretty much open a whole world, the supply chain world. Now we can quantify and price how much it will cost to fix it, how massive deterioration, and actually compare it to other places. So disability is is definitely crucial in paving our way forward in being more efficient, not just in assessing the infrastructure, but impacting and saving a lot of money on the supply chain. speaker-2: You are holders of one of the largest visual repositories of civil infrastructure in the world, analyzed down to every defect in every year. What could you learn from that archive that nobody has thought to ask yet? speaker-0: We can learn how infrastructure behaves. I'm I'm using the word behavior, which is very human, but we have a leather jacket here, which you know protect us from the cold. Bridge has the same pretty much the same reaction when they are on the northern part, exposed to snow or down there exposed to to heavy sun, the same in tunnel, the same as dams. So the environmental changes with the specific materials and de facto have has their own behavior. As far as we know, doing concluding this one in scale has never been done before, and we're stu and we're collecting the dot and presenting the behavior. That's why I use this word of different materials and different environment reacting to different damages. speaker-2: Just to follow up on that, out of curiosity, based on some of the information we now know about how different structures age, are they building new structures differently than maybe the classic way they used to? speaker-0: Absolutely yes. thanks for science. out there. We're using better materials. we're planning just after the Second World War, bridge were planned to last sixty to seventy-five years. Now they're planning to last hundred years and even more, the materials are better. so yes, there is a conclusion implementation out there in w when it comes to constructions and materials and we see things improving, but we have to keep in mind that the majority of the infrastructure around us is aged. It's not the new one. Okay, the new one, think about your community, your state, how many new bridge tunnels, culvert you see every year. Not many very small not too many. How many deteriorated bridge roads you see every day? That's yeah. So that's yeah, that's the problem and that's the challenge. speaker-1: The other thing that that I think I find also really fascinating about what you do is that you don't just talk about the state of a structure. You also now have built probabilistic models that can sort of talk about where that structure is heading, right? You can sort of talk about give an overview of what's going to become of that structure. my question really is how confident can you be? about such projections, especially now that we know that probabilistic models are subject to hallucinations and bias in the information or the outputs that they give sometimes. speaker-0: Yeah. You know, it's pretty much connect all the the dots above the the capacity of the structure we analyzed and our ability to analyze them to a different level. But you know we c you know the the story about the tiger chasing two people, you just need to be faster than the one beside you. So we compare our performance and what's happening out there based on the existing practices. There's existing practices are very poor with the data. And the level of analysis. So we're taking what was approved by government as as official forecast models. We benchmark quantitively and qualitively with our with our data, and we are much, much better performed and much more accurate in in in defining the this specific deck on the bridge. We're discussing just bridge, but it's the same for culverts, dams, and so on and so forth. Is gonna deteriorate because if the official benchmark is 600 structure, we can benchmark against 7,000. Okay, with the history of each one in the last last in the in the recent two decades. So we're much better in the numbers, which definitely affect on the air curves. speaker-2: There's been fears of AI taking over people's jobs and how it changes different workforces and different careers. I think you have an interesting perspective. You took a relatively very human, blue collar, out in the field job and teamed up these at least in the public sector, these engineers with the tools to use AI to work smarter, faster, et cetera. Was there any apprehension by these different engineers on working with AI? Were they fearful? Did they prefer to do things the old fashioned way? What was the onboarding like in that process? speaker-0: First all, we have to recognize, at least in the United States, there is roughly 27,000 missing people, okay, people who are jobs that are not equipped with the right with the right civil engineers to to monitor and and and work out there, support the communities on the public sector. So one, there are not not enough people out there. But the old generation that left the government and left the public and the private sector a decade ago was really against any software. It was the notebook, whether it's a spreadsheet or a real notebook, and that's it. we see that the the new county engineers, civil engineers, the people responsible for the infrastructure, really adopt a different approach. They want someone to very sophisticated someone or something, very sophisticated Helping them improving their job as as long as they can take the decision. And our job is to recommend and provide the justification. They are so happy to work with AI. So it's side by side. And we're very proud to say that. And and and people actually embrace it because if you go to the field or you need to claim for a budget, you want the evidence, you want a very good assessment, you want the justification why you ask for. additional two million dollars to fix this or that world. That's our job, okay? To provide him the backup for every decision. speaker-1: so this is my last question to you. And it's it's kind of writing off the question that Leah just asked. I would like you to tell me, and generally the everybody listening or watching this right now, what the AI and engineering future looks like. And I'll tell you why I'm asking that. I'm imagining a scenario where an engineer who knows maybe the bridges or the structures that he's been working with over for like thirty years, has great experience. The engineer thinks one thing and then the AI agents that you've built is saying another thing, right? And there's that kind of friction that then occurs. So I just wanna know, I'm curious about what you think the future would look like for engineers in the blazing world of AI and for, you know, AI in the evolving world of engineering. speaker-0: Yeah. So assuming we'll have we still have humans around us in the in this future in your future. eventually eventually we Yeah. speaker-2: Maybe in that specific industry that it would be completely run by AI, perhaps. speaker-0: I'm I'm not so sure because eventually the budget allocation is is driven by people with different interests. It can be pure engineering interest, it can be political interest, geopolitical interests. and and and and to your question, I I think that AI will make our life much easier. With one person can if if if one person now in with no hair because because of the headache Is is managing a few hundred assets. in a few years he could manage ten thousand assets. But I think that people will still drive the wheel to that sense. Will navigate where this one is going, what's the real priority, which combines the human factor or the social factor with the very strict engineering factor. And if I need to be even more visionary, I don't think we should lose the ethical part of it because If it's just AI, we can ask questions like, okay, who's paying for the AI, who's interest with the AI? So I want the AI to be objective and people to react to the environment. So that's at least my perspective. speaker-1: Thank you, sir, for an excellent conversation and for really great insights. It was great having you on the show. speaker-0: Thank you very much for having me. speaker-2: And thank you to everyone listening. This is Machine Dreams. More honest, unfiltered conversations with the people shaping the future of AI are on the way. See you next time.