The Mayflower Autonomous Ship: AI and Automation at Sea
28:58 Share

Hello, hello. Welcome to Smart Talks with IBM, a podcast from Pushkin Industries, iHeartRadio, and IBM. I'm Malcolm Gladwell. This season, we're talking to new creators, the developers, data scientists, CTOs, and other visionaries who are creatively applying technology and business to drive change. Channeling their knowledge and expertise, they're developing more creative and effective solutions, no matter the industry.

Our guests today are Brett Fanoff and Don Scott. Brett and Don are responsible for creating the world's first unmanned, fully autonomous ship to cross the Atlantic Ocean, a research vessel they've dubbed the Mayflower 400. Brett is the director of the Mayflower Autonomous Ship Project, and Don is the CTO of Marine AI.

On June 30, 2022, the Mayflower 400 successfully completed its voyage from Plymouth, UK to Plymouth, Massachusetts. It's both an homage to the original Mayflower, which crossed the Atlantic 400 years earlier, and a bellwether for the ways autonomous technology will push the boundaries of maritime exploration in the next 400 years.

On today's show, the unlikely origins of a self-directed ship, some ocean misadventures, and what AI and machine learning will mean for the future of seafaring and beyond. Brett and Don spoke with Lauren Ober, host of the forthcoming Pushkin podcast, The Loudest Girl in the World. Lauren is a longtime radio host and reporter, helming shows like NPR's The Big Listen and spectacular failures from American public media.

Okay, now let's get to the interview with Brett Fanoff and Don Scott. Don and Brett, it's really great to be talking with you guys today.

I was wondering for each of you, what is the draw of the sea? I mean, it's like this expansive place. It feels so unknown in so many ways. But I'm curious, like, what is the allure there? For me, it's I wanted to be I wanted to do aerospace. So I always feel like I'm like the poor cousin of aerospace. But it isn't. It's actually it's harder to do the underwater stuff. It's closer. It's just harder than being in space. It's

It's incredibly hostile and wildly unexplored. What I like about it is that you can take a bucket and go down to the beach, get a bucket of water, analyze the bucket of water for the next 20 years and

you know, chances are pretty high you're going to have a couple of things in there that nobody's ever seen before. And that's every bucket of water everywhere in the world, right? So I like the idea that you get to discover something new all the time, and it's also hard. It's a difficult place to work. So it challenges you to come up with new ideas and new ways to do things and new materials. And that's what I like about it. I don't know, Don, what about you? Yeah, I mean, there's obviously an allure and a draw. There's some great descriptions about why people are drawn to the ocean.

talk to the authors and the poets, you know, it's definitely a real sort of visceral feeling that people get. I think you'll find that the people that are involved in ocean engineering or marine science, like, you don't just sort of fall into this career by accident. You make proactive decisions to get involved in that environment. So you have a bunch of people working there that want to be there and sort of have this understanding of...

that this is the place they want to be and this is where they want to work.

So that becomes a very positive work environment, workspace, because everyone's, they want to be there. So there's that. Yeah, it's highly collaborative, isn't it? It's like anything, there's personalities, but it tends to be a lot of fun more than anything else. It's challenging in all the ways that make life interesting. And then it also tends to be a good time. Yeah, you can't work in the ocean by yourself. Well, you can, but it's kind of hard. So like Brett said, it's an incredibly collaborative environment.

I mean, if you want to be doing anything of significance, you have to be working as a group because you need to rely on each other. It is an incredibly dynamic, hostile environment, very humbling. So you find you're going to achieve success

as a collaborative group as opposed to some sort of lone wolf type attitude. Right. Okay, so we're here to talk about the Mayflower Autonomous Ship Project, which obviously is very cool. How exactly did you guys decide to build an autonomous ship and then model it after the Mayflower? I mean, it was just a hold my beer kind of thing. I'm sure. Yeah.

It really is. It really was. Oh, it really was. Yeah. What it really was, it was... So in meeting with the city of Plymouth on something else, they were talking about what they were going to do and maybe build a replica ship of which there's already one. And I thought that wasn't the best idea. You're talking for the 400th anniversary. Yeah. And so I was a little bit

in my comment as to how they wanted to proceed with a possible replica. I think you said it was a stupid idea. I said it was stupid. And there was more I couldn't resist. And I said, there already is one, you know, and it's just, I grew up near there. And so they said, all right, smart guy, what are you going to do? I was like, oh, we should build one that

us technologically and from an engineering perspective and sort of invokes the spirit of the original risk-taking and do something that informs the next 400 years. And everybody was like, yeah, you should do that. And I was like, you know what? I will hold my beer. And so I called Don after the meeting and I was like, oh, Don, we have to build an AI. I need Captain Watson because we're going to build an autonomous ship to cross the Atlantic. And he was like, great.

And so, yeah, and it was just that, literally that glib. But it also, I mean, he and I have been working on unmanned systems and autonomous systems for a long time together, 20 plus years. And so I wanted to see where we could get to, like how hard could this be, right? I mean, AI, sure, let's do it. And then,

So we built a ship. You mentioned capturing the spirit of the original Mayflower journey. And I wonder what exactly were you trying to capture? Was it the spirit of taking risks or was it doing something that hadn't been done before?

What we were trying to do, we knew was really hard, right? Like, and it was a huge amount of risk to undertake it. Brett's the real risk taker. He's the one with the big ideas and wants to take the risk. I'm a little more cautious and sort of pragmatic in the sense of, okay, what's it going to take to do that? We actually didn't think we were going to make it or I fully expected at some point the ocean would get annoyed and smite us. You know, pilgrims, like that to me is what's interesting. Pilgrims took a risk, right? So every one of them,

fully expected that they would die, if not on the voyage, within like the first year, right? That's how it was. Yeah, sure, sure. And it was worth it to them to take that risk. So our risk is infinitesimal by comparison, right? It's tiny. What was our risk really? We'd lose a ship we spent some money on. So what? Right. The knowledge about how to approach these problems and the experience that you get to give people

to take risk at that level from an engineering perspective is really important, right? Somebody had to do the first open heart surgery and took a risk. Now we're not doing open heart surgery, right? No one's going to die. So what's appealing about the risk thing is it has a technical risk and a environmental risk. And then there's a legislative and regulatory risk because we had to have our fights with various agencies about the fact that they didn't have a law that said we couldn't. So they didn't get to say no, just because they didn't want us to.

And at the same time, trying to create a reliable machine and then some sort of an AI machine learning based system that would be safe, whatever that is, in the middle of the ocean. It's really interesting and gives people a lot of purchase for different people with different skill sets to collaborate. Brett and Don started developing the Mayflower autonomous ship in 2016. It took them six years to figure out both the software and the body of the boat itself.

In that time, over 70 people contributed to the project. Lauren asked Don and Brett what it really took to go from Hold My Beer to an actual ship. You know, it is mind-boggling when you think of how many people are involved, how many people are touching this project, how many interesting minds doing interesting things. But you have to funnel it all into this one project.

Well, I don't know if it's that way. I mean, I guess you could say there was one project, but there were lots of projects. And so, you know, there was sort of the hardcore group of people that are trying to build the actual software that works. And then there's the guys trying to build the hardware and they have an interface, but they're parallel pursuits that don't have direct overlap. And then we said yes a lot to anybody who wanted to help because we learned from experience that most people don't last in terms of

the ability to stick out four or five years focus on a project is very hard. And so the people that wanted to stick it out and bring it to fruition ended up, you know, sticking it out. And that was great. You know, and then there were all sorts of different things. There was a group making a web interface so that they could show the world what we were doing. And, you know, then there was a PR group that was marketing things and sort of talking about how we tell the world about it and we would support them. But it's hard to describe it as one project, I guess, would be my position. It's lots of

Interlinked programs, yeah. Right, sure. I get that. I get that. Can you tell me more about how automation is built into the ship and how it works? Oh, well, there's tons of automation in Mayflower. I mean...

Mayflower is like most robotic systems, right? So you peel it open and you find, you know, programmable logic controllers and motor drives and all sorts of other things, sensors and industrial automation that you'd see, you know, in an elevator or an escalator or industrial machinery for manufacture. And that's one sort of layer of it, right? So you've got the basic analog control, then you've got sort of a veneer of automation and then what I would call sophisticated automation, which Don and I have worked on for years.

decades in the marine space. So all that's in there. And, you know, Don and I talked really early on. If I just wanted to get across the Atlantic, we could have bought an old fishing boat, filled up the fish holds with diesel fuel and put a cheap autopilot on it and sent it. It probably would have got across. But so what? It's not reducing risk and it's not unburdening a person and it's not doing anything really clever or sophisticated.

And so what we were more interested in was getting to a point where instead of having to tell it to do everything, saying, go do this task, write a goal, like go to Plymouth. Right. And then while you're doing that, oh, by the way, while you're doing that, collect all this science data. And if you see anything unusual, tell us. And while you're looking for all these unusual things and trying to achieve your goal, don't hit anything. So then what role did IBM's technology play in all of this?

Yeah, I mean, their technology is all over the ship. Probably the main contribution, it was the decision-making process. It's an automation tool, ODM, Operational Decision Manager.

It's actually a financial services tool. It's for making decisions about the viability of a transaction, whether it's fraud or not, or a loan, or let's say. And we were being presented this by one of the ODM engineers. And I remember sitting in the room with Brett thinking, what in the world does a financial services product have to do with marine navigation?

And they sort of were brought to realize by the IBM engineer how this isn't really so much about financial services as it is about making really difficult decisions in a really complex environment, which is what they do with financial services. But it's also exactly what we needed to do in re-navigation. And when the system was actually running, it would create a log essentially of why that decision was made.

so they can validate that decision and verify and validate that that was in fact the right decision. And so that's one of the key IBM tools that are on board. Well, one of the things you might want to consider about that is the fundamentals, right? The theoretical underpinnings of all the AI that we're deploying now have been sort of understood for decades, right? And so now we just happen to live in a world

where the microprocessors are up to snuff that they can deploy some of these very sophisticated theoretical and reality, and all of which IBM's been involved with from inception, based on its pedigree as international business machines. There isn't an IBM product that I can think of that we haven't tried to utilize or deploy. So it's everywhere in the ship. I think a lot of people think of technology as a creative pursuit, but...

I imagine building an autonomous ship from scratch takes a lot of creativity. And I'm wondering, do you guys think of your work as creative? You know, engineering is essentially design. Technological innovation, sort of, you think of it as a very logical process.

And there is that for sure. But there's an incredible amount of innovation involved too. Like there's no template for what we were doing. And, you know, we call it white paper design where you're basically given a blank piece of paper and a goal, which is, okay, a ship that's going to cross the Atlantic. Okay, come up with some ideas, right? So, I mean, it requires major conceptual leaps and then the technical skill to realize those leaps.

You're not going to make any advances just doing things the way you've always done them. You need to stretch. And the only way to stretch is with implementing new ideas. You can spend a decade, we call it PowerPoint engineering, where you do nothing but think of things but don't actually do anything, as opposed to what we call full contact engineering, where you actually build the boat, write the software to go on the boat, and send it out on the water.

Get your teeth kicked in. Get seasick, you know, all that sort of fun stuff that happens when you're out on sea trials. And because that's where the actual learning is happening. That's where the actual development is happening, is being out on the ocean. Crossing the Atlantic is no small voyage for any vessel. But the Mayflower Autonomous Ship Project is more than just about sailing from point A to point B.

Automation and AI have game-changing implications for the way we design the next generation of vessels and the way these vessels will behave and interact at sea. Ships will be able to gather data from the ocean by themselves, providing humans with critical information we need to address problems like global warming, ocean pollution, and our impact on marine life.

For instance, the Mayflower 400 can sample ocean water from microplastics and record audio of whale vocalizations. Taking the human factor out of a ship allows us to explore new designs and functions that haven't been imagined before.

Lauren asked Brett and Don more about this. What are some of the benefits of having an unmanned vessel? Like, how does automation push the boundaries of what we can do out in the ocean? Well, there's a few major variables, right, or a few facets to that. One is you can do some risky things when you don't have the people there, right, because no one's going to be lost at sea. And then the other thing is you can drive costs down. And I mean cost financially, but also environmental cost, right, because you can...

use far less energy to accomplish a similar goal. And then what that allows you to do is have more, right? So instead of, say, having one $50 million or $100 million research ship, which is the kind of numbers you're talking about to take scientists to see, you can have 20 or 30 or 40

million dollar or two million dollar ships that go out and work collaboratively with space-based assets and with one another and collect vast amounts of data from disparate parts of the ocean and then you use that data to create information that informs where you send the man vessel right so that they get the most out of their time at sea so it's about enabling the people it's about leaving the humans to do the uniquely human part which is have the insight and the intuition and

the creativity and so you know that's why it's important and we're gonna see an increasing amount of this and I think it's also important for people to get comfortable with the idea that these things will be roaming around and that it's okay yeah and on an interim basis I mean we're also talking about this same technology that allows a ship to sail autonomously also can be used to assist a human crew now you know basically

Be another set of eyes and ears. Be a watchkeeper for a manned vessel. Right. I want to know more about the AI captain. How did you build it so that it would be comparable to the way a human captain might direct a ship? What we're trying to do is augment the person, right? We're trying to let them...

be more of a person than sort of they don't have to watch the radar. They don't have to watch the cameras, right? The machine can do all that. And then if it can't do something safely, if it can't come to a solution, it can ask a person, send a little text, say, I don't know what to do. And then a person can, in a very calm way with no stress,

tell it what to do. But in the interim, they're doing something more important, like looking at all the information that's being produced by the instruments and having insight. Ever since we started sailing, there's been expectation of how ships interact with each other at sea. They've been codified by the IMO. They're called the Regulations to Prevent Collisions at Sea. We just call them COLREX.

But they're quite nuanced. Like it's not like they're called rules of the road, you know, after like the idea of like cars, but they're, they're much more nuanced than like rules for cars. How you act depends on the type of vessels that are interacting. Like if it's a sailboat or a fishing boat or a container ship or a pleasure craft.

Like imagine if you're driving your car down the road and you're at a stop sign. And then depending whether you could go or not, depending on whether the other car at the stop sign was a truck, a bus, or, you know, or something else. Sure. Like the rules change. Anyway, so that's where humans are really, really good at, is this nuanced understanding of these rules.

Squishy rules. Squishy rules, yeah. And that's where we've done a lot of our work on, is in that area. And that's the hardest part of this whole puzzle, to be honest. Right, right. I wonder if the ship ever got into any sticky situations that the AI captain was able to get it out of. One time we had a sailboat come at us in the night.

head-on reciprocal course no lights on no radar reflector everybody was probably asleep and they just had the autopilot on and um we easily could have speared them or they would have actually hit us because they were in violation of all the various regulations but um but that's common right at sea when you're crossing it's so unlikely it's so fast that you're going to run into somebody

but it happens. So we, you know, the ship took appropriate action and moved so that that wouldn't happen. But it's not like, it seems very dramatic at the moment, but, you know, you see these things coming miles away and it unfolds at like five miles an hour or something, right? So it's, yeah, so it seems more nervous than it is. And I mean, weather was challenging and we had some failures, technical and mechanical failures in the ship that were very, very challenging. But from the AI captain perspective, the only time that we got annoyed was, um,

There was a research ship that shall remain nameless from a university that was coming along and was going to cross in front of us by 10, 12 miles, which is fine. And they were going along, but they clearly saw us on their...

either their radar or their automated identification system which we broadcast and they just at some point turned and came directly at us at an angle that it's the it's the i'm messing with you angle yeah the angle that allows them to maintain right of way but makes it very very difficult to understand their intent and take action so the ship was kind of like if they had persisted it would have ended up kind of going around in circles trying to avoid them but

But fortunately, we had a support boat that was coming out of Halifax to meet it, and it physically got in between the Mayflower and this research boat and said, what are you doing? Oh, we were just going to take a look, but we weren't going to get any closer than two miles. And it's like, well, what are you going to see from two miles away? They absolutely were going to come over and take a much closer look because they didn't understand that the vessel existed.

was trying to avoid them. You know, when they see these unmanned systems at sea, they're just dumb robots, right? They just float around with wind or wave power, right? There were a bunch of scientists coming back from like a six-week cruise and they were like, oh, that looks interesting. Let's go take a look. So that was the only thing that was annoying. Other than that, it was getting into and out of port. Getting out of Plymouth was a little challenging. Once we got outside 12 miles, we had a lot of fishing boats to dodge, but that was fine.

And then out in the deep sea, it's mostly the sea that you're concerned with. And the fishing grounds are always the trickiest place because... Yeah, because fishing boats do whatever they want. Yeah, and they're like container ships. They're not going to change course unless they have to. So you can pretty much understand what they're doing. Fishing boats could be going along a nice straight line and then all of a sudden do a 180 or worse, a 90 degree turn.

And they don't care about you and they just expect you to avoid them. And they literally, there's no one in the wheelhouse probably. They're all on the back deck. Yeah, but those are the rules too. We're supposed to avoid them. And so, but what Brett got earlier was things evolve very slowly. Like things don't happen quickly at sea. It's sort of like, okay, there's a ship. It's, you know, it's 20 miles away. I've got a little bit of time to figure out what I'm going to do. You don't ever try to put yourself into a situation where there's a risk of collision.

So you make decisions so you don't put yourself at that risk, right? So I'm not going to cross the street at the busiest place. I'm going to cross it at somewhere safe. Fishing boats, container ships, scientists on a cruise, the vast majority of vessels at sea are still of the not-autonomous variety.

To wrap up their conversation, Lauren asked Brett and Don where the technology they've developed is headed, what it means for the humans who work at sea, and what's next for the two of them. What do you guys think this type of automation means for the future of the maritime industry and people who work in it?

I think first of all, like we mentioned, Brett and I have both worked in the ocean community for decades, our entire careers. Like we have an incredible amount of respect for the people that work in this area.

And this isn't about a replacement technology. It's an augmented... What's the right... How do you say that word? Augmented intelligence. There you go. I mean, look, ships have always been the leading edge of technology in almost every society up until the 20th century where we started into flight. And now they're kind of resurging into really new technological areas. But

The point I'm trying to make is there was a time when there were no propellers. There was a time when there were no rudders, right? It was just sails and steering oars. And then, so it's been this evolution in technology. And ships have always been right at the absolute forefront of it from design and engineering and material science and everything.

You know, we've seen this sort of long evolution of technology, and this is just another thing. So I think you're going to see lots of areas where really smart sort of machine learning models help to improve efficiencies. And so we're at the advent of a new way of thinking about design and implementation of very sophisticated technologies.

solutions that are based in vast amounts of data analytics that are hitherto impossible to address. What is next for the Mayflower autonomous ship? We may do a few things with the Coast Guard, and there's a few other folks that want us to do some

work on national marine sanctuaries, looking at cetacean populations. So we'll do that kind of thing with it, and more and more people will get involved in its day-to-day operation, and we'll have less sort of day-to-day input, which is fine. And then the AI captain is going into a whole bunch of other projects and programs, and we're just starting off on a new design for a much larger ship for vast oceanic voyages, so...

Maybe even a circumnavigation. Wow. That's quite an effort.

Yeah, and then we're going to connect with NASA, with the International Space Station and satellite networks and sort of have them work collaboratively. So the space assets see things and they know there's another ship asset. So it's almost like a satellite in reverse. It's like the inverse satellite at sea. So it sees something from space and it says, oh, a ship such and such is over there. Ask it to go and look at that and tell us if what we're seeing is right or collect a sample. And those things will work quite collaboratively without people.

You kind of opened up Pandora's box here. So we did this and now there's all these other things that we can do. So, yeah. And we just have to pick one that we can do within the remainder of our lifetime. There you go. Well, I hope you I hope you both get to do all the new things that you want and have capacity to do. Thank you both so much for your time and good luck with future journeys and projects. Thank you. Bye, everybody.

In the centuries-long evolution of maritime technology, the Mayflower autonomous ship represents an inflection point. The ship's success indicates that artificial intelligence and automation are tools ready to be normalized within the nautical industry, and that the advantages they provide will change the way we conceive of shipbuilding. But the technology aboard the Mayflower 400 has implications beyond just application at sea.

Brett and Dawn's project has shown that the potential reward for innovative risk-taking is to achieve something unprecedented. And that's true for any industry. But like the original Mayflower voyage 400 years ago, it may require a leap of faith.

On the next episode of Smart Talks with IBM, what does it take to create a sustainability-focused global supply chain, innovative and equitable enough to connect our modern world? We talk with Sherry Heinisch, IBM's Global Sustainability Services Leader and Offering Leader for Sustainable Supply Chain.

Smart Talks with IBM is produced by Molly Socha, David Jha, Royston Reserve, Matt Romano, and Edith Rusillo, with Jacob Goldstein. Our engineers are Jason Gambrell, Sarah Bruguere, and Ben Tolliday. Theme song by Gramascope. Special thanks to Carly Migliore, Andy Kelly, Kathy Callahan, and the 8 Bar and IBM teams, as well as the Pushkin Marketing team.

Smart Talks with IBM is a production of Pushkin Industries and iHeartMedia. To find more Pushkin podcasts, listen to the iHeartRadio app, Apple Podcasts, or wherever you listen to podcasts. I'm Malcolm Gladwell. This is a paid advertisement from IBM.