S01E07: The future of water is digital, with Sam Yenamandra from Murrumbidgee Irrigation

In today’s episode, we get a sense of what it’s like to manage a water system here in Australia with the help of our guest, Sam Yenamandra, the manager of Asset Performance at Murrumbidgee Irrigation.

Murrumbidgee Irrigation has been going through something like a technological revolution for the past 20 years — driven by the need to deliver water more efficiently and reliably, and provide greater flexibility to customers they serve.

As you’ll hear from Sam, data is at the heart of this revolution. And what they’re doing now is only the start of a massive – and global – change in the way we feed our planet.

This episode was put together by Hannah Feldman and Joseph Guillaume, both of whom are part of the ANU Institute for Water Futures. The ANU Institute for Water Futures collaboration includes the ANU School of Cybernetics and ANU Fenner School of Environment and Society, both of which are involved in Algorithmic Futures Policy Lab.

Listen and subscribe on Apple Podcasts, iHeartRadio, PocketCasts, Spotify and more. Five star ratings and positive reviews on Apple Podcasts help us get the word out, so please, if you enjoy this episode, please share with others and consider leaving us a rating!

The Algorithmic Futures Policy Lab is made possible with the support of the Erasmus+ Programme of the European Union. The content on this website and any material herein reflects only the author’s view. The Education, Audiovisual and Culture Executive Agency and the European Commission are not responsible for any use that may be made of the information it contains.

Episode credits

Hosts: Hannah Feldman, Joseph Guillaume, Zena Assaad, Liz Williams

Producers: Hannah Feldman, Joseph Guillaume, Elizabeth Williams

Sound editors: Hannah Feldman, Cyril Burchard

Thanks to Nicolas Paget from CIRAD for helpful feedback on the narrative.

Episode Narrative

Liz: Hi everyone, I’m Liz Williams.

Zena: And I’m Zena Assaad.

And this is the Algorithmic Futures podcast. 

Liz: Join us, as we talk to technology creators, regulators and dreamers from around the world to learn how complex technologies may shape our environment and societies in the years to come.

…

Zena: In today’s episode, we’re going to give you a glimpse of how technology is changing the future of agriculture. We’ll start our exploration here in Australia, with water.

Water is, of course, the lifeblood of our agricultural system, and is also a highly contested asset on a continent like ours, where drought is a frequent and sometimes prolonged occurrence.

Liz: We’ll get a sense of what it’s like to manage a water system here in Australia with the help of our guest, Sam Yenamandra, the manager of Asset Performance at Murrumbidgee Irrigation.

Murrumbidgee Irrigation serves over 2,300 customers over an area of nearly 380 thousand hectares in the Murray-Darling basin. They have also been going through something like a technological revolution for the past 20 years — driven by the need to deliver water more efficiently and reliably, and provide greater flexibility to customers they serve.

Delivering water in the best way possible is a key part of the story to help farmers deal with the risks brought on by climate change and boost the resilience of supply networks that underpin intensive agriculture.

As you’ll hear from Sam, data is at the heart of this revolution. And what they’re doing now is only the start of a massive – and global – change in the way we feed our planet.

Zena: Before we have Sam join us, I’d like to introduce a couple of new voices. Hannah Feldman:

Hannah: Hi, everyone!

Zena: And Joseph Guillaume:

Joseph: Hello!

Liz: Hannah and Joseph are both part of the Australian National University Institute for Water Futures. Hannah is based in the ANU School of Cybernetics and Joseph is in the ANU Fenner School of Environment and Society. We’ve invited them to put together this episode, so they’ll be taking over the narration shortly. First, though, I asked them to share what brought them both to the Institute for Water Futures.

Hannah: My pathway to the Institute for Water futures has been a really varied one. I started many years ago in astrophysics, working with many scientists on some of the world’s largest radio telescope projects in Western Australia, but as a young scientist, I realized it was truly the amazing people in astronomy – as well as the stories that they had to tell – that were more interesting to me than really looking at the stars. I changed pathways and I did a Masters in Science Communication instead. For many years, I worked with non-urban remote schools as well as the teachers and high school students exploring what science, technology and design might look like for them in a changing future. Eventually I came back to university life and I completed a PhD in Science Communication. In my PhD, I specifically looked at how young people engaged with climate change, as well as the politics surrounding it, and really – what happens when politicians tell them to go back to school. These days, I bring intergenerational perspectives to water and technology issues – a Venn diagram that’s only growing more complex in a fast-changing climate, society, and technology space.

Joseph: And mine started in computer science and environmental modelling. I’ve always been fascinated by simulation games and decision making. Since my PhD, I’ve been focused on better understanding and improving how we manage uncertainty.

I started off looking at uncertainty in groundwater management, with a case study set among the vineyards of McLaren Vale in South Australia.

I then spent 5 years in Finland working on water scarcity from a global perspective, with idyllic breaks to pick blueberries and mushrooms and go swimming in lakes. I came back to Australia to further explore how we prioritise efforts to address uncertainty.

Hannah: Joseph, I’m so excited to be hosting with you today. One of the things I love so much about being with the Institute for Water Futures is all of the varied, amazing experiences people have had on their pathway into water, and I think today’s episode is such a nice example of three of those pathways converging. So I’m super stoked to be here. Why don’t we kick off today by hearing a little bit about how you got into water research?

Joseph: So I rather fell into the topic of water. When I started my PhD there was an opportunity to look at the topic of groundwater management and the role of uncertainty within this. And a great part of what I’ve enjoyed in this work is that it’s action research, working with real world problems, with real people, with communities, farmers, governments, with workshops and interviews and so on. I’ve had the chance to work in a variety of contexts looking at how we manage uncertainty and how we prioritise efforts to address uncertainty. This is a topic that’s very important to me in terms of understanding how people currently do their work, how research can contribute to the problems that they’re facing and helping them think about how they want to improve.

My connection with Murrumbidgee Irrigation really took hold through a common interest in digital twinning – that’s the idea that we can build a comprehensive, digital view of a system that can use to progressively improve our understanding over time.

The resulting digital twin is a tool that can help us reason not just about how well the system is doing, but also about what information and analytics will be most useful.

Sam’s Asset Performance team has proved to be the natural home for the digital twinning process within Murrumbidgee Irrigation.

As we’ll discuss later in this episode, their work is focused on understanding how well the whole system is working, and developing the tools needed for operations and management to make decisions.

In addition to working on specific projects together, I’ve greatly enjoyed joining the Asset Performance team’s weekly meetings for over a year now, and I’ve had the pleasure to get to know Sam, his work, and his story.

Let’s start from the beginning.

Sam: I was born in Bangalore, India. So a city in south India. Got a lot of representation from global tech companies, bustling ecosystem. It’s called the Silicon Valley of India, and it was a really, really good learning environment for me growing up.

I was also very fortunate to have had access to world class education in Bangalore. With that as well, it was a very competitive setting. It was almost like an intellectual crucible. In that, I suppose, the biggest gift that I was given was the freedom to really pursue stuff that interested me, topics, subjects and whatnot without too much weightage given to how I scored in standardized testing and whatnot. So it was a really good environment to go and learn about something, become proficient about something, put that on the shelf and go find something else to go and learn about.

I’ve never had that real strong desire to specialize in any one area, I suppose. That also goes to the way I’ve seemed to have gone through my academic career and the way I seem to be doing my professional career as well. There’s also the impact piece as well of whatever we do, from a very early age, it was said that, hey, you’ve got to use what you’re given to leave the world a little bit better than when you found it. You don’t need to win a Nobel prize, but hey, do your part. That stuck with me and I think it’s a big part of who I am as well. So although these days, I would kind of balk at someone who says I’m a specialist. I suppose you could accurately say I’m a specialist at being a generalist. And that’s me in a bit of a nutshell.

The other thing is I read a lot in my spare time, and it’s pretty much reading to challenge existing paradigms and my existing worldview. So I really try to read stuff that’s in my blind spot because I like being surprised and I like figuring out that there’s something that, oh, wait, I don’t know that as well as I could know and whatnot. So I think that’s another big influence as well. So in terms of early influences, that really shaped the path for me.

Joseph: What path led you to Murrumbidgee Irrigation?

Sam: So I spent quite a few years working in a variety of tactical and strategic roles in a variety of different industries. Immediately before moving to Murrumbidgee Irrigation, I was in hydroelectricity. It was a fantastic opportunity to learn from some really clever peers on some really cool machines and really develop my skills as an engineer. Towards the end, we just started a family and my wife and I sat down, and we were like we can’t live as remote as we were back then. So we had to move somewhere. What really brought me to MI was there were a couple of things. The first was the caliber of the leadership team that I met with during my first interactions with the company. Right off the bat, I knew hey, I could learn a thing or two from these people in very different domains and stuff that I’m not good at. It seemed very interesting and I knew I’d learn and grow here.

The second was it was amazing how little I knew about modernized networked irrigation. All this investment from the Australian government in the sector, and I knew nothing about it, I must confess, before actually first meeting with them and preparing for the interview then. What really struck me was this is a sector that’s sort of ripe for a bit of disruption because it doesn’t take the front pages in anything. Water’s so critical to everything we do, but arguably, at the time, I was amazed at how little I knew about the whole sector. So those were the two reasons that pulled me to MI, and I’ve largely stayed here for the last three years.

Hannah: Can you tell us a little bit about basically what Murrumbidgee Irrigation does and what’s your team’s role within that as well?

Sam: Murrumbidgee Irrigation exists to enable regional productivity through water and how we do that is really to deliver the water in the best way possible. There’s three overarching pillars that help us accomplish our core function. We work as one team to meet our customers’ needs. We try to stay ahead of the game by using all this modernization to drive us forward, and we challenge ourselves to do different and really challenge our ways of working.

Hannah: Sam leads the asset performance team. Its core business is in technological transformation; they develop the data-driven tools and services that help Murrumbidgee Irrigation, or MI, as it’s often called, accomplish its core role – not just now, but into the future. We asked Sam to tell us a little bit about the Murrumbidgee Irrigation Modernisation program – a project Sam’s team is involved in that is meant to help Murrumbidgee Irrigation build for that future.

Sam: So we’re 100 years old.We were conceived at a time when big infrastructure was the name of the game, and our infrastructure was built remarkably well. It’s largely weathered the test of time, but they are aging and they were aging in 2009 when the organization first started developing a modernization plan. It coincided with, really, investments by the Australian government in improving water efficiency to help mitigate the impacts of the Murray-Darling Basin plan. So that’s the genesis really of our whole modernization program. We’ve largely carried through from 2009 with multiple rounds of investment by the government, and we’ve progressively modernized big chunks of our network. Currently, 80% of our network is fully automated. We have another program that we’ve just kicked off to take that to 96% of the total network. So we’re very much pushing the limits of full automation with the modernization that we started in 2009.

Hannah: Can you tell us a little bit more about this? At a high level, step us through what that automation looks like – how does automation work across the network and what data is collected, what do you use? And can you give us a little bit more on how expansive this system is?

Sam: Yeah, so I think it would really help starting with scale and context of how big we’re talking. We have over 3000 kilometers of channels, over 1700 regulators. So a regulator is any device that regulates the flow of water to meet objectives. Over 4000 customer outlets; so lots of customers. We’ve got 50 pump stations of varying sizes, and we have 15 node towers that really form the backbone of our communications infrastructure. So that’s, in a nutshell, the scale of the system.

In high allocation area, we can easily deliver 800 gigaliters through our network. For context, that’s roughly 70 to 80% of the required annual release of the Snowy scheme on the Murrumbidgee River catchment area. So we’re talking large volumes of water. The quirk of automation, and this goes to your question about how does it work, is it’s a series of networked regulators that talk to each other. It’s a reactive system in that you notify the system that you are going to take water, and it basically propagates that all the way to the topmost point in the system.

So you could be effectively hundreds of kilometers away from the top of the system, but your order gets propagated all the way to the top. And the automated infrastructure does that in a way that maintains water levels so that all the other customers between the person taking the water at the bottom and the top of the system stay within limits.

[musical interlude]

Joseph: Let’s pause for a moment and process what Sam’s just told us.

Their network has physical components – things to help water flow or stop, based on who needs water when. Each of those physical components is networked, which means they have the capacity to send and receive data.

And all of those devices are designed to act on that data to collectively manage a network that spans 3000 kilometers.

Hannah: Consider the scale of their network. A 3000-kilometer long network, if stretched into a single channel, covers more than the flight distance between Melbourne and Perth. Or more than the distance between London and Athens – and back again. And most of it is now managed by automation. Think about small computers, batteries, and solar panels at every regulator on the network, constantly talking to each other and constantly talking to the base stations via wireless networks, with the code needed to respond to data, raise alarms, process instructions, and provide specific information about what the equipment is doing and what’s going on in its environment.

Sam: So it’s quite remarkable. It’s sort of a huge coordination exercise that’s engineered into this modernized infrastructure. That’s how it functions. Now as to what data it generates, to do this, it’s continuously generating flow rates and levels and a bunch of other diagnostic data at each of these regulator points. To do the core function of delivering water, you don’t need that many data points, but you need them in a regular fashion, and one that the controller can use to respond fast enough to a change or an order or whatnot. So it generates a lot of data, but if you look at it, not a lot is required to move water from point A to point B.

Joseph: Murrumbidgee Irrigation is only one of the irrigation companies drawing water from the Murray-Darling River basin. This area, which provides over 2/3rds of Australia’s irrigation water, is incredibly important to Australia’s agricultural system.

Here, Sam tells us about how the Murrumbidgee Irrigation system he just described connects to the larger water system in New South Wales.

Sam: To give you an overview of how we operate with the rest of the river, you’ve got a series of dams. They’ve all got target storages and they’ve got certain capacities. There’s some smarts that go into working out how that translates, along with weather data into annual allocations every year for everyone with an entitlement. And depending on where you are in the catchment, you get a very good idea of how much water people are entitled to take. You’ve then got that translating into some form of demand. So in customers with the intention to plant things and grow things, you can aggregate that across the entire network to get an understanding of how much water you would need to basically get from the river to supply everyone.

We work closely with WaterNSW to put in orders seven days out from when we need them. They coordinate bulk movement of water across the state to our off take points in the network. So really, the management of the dam, the airspace, the capacity allocations are sort of extra water that may come in with a lot of rain, all of that is coordinated by WaterNSW and we work with them to get our water orders in so that we can get water for our customers. The other thing that happens is there can be supplementary events on the river when there’s surplus water. That’s usually announced by another organization, and again, we work with them to deliver to our customers. So it’s really about us ordering enough water to cover our customers needs, and the bulk river operations is predominantly handled by another agency, which is WaterNSW.

Joseph: We now have a sense of the current state of the Murrumbidgee Irrigation system, and how it operates within the larger New South Wales water system.

This is a region that has faced cuts in water allocations, and is facing increased temperatures and less water in spring in the future.

Water prices have increased, too, so it’s more and more important to make sure every drop is being productively used, whether it’s for agriculture or the environment.

At the same time, the Australian government is looking for improved on-farm productivity with a target of a farm-gate value of 100 Billion Australian dollars by 2030.

Hannah: So, like many other places, farmers often face increasing costs but decreasing income, so there’s a constant need to find ways to make life easier and do more with less. For example, allowing farmers to order water on short timeframes even though it takes seven days for water to arrive from upstream dams. Changing crops and irrigation practices have also increased the need for higher flow rates, to be able to get water onto crops more quickly than before.

Joseph: In this context, it’s worth comparing the new system to how an irrigation network like this worked before automation became the norm.

We asked Sam to help us understand what the Murrumbidgee Irrigation network looked like before its digital transformation.

Sam: Prior to automation, everything was manual. It really relied on coordination between customers, people within the organization and the person who drove out to each of these regulators and manually opened and closed them. We were operating 1700 regulators entirely manually. It was amazing that it worked for so long, but in the 21st century, I think automation is definitely the way forward because of all the benefits that you get from it.

Hannah: Can you paint me a bit of a picture of what these regulators look like? When you say they were operated manually, what would I be doing if I was out there operating them? What do they look like? What do they feel like?

Sam: So full disclosure, I’ve never had to manually operate a gate before.

Hannah: Understood.

Sam: But essentially, they’d be predominantly hand wheels on… Just spindles. You turn a hand wheel and there’s a door that moves up and down. That’s one form of gate. There were a few others, but broadly speaking, it was just basically a combination of people turning hand wheels and cranking things to get them to open and close. Or sometimes, maybe manually putting in boards. So moving boards up and down to get the levels and measuring it with something they call a water wheel.

Hannah: Managing this manually was a massive human endeavor. Adjusting this system manually would have required an extensive network of people, all tasked with the job of keeping the water flowing where it was most needed. This wasn’t just about manually moving doors up and down, or shifting boards to redirect water. It was also about making sure the network was working as it should be. This would have been done by human observation before – someone tasked with changing a regulator picking up the need for a repair, or a farmer calling up and complaining that water they were meant to receive simply wasn’t there.

[musical interlude]

Joseph: The ways in which the system can fail – problems that affect Murrumbidgee Irrigation’s network — are called failure modes, and with the modernization program Sam and his team are involved in, the way they detect and manage these are also changing. All the data coming from across the network comes into databases, and analysis software running in server farms help make sense of the data to detect issues and help inform maintenance and planning decisions.

Developing new algorithms and processes is part of what MI has been working on with researchers – including me and my team at the Australian National University.

We asked Sam to tell us more about what failure modes are and why they’re important.

Sam: Ah, Joseph, it sounds like a question you asked me a while ago, what’s this failure mode you’re talking about? Essentially, so all assets have a function. So they have a job to do and just as they have a job to do, they can fail to do their job in a variety of ways. So when an asset are not able to perform its job because of various reasons, we define that as a failure mode. Failure modes, as well, can have multiple or a myriad of root causes associated with them.

So one common failure mode could be that the regulator fails to move, so it fails to perform its core function of actually regulating the flow of water. It could be related to a bunch of hardware and software faults. We have diagnostics to unpack that to a great extent, and where we don’t have diagnostics, we have inspections.

Joseph: So what happens if a regulator fails to do it to job?

Sam: That largely depends on what way that regulator is. So if you have a regulator failure in a site with more than one regulator, can you live with one out of service? If you have a regulator fail on a site with a single regulator, that’s probably DEFCON 1, because you can’t move water then, well you can move water, you just can’t move water in an automated fashion. So you still have fail-safes built into it, but largely it really depends on what’s broken where. Which is another really interesting part of what we do here at MI, is we’re able to kind of unpack that as well to an extent. So not all failures matter equally, and some are more critical than others. And how you find the more critical ones and respond to them in an appropriate fashion and provide certainty that the supply’s going to get to where it needs to be.

Hannah: What Sam’s getting at here is one of the major changes automation has introduced to their system: data. We asked Sam to elaborate on the kinds of data Murrumbidgee Irrigation’s system collects, how that’s currently being used, and how they’re thinking about using it in the future.

Sam: This is the part that’s really interesting for me. As I said, there’s not too many data points that you need to basically move water from point A to point B. We’ve got plethora of instrumentation and other data points that give us all this really wonderful and diagnostic and condition monitoring event data. We’ve also realized that, broadly speaking, automation kind of contextualizes where everything is in the network to a degree. So right off the bat, when you add new data points, it’s already contextualized to what’s already there. So you’re getting a scalable information spine that lets you do things with it.

What we’ve also realized is that we’re able to use automation in ways that was beyond what was originally probably intended for them. So really, some examples are we can put simple calculations are using automated data to work out things like channel condition and channel performance, and start to do some condition monitoring on things that previously you couldn’t condition monitor. Channels don’t have any instrumentation on them. So you’re kind of using regulators feeding channels for diagnostic information. The other thing you can do with this data is things like anomaly detection and better main network maintenance planning, because for the first time you’re able to understand what’s actually going on across your entire network. Whereas in the past you were broadly using first principles or making some conservative estimates. You can actually test those things and build some tools.

So risk management is another big application area that we can use this for. So we’ve got a requirement to meet customers’ supply obligations, and you can use this rich source of data to contextualize the whole risk management framework as well and try to get your bearings on what you should be worried about, what you shouldn’t be worried about. So again, this feeds into our core objective of delivering water in the most efficient way possible.

The other thing that we’ve really had a lot of success with is this vast amount of data. We’ve gone from proving failure modes on one, two, 10 assets to proving failure modes on hundreds and thousands of assets. Because the nature of irrigation is it’s a very modular application — so it’s similar or the same tech multiplied by order, by how many of times you need to scale to actually get water for point A to point B. So that lends itself to really chasing down failure modes across thousands of assets.

We’ve managed to find previously hidden failure modes that we have patched to our network, and it’s got broader applicability to every other utility that uses similar technology. So it’s almost direct applicability because it’s the same tech. So any improvements you make in one spot, you can almost patch for all the other, however many gates you have in the world. So that is something we’re really starting to push towards as well.

Joseph: Sam’s story suggests that their efforts to automate their system are bringing on significant changes – both for the way Murrumbidgee Irrigation works, but also for their customers – who are shareholders in the business.

We asked Sam to tell us about some of the flow-on effects of these changes – both for the organization, and for the communities that Murrumbidgee Irrigation serves.

Sam: The first tranche of impacts have largely been at finding failures before they’ve become too big, or before they become an issue. So the detectability of stuff being broken traditionally, you either have somebody in the field inspecting an asset, finding out, “Hey, that doesn’t look right,” or a customer who says, “Hey, something doesn’t feel right.”

But what we’re finding through the application of these tools and techniques is that we are detecting a lot more of the failures behind the scenes and mitigating them. And failures are, I should clarify, not a bad thing, they’re a reality of life, things fail. And it’s a good thing they fail, it’s even better if you can understand how they fail and figure out how to maybe stop or mitigate them failing. So that’s the big immediate pivot that we’ve seen.

We’ve also had a lot of success with using outlier detection, some AI machine learning on satellite data to work out which meters could be outliers, and which ones could potentially have a fault to them. It’s a very coarse tool, but we’re using the tool to really be your giant rock catcher. So this is a filtration metaphor, it’s we don’t really fine levels of filtration, we just want a big rock catcher to find all the stuff that doesn’t make sense so we can dig deeper and go and find what’s happening. And we’ve had quite a few hits with that.

The other big sort of benefit story has been weeds are a huge problem across our integrated delivery network, they’re also seasonal, so they can flare up, they can sort of die down when the colder months happen. And we have some tools that help us quantify the level of, or the loss of performance in a section of channel due to things like weeds and silt. So silt loading is another unseen, but also a particularly bad root cause for a failure mode.

Hannah: Sam’s talking here about plants and dirt infiltrating their channel network. It’s something an outsider might not consider when they’re envisioning what this network of water channels looks like. But it’s a serious consideration.

Joseph: In a semi-arid climate, sustainable irrigation requires fine grained control over what happens to the water that is taken from the river. We want to deliver the water where it is useful, both for irrigation and for environmental water at the right time and with as little unproductive loss as possible.

Weeds and silt slow down water and decrease capacity. You can’t design these problems out of the system.

So the data they collect are now allowing them to figure out where they need to remove soil and plant matter from their system.

Sam: We have the ability to go and triangulate within a 10, 20 kilometer section of channel, where exactly is this constriction? Because you don’t need the whole channel to be constricted; it’s likely to be just a part of it. So we can triangulate very quickly on a big long channel where exactly it is and go and intervene and improve things.

One of the things we’re trying to scale is this very concept of working out how much performance we lose in a channel across not one or two pools, but hundreds of pools. And we’re currently in the process of scaling that. And we believe that when that’s operational, that will allow us to bundle up the risk and manage it on a much bigger scale; where probably more of the efficiencies lie.

Hannah:  Do you have any specific examples, any stories of triumphs or catastrophic failures that really stick with you based on the implementation of this system?

Sam: There was one particular metered outlet that came up as an outlier in this tool. I know this one because I was in the car with the technician driving around and having a look, and we got to the site and right off the bat you could see… So just for context, the majority of our flow meters are electromagnetic, so they just basically measure velocity of the disturbances in the magnetic flux and then translate that into current, and we get a response. So it showed up as an outlier, we went to site, we had a look, and it looked like there was actually a generator next to the outlet… So the outer casing was blackened because of heat. Now there’s very few ways in which a electromagnetic meter will fail, and one of those ways is if the internal resistance changes somehow. So that can change if you have moisture and dust ingress, or one of the root causes. So we saw the compromised casing and we were like, “Woo.” I don’t … We didn’t have any alarm that would’ve picked that up, but it certainly is an example of if you apply outlier detection well, there’s failure modes you know about, there’s failure modes what you don’t know about, and really what you’re trying to engineer is you’re trying to build a Swiss cheese model of safety. Have you seen the Swiss cheese model of safety?

So what you’re trying to do is just layer it with Swiss cheese and really hope that you’re catching your failing mode one way or another to really manage the risk. Yeah, that’s probably one example.

Hannah: What about any wins, any really good examples that come to mind for places where this has gone well, and you’ve just been like, oh my goodness, this has saved the day, or we’ve done something really cool with this?

Sam: I think the biggest wins have been we had persistent reliability issues with one aspect of our gates, and we were able to establish a hypothesis on what could be happening based on root cause analysis of the failure in question. So we were able to narrow down the field of search, we then ran an analysis on all of our gates in the database. And we were able to conclude, not only did we find evidence to support one of the probable root causes, we were able to detect that that same failure mode was likely to have occurred in multiple other sites that we didn’t know about purely because there was a bug in the software.

So that’s another example of a win. That was a big study, I think about roughly 1500, 1600 sites we were able to analyze the data for, that was a good demonstration of finding something and then seeing where else it existed in scaling that analysis.

Joseph: So what has your experience been with the operations and maintenance teams reactions to this new technology? What has your relationship been from that perspective? How receptive have they been?

Sam: Our operational maintenance teams have been really fantastic with working with us to try to change our approach to using more of the tools and technology. We have a lot of candid discussions on, “Hey, how can we improve? How can we really bed down this continuous improvement and how can we do better?”

It’s about waking up every day and coming in and trying to find that extra step. And to that end, they’ve actually been quite fantastic to work with. They’ve also largely let me work with them to disrupt their workflow. We’re unified in the desire to make improvements, which largely helps. There’s a very cohesive approach, which really speaks to our whole one team culture where we’re all trying to pull in the same direction, ultimately for the benefit of our customers.

And that’s been quite a good learning process as well. Also in terms of skills, I think the biggest skill when you’re talking about this sort of pivot really has to be the getting comfortable with the discomfort of making decisions with incomplete data. So anyone can make decisions on complete data. Formally in my engineering practice, you wouldn’t make a decision unless you were really sure about something. i.e. you’d want complete data.

But what you’re sort of seeing with the proliferation of big data analytical techniques, data science techniques, and whatnot, and stitching it all together, it’s that discomfort with incomplete data, but also having a plan B. I almost frame it as a positive asymmetrical risk. So you lose the risk of managing that failure, it’s small endeavor to manage the risk of that. But if you succeed, there’s a step change improvement.

So really about trying to compensate for the early stage development of these tools by tailoring the final applications in terms of positive asymmetrical risk.

I suppose Joseph and I are constantly dreaming up ways to really challenge ourselves on how we can embed these tools. Because the real benefit of what we are doing is only landed for our customers if we translate the world class innovations that we’re making in the organization into real world impact. So our job is kind not done till we put rubber to the road and we’re able to make these changes that we believe are there.

[musical interlude]

Joseph: Following on from that, if the industry was to adopt these kinds of things more broadly, this is obviously a challenge that everybody’s facing. So if you were to recommend to others, have a look at this technology, what kind of risks would you warn them about that they’ve got to look out for?

Sam: So I can talk to our own learnings from the experience of developing these tools. Broadly, you need to first off acknowledge that there’s a chance that you’re wrong. So the chance of being wrong is always – is real. Then you are really looking to surround yourself with diverse perspectives. I’m talking diversity in like all its forms, skills, knowledge, experience, perspectives, and that really helps you find all the gaps that you’ve missed.

And again, it comes back to really being comfortable in the fact that you’re having to make decisions on incomplete data, and managing the risks that come with it. Just because the data is not perfect, does not mean it’s not useful. And to be able to leverage it, you just need to get quite comfortable with really challenging some established paradigms.

We all love surety. We all love surety, we all love predictability, and unpredictability is really bad, but nothing is 100% predictable, nothing’s 100% certain. So it’s really about managing the uncertainty well enough that you’re able to realize the benefits.

The other thing that we’ve really found is changing the questions you’re asking often leads you to different approaches to solving long entrenched problems. So, you know, you can borrow from different domains of expertise and actually see if that actually is relevant in your own sort of context.

And yeah, so it really… I suppose that those are some of the learnings that we’ve encountered along the way. Obviously depending on who’s attempting a similar exercise, their context will vary slightly, but broadly speaking I think that kind of encapsulates what they could expect.

Hannah: Risks are inherent to any change in a system like this. But they are often necessary to make progress. In water, and in digital agriculture more broadly, the kinds of changes Sam is sharing with us are needed to adapt to our future climate.

Joseph: Climate scientists predict there will be more uncertainty – in the rain, the weather, all the things producers rely on to supply us with what we need to live – and there are more extreme events placing pressure on infrastructure.

There’s a need for change, in other words – one that Sam and people in digital agriculture more broadly are aware of.

Hannah: We asked Sam how he thinks about managing the risks inherent in making changes to Murrumbidgee Irrigation’s approach, especially when implementing something new.

Sam: So coming back to our discussion on failure modes, all our assets have a function to perform, and when they fail to perform that function we jump in and we intervene. But broadly speaking, we’ve also got a robust set of KPIs for each of our asset classes.

KPI setting is an interesting one, you have your outcome KPIs, which broadly determine how well you’re doing, and your process KPIs, which really say what the things you’re doing to drive your outcome KPIs forward. So we’ve actually got a measure of overall health of the process running. So when we make change, we pair that with some robust risk management frameworks; and it’s nothing has really rocket science.

If you look at any process industry with lots of risks, they have really, really good risk managers, they have all these robust processes, standardized ways of implementing change that balances some of the risks associated with doing new things and modifications. So we just seek to where possible borrow from that, build that into our processes, and we have sort of a commissioning period, if you will, when we implement we monitor it and make sure everything is going as expected. And if there’s any sort of changes or something we see that doesn’t make sense, or shouldn’t be happening, we’ve always got a fall back.

We’re not afraid to make changes, but we put a lot of thought into working out what are the risks of making this change? And we really try to do that depending on what the final consequences of the risks are. Any risk management approach is stratified to have more controls the bigger the risk, and when the risk is not huge, you can put in place some rudimentary checks and balances to make sure you’re on the right track. So we’ve really tried to focus on making the changes with the appropriate risk management controls in place.

And that’s largely borrowing from my experience in hazardous process industries, and power generation by definition in Australia is a risk management exercise, you’re constantly trying to manage market risk and have the generation capacity when you need to generate. So I’ve largely been able to borrow from that approach in our work at MI.

Hannah: Sam gave us a good explanation of how Murrumbidgee Irrigation is thinking about risk when implementing changes to its approach from a risk management perspective. But we wanted to know how this approach fits into the broader conversation Murrumbidgee Irrigation is having with its customers about its present day operations and the future that it’s building towards.

Sam: We have a communications team that send our periodic newsletters where we talk about these innovations and keep our customer abreast and different things matter to our customers at different times of the year. So, it’s about trying to put in front of them information that matters to them at that point in time. What we do is we innovate all year round, but what we share with our customer is probably what’s top of their mind. And that’s currently the mechanism that we communicate with our customers with. The whole company is very straightforward with what we find, what we don’t find.

Joseph: Murrumbidgee Irrigation’s customers are farmers. They are also shareholders. We asked Sam: What does success look like for a farmer?

Sam: So, I come back to the core job of MI, which is to deliver water in the best way possible. All of what we’re doing is really feeding into that singular goal. If there are better ways to move water from point A to point B, we want to take it. The ultimate benefits will flow through to our customers who are also our shareholders.

So in terms of the impacts to the end user, the better of a job we can do in finding the best way to move water from point A to point B, be it it’s more efficient, we’re managing the risks of network condition a lot better, we’re managing the risk on their behalf so that they’re always getting the water that they need, and/or we’re managing asset failure risks that could impact their ability to get the water. So everything we do is to try to move and push the needle in that — trying to move water in the best way we possibly can.

Joseph: So if we take a specific example, if a regulator fails, what impact might that have on a farmer?

Sam: If you’re talking about asset failures, and if asset failures coincide with periods where we need to get the water to a farmer, then yes, that’s something that we have to jump straight away.

Hannah: Talk to us about how quickly you can respond and what does response look like?

Sam: Should an asset fail in the field, it’ll generate certain alarms that’ll alert whoever’s first response that, “Hey, I’m broken, come pay attention to me,” depending on what that failure is. If there is a remote solution where we can fortify and restore a site remotely, we can do that. If it’s a failure mode that is unlikely to impact end users, we can come back to it first thing in the morning and just get out there and visit site, fix it. If it’s something that warrants immediate attention, we’ve got a robust on-call framework where we get out to the site, and we’re able to do whatever it’s needed to make the site safe and maybe assure the farmer that they get their water.

Thinking back to that failure mode we talked about where a gate doesn’t move, what that could look like. If the gate doesn’t move, it raises an alarm, first response looks at and says, “Hey, okay. That could be bad. I wonder what the demands are over the next 12 hours.” Or, “Wow, they’re currently watering. What do we do about it?” “Okay. Let’s just go and see if we can move the gate manually, somewhat if we can, depending on what’s actually broken.” So that you can just deliver them the water and maybe revert control to the upstream gate.

So you can reconfigure the system to get around some failure modes and still supply the water. So, there’s a whole triage process that we operate to make sure that where we’re appropriately responding to this high risk events in the field. We have half a million instrumentation points in the field, and we’ve largely gone and understood what warrants the whole, “Let’s send a call out to somebody at night,” versus, “Okay, this is something that can wait, or this is something that is going to fail given time, but it’s not currently a high risk.” So we’ve got those triage processes in place.

Hannah: Cool. So, talking about the kinds of things that might be at the top of your customer’s minds, one issue that’s come up globally with these kinds of automation systems is of course around confidentiality – particularly of farmer information. And I know that it’s still early days, but how do you see information sharing with farmers changing into the future? What I want you to think about here is things like data sovereignty, sharing the cost and benefits of that information, and how they might trust or distrust the system.

Sam: I think it’s a bit premature, as in we haven’t confronted that yet. We are talking about it. And I certainly see that, like with any mature data endeavor that would naturally become something that you have to factor in and account for and manage somehow.

The bulk of our work is asset focused. So assets are beautiful in that there’s no care about bias and they don’t really care about, “Oh, you’re giving that regulator more attention than me.” And so, it’s largely kind of very immune to an extent to any sort of biases. And crucially, whatever we find, we make a lot of improvements internally to the ways we do work, our processes, we help our key technology partners make improvements to their products for the benefit of our customers. We try to champion that whole relationship in the discussion as well.

So, the honest answer to your question, Hannah is I don’t know because it’s very early days currently, and I wouldn’t want to speculate either because I tend to have some fairly left field ideas for things.

Joseph: So there is a follow-up question there, though, which is that you have been very careful this whole interview of not revealing certain confidential information about particular cases and that kind of thing. So, what sensitivities are there around the types of information you gain about farmers? You don’t have to be specific just in terms of general, what are the reasons that you keep things confidential?

Sam: There’s currently not a real demand for the data we are generating outside of the organization. I think farmers are really focused on doing what they do really well, which is trying to grow things in the best way they can. So, aside from ad hoc investigations, where we work with customers to say, “Hey, this is what we found. And this is what we’ve discovered,” there is no organizational-wide methodology for broad dissemination of data. We do publish fact sheets and other information for our farmers to go and access on our internet. But the honest answer is we probably don’t do enough of it because it’s still early days.

And it has to be an organizational-wide response to this because this is a big problem. And this is a problem that everyone is grappling with. And I have full confidence that in time, we will find an approach that works for our customers. Right now, I don’t have an answer to your question, unfortunately.

Joseph: It’s clear that there are no easy answers to these questions. People are grappling with these issues in many places, and sectors throughout the world. Sam shared with us a sense of why Australia’s an interesting place to be thinking about some of these issues as they relate to digital agriculture.

There’s a sense that there’s an opportunity to help shape the future in a socially responsible way – not just for Australia, but more broadly.

Sam: Australia has one of the highest rates of modernization in agriculture in the world. And we’ve also got some free trade agreements where the geographical arbitrage is not a thing anymore. So, really about how do you extract the value from the data that’s already latent, how do you unlock new data sets and share the value broadly across society? These are big questions. I certainly wouldn’t like to be a politician trying to answer them, but certainly, one thing that is clear though, is a sector-wide approach is something that would be warranted.

So, if you did piecemeal, that would come with the pluses and minuses of a piecemeal approach. If you went big and rather did a nationwide approach, that also carries a single point of failure and that if you get it wrong on the national policy level, that also has inherent risks. But this is a big question.

Hannah: I’m going to poke you, Sam. I’m going to get you to speculate.

Sam: Oh, don’t get me to speculate.

Hannah: I’m going to go completely off script now. But with all of this in mind, we know that water planning has had a bit of a rocky history in this country. We know that this agricultural revolution is going to make really big cultural shifts, whether that’s at the policy level, at the user level, both. I want you to picture 50 years from now, what does it look like? What are the key differences between, say 10 years ago and 50 years from now, that you see emerging? What does your utopian agricultural revolution look like?

Sam: There is 100% chance that this is 50% correct. So, what we’ve noticed in the last 10 years, for example, the innovative work we do here at MI largely would’ve been much harder if we had tried to do it 10 years ago, the explosion of data in the consumer realm has really far outpaced anything in an industrial setting. I believe that’s largely to do with the industrial processes. You need to have a good causal handle on if this does this, this does that. Then that leads to this whole concept of data density, so you don’t really have as much data points that you need to manage.

Whereas the consumer realm has just kind of exploded with data. Now, extrapolating from that to the next 50 years, you’d almost have to think that due to the rise of quantum computing, vastly increased amounts of processing power, bigger data than we currently have. So, I think, the modeling and the simulation aspects of water would be a lot more accessible to a lot more people, a lot more easily.

Hannah: And what would that enable?

Sam: It depends on the intent. It would either enable large scale optimization of a finite resource that we all need, or it would lead to a fractured approach that has the winners and the losers. And mind you, we’re talking 50 years out, so this is just what Sam thinks. Again, no guarantees for accuracy or all the other usual disclaimers that you get.

Hannah: Well, we won’t come back for you, Sam, in 50 years time and hold you accountable. Don’t worry. But this is fantastic to hear, your thoughts. So, continue on this train.

Sam: So broadly speaking, compute might not be an issue anymore. Who knows what sort of software frameworks we’ll have. We’ll have a proliferation of data. You’re talking geospatial imaging already with a lot of what we are seeing with, low earth orbit satellites. So the data sets will be more and more. We had a very exciting look at some really cool technology, which we thought was a bit too exploratory for us to fully jump in on, which was water-penetrating LiDAR bathymetry. So you’re talking, if you can scan huge swaths of land and get plus or minus 20 mil resolution bathymetry, that’s also a game changer because typically, your channel survey or land survey is quite involved. And for some applications, you need high accuracy, so plus minus mil, or one mil, but for some you don’t. So for simulation modeling, you can get away with coarser data, but you can basically brute force it with more compute.

[musical interlude]

Joseph: Bathymetry is basically a way of creating a topographical map of an underwater landscape. Underwater mapping like this is usually a challenging process to undertake manually.

Hannah: And LiDAR, short for Light Detection And Ranging, is a laser-based remote sensing method used to gather the same kind of data. You basically shoot a special kind of laser down into the water from above and see how long it takes to reflect back. This is called a “time of flight” measurement. Couple this with precise GPS data and you can create a 3-dimensional map of an underwater surface. Or a topological surface on dry land – because with some modifications, you can use the same technique.

When Sam talks about plus or minus 20 mil, or plus or minus 1 mil, he’s talking about the accuracy of the measurement in millimeters. So if your measurement is 300 plus or minus 20 mil, you can be pretty sure that the true depth of your measurement is somewhere between 280 and 320 millimeters. This isn’t very accurate, but might be totally fine for some purposes. Let’s get back to our conversation with Sam.

Joseph: So what involvement do you think farmers, government agencies, others in the community, the developing tech industry — what involvement do you think they’ll need to have to fully realize the benefits?

Sam: Let’s look at formula one — We’ve got the parallel that we need. So in formula one, you have 10 manufacturers, one set of rules, everyone with their own interpretation of things, and they all go off and spend millions of dollars and do lots of engineering analysis and come to the first tests with a bunch of race cars. The amazing thing is on average, they’re only 5% off of each other in raw speed. They’ve taken all these different sort of approaches, but yet they’ve found themselves in a sort of narrow window. And that’s largely a function of the fact that there’s a rule book, which I’ve had the good fortune of reading, but immediately made me realize why I’m not a Formula 1 engineer.

And the rule book is the sort of baseline. And you’re still opening up the space for that innovation to exist. You’re still sort of getting teams coming up with clever concepts that everyone else looks at and says, “Hey, wait a minute that doesn’t look right.” And subsequently the rules are updated.

So this whole concept of having a set of defined rules, but enough flexibility for people and organizations to innovate, that could be a very good working model. The other part of that model is it’s not set and forget. You don’t write a rule book and walk away and 10 years later anything that isn’t looked after turns to chaos. So I mean, there’s constant revisions, constant discussions about, Hey, there’s the legality of certain implementations and approaches and stuff and technical directives and whatnot.

So you really need an involved approach that sort of gives people some background or gives organizations a bit of background whilst also letting them innovate in the space.

Hannah: I love this Formula 1 analogy. I’m wondering what kind of workforce changes you might see in this system based on this. With all of these different developments and changes and so on I figure we are going to have implications for the workforce and how regions are going to be developing and enabled to develop in that area. I’m wondering if you can talk a little bit about what kinds of workforce changes we might see. You’ve touched on this a little bit already — that it used to be a technician that went out to the sites to crank the handles and so on, and now that it’s a computer, no one does that anymore. That’s just one example, but what else are you thinking about when it comes to workforce changes, particularly out in these regional areas?

Sam: Again, it probably is helpful to look at a parallel as well, 70 years ago, actually longer than that possibly when they first conceptualized this whole concept of the Snowy Mountain hydroelectric scheme, there was nothing in the way of meaningful hydro aside from a few pieces here and there. And over the course of building and operating this infrastructure, a lot of the skills that were needed to keep the infrastructure operating were developed and honed and they really thrived.

So part of the discussion around what skills are needed for the transformation in the agriculture space is really around the transformation in the agriculture space almost helps develop the skills that are needed to sustain the transformation in the agriculture space. It’s a cyclical thing. And there are definitely benefits to scaling an initiative as well.

I think in terms of what will become quite important, and if it’s not already becoming important, diversity as a means to manage bias in AI and algorithms and whatnot is already a hot topic. That is going to continue because we don’t see the world as it is. We see the world as we are. And the more diverse your sort of perspectives are that will lead to better outcomes in this data rich environment.

So I think that being said, moving towards landing and leveraging a lot of these systems and trying to do them at meaningful scale is probably going to be the biggest accelerator for skills that we’ve seen because right now everyone’s, I mean, I’ve been quite candid about the fact that I knew very little about networked irrigation until I started here. You’d almost have to expect that would be the same for somebody unfamiliar with the industry.

So I think there’s the whole pivot and drive towards operationalizing a lot of these things will help greatly balance that question because diversity in all its forms is super critical. If you can’t look at the problem in 20 different ways as part of a team, you’re missing all these things that could potentially introduce risk. So I’m not too worried about that because the Universe has a funny way of sorting these things out. And the best way to sort out a deficiency in one form is to actually have a sort of almost a balancing value proposition on the other side. Hey, this is a real problem. You need diversity to solve this problem. So what happens, there’s a push pull lever and it sort of, kind of evens out. And that being said, there’s also other bigger picture discussions that need to be had on a societal level about STEM education.

And I can only speak from my personal example. Here at MI, we’re really challenging ourselves to look at the stereotypes, any job is accessible to anyone. We’ve had lots of really good wins and sort of improving gender diversity in roles that traditionally have not been gender diverse. And that has been a concerted effort, and I’m happy to say we’re trending in the right direction.

Joseph: Diversity can come from the people and organisations one collaborates with, as well. I know from my work with Sam that Murrumbidgee Irrigation actively works with technology companies and researchers like me to drive innovation.

I asked Sam to share why they were taking this approach, and how they were going about it.

Sam: We’re exceptionally fortunate to have a culture at MI where we really challenge ourselves to innovate and continuously improve. We’re very lucky that we have fantastic relationships with the world class research organizations like ANU. And we’ve also enjoyed a really close working relationship with our key technology partners, largely because we try to frame the problem as a win-win. So there are wins for our customers, but equally so, there are wins for their customers too. So this is a journey. It’s not a sum-zero game. We really try to be fair and firm.

With regard to MI, I think we’ve made a lot of investments in data skills and also engineering skills. And it’s not typically the norm in this industry from my observation. I suppose, in the long run successes, it’s really about taking everyone on the journey, finding a good solution to an entrenched problem.

Everyone is part of the solution. So all our partners are part of the solution, so I think that approach certainly helps. But I think the biggest factor would have to be the culture in this organization. I think it comes right from the top. We are always looking for those improvements. Innovation is encouraged and nurtured here. We’re all trying to pull in the direction of adding value to our customers and doing our core job of delivering water in the best way possible. And we largely find that with that mindset, we’re able to tackle problems with novel approaches. So, the palatable solution, the proof’s in the pudding. If you make an improvement, you have justified the approach, and it’s not the other way around. It’s not the approach tail wagging the problem dog. So I think that’s largely the reason why we’ve been able to operate the way we have and why we’ve been able to take everyone on the journey.

Joseph: So, one of the models of innovation that’s often criticized is a linear one. Researchers come up with something. Tech companies implement it. And MI is the customer, they just take what is given to them. What’s your reaction to that kind of model?

Sam: It is an interesting model in that you look at digital technologies and big data – that’s been around for what? 20 years? Not even?

When you really reflect on that old all saying, that there’s nothing new under the sun, I can draw an example from a previous life where I was working on this Japanese hydro turbine, built in the 70s, and we were having some trouble with it. We had been operating it for 50 odd years. We were having trouble with it. We went back to the manufacturer and we said, “Hey, we are having all these things. This is what we’re noticing. What’s happening here?” And the manufacturer just basically told us, “Hey, you know more about this than we do. So, good luck.”

And I feel like the whole discussion, that traditional innovate tech supplier builds and deploys, and then you take what you’re given, it robs people and organizations of the agency, and given enough time, that’ll become quite clear. When you’re maintaining plants that are 50 years old, you realize you’re it, and you have a lot more say than you think. And given that modernization really kick started in 2009, maybe a few years earlier in some places in Australia, I think it’s early days.

[musical interlude]

Joseph: Sometimes, we hear a different story: a dystopia, where all the technology is owned by a few large corporations. They own the tractor, the irrigation system, the robot pickers. And everything is operated remotely. When there’s a problem, you call someone (or some chatbot) overseas – or in some big city somewhere that’s never even been near a farm. This is not the story we hear from Sam. We asked him to share his source of optimism.

Sam: I suppose my optimism comes from the fact that I believe people are capable of great, good, if their motivations and intentions are in the right place, of course people are also capable of terrible things. But we all have the human ability to sort of move towards something that’s better than what we had before. And it’s great that there’s so many entrepreneurs out in the world that sort of seek to do the whole for social good piece.

So I suppose really my optimism comes from understanding how people can actually really cause disruptions and disruptive innovation, and also try to use that to further things in a way that leaves society better off than when they found it. Of course, there’s the other side it, but you can only control what’s within your circle of control. So that’s just my perspective, Joseph – I don’t know if I answered your question.

Hannah: And finally, we asked Sam what the term socially responsible algorithm means to him.

Sam: I must admit this question. I had to reflect on quite a lot, and I think it all comes down to principles. I asked myself what are some of these principles that could apply? The first one is bias of really a bad thing? Nothing is without bias. So if that’s the case, then why not have bias with the maximum social impact? I don’t know, that’s a foundation for a lot of the whole mindfulness practices and stuff, really about how you can sort of bias yourself in a way for positive change. And surely the same has got to be surely the same can be said for design of algorithms and whatnot.

There’s also got to be an equal care and we pay a lot of attention to building them, but do we pay a lot of attention to using them? And using them in a way that balances some of the risks of the self-learning algorithms that can get it wrong.

Speaking of getting it wrong, we also really need to acknowledge that all algorithms are wrong, some are useful. This is that whole play on words for the model space. And if that’s the case, then you can almost manage the risk by having systems in place to balance, vetting and voting and whatnot. There’s a reason why critical components in a nuclear reactor are triplicated, because stuff will break. And the biggest thing is by expecting stuff to break, you kind of build a certain level of error proofing for these algorithms. And it’s also about redefining the problem and applying the right tool to the problem. It’s very tempting to say, “We’ll just apply black box solution to this” and just walk away. But that’s also fraught with risk. I mean, if you don’t know what you’re applying, it’s like saying, “I’m going to press that red button. I’m not sure what that button quite does, but there’s a 50% chance that it’s not that bad.”

So it’s almost like you need to know what you’re getting yourself into. And part of that is about what is the problem you’re trying to get this algorithm to solve? We’re not at that phase where we don’t have Jarvis yet. We don’t have all the the sci-fi writers predicting all this really cool self-aware type computer platforms yet. But until that time, let’s maybe define the problem a lot better and find the right tools for the solution.

That being said, it’s interesting with this tool analogy, because Adam Savage has written a wonderful book called Every Tool is a Hammer. And for whoever’s listening to this podcast, this is a spoiler. So you might want to pause, but the whole premise of the book is that there is more than the intended use for something, for a particular tool. Quite literally the innovative work we are doing here MI is really a testament to that whole mindset. But algorithms are not immune to this. You build an algorithm for one thing and you’ll have other applications, which you may or may not know about, which may or may not be introducing risks from the whole social responsibility piece. So, that’s just my take on it. Of course there’s a very good chance I’m incorrect. But I suppose if we are allowed to speculate on how we can build a better future, that would be what I would say.

Joseph: Sam has described an industry undergoing a major transformation, underpinned by data.

The focus is on delivering water more efficiently and reliably, and with greater flexibility for customers, but the big picture involves organisational, cultural, and even societal change.

What is at stake is the capacity to deal with the risks brought on by climate change and the improve the resilience of supply networks that underpin intensive agriculture.

This podcast has focused on Murrumbidgee Irrigation’s experience, but the trend is for this technology and supporting algorithms to scale out and up. As more irrigation districts adopt this approach, new data and experiences will pave the way for improved analytics, models and algorithms. They will also help identify further issues and possible solutions.

Scaling up, there are already steps being taken to apply these approaches at a state or national level, aiming for a water grid that delivers water in the best way possible. This requires bridging systems across organisations both at a technical level and in terms of how decisions are made.

While inequality of access to technology, internet, and technical expertise remains an issue even in Australia, there is a worldwide trend towards adopting new measurement tools and digital apps.

Digitalization of agriculture is recognised as an agricultural revolution, with a rich array of opportunities for countries to show leadership, and to experiment with managing risks and how best to seize opportunities.

Hannah: For social responsibility of algorithms, Sam has painted an optimistic picture, while recognising the importance of a really considered approach, complete with checks and balances — knowing what you are getting into, and expecting things to break.

Algorithms and automation will continue to play a large role in decision making and control of these systems. As they progress, the problem of alignment of objectives becomes more prominent, with the potential for data to be used or abused for unintended purposes. Some farmers might be favoured over others. A minority might be able to increasingly impose on others their own vision of the future (or even for our handle on the future to get away from us).

The idea of positive bias and keeping humans in the loop with checks and balances provides some of this guidance that we’re looking for, but fundamentally in a context where data and algorithms can be used for multiple purposes, with changing objectives all the time, the challenge is in how to involve all stakeholders, young and old, in shaping the system over time.

It seems fair to say that when it comes to digitalisation of agriculture, transformation is happening – kind of whether we like it or not. The question now is how to make sure we can collectively construct a future we can all look forward to.

…

Liz: Thank you for joining us today on the algorithmic futures podcast. To learn more about the podcast, the Social Responsibility of Algorithms workshop series and our guests you can visit our website algorithmicfutures.org. And if you’ve enjoyed this, please like and share the podcast with others.

Now, to end with a couple of disclaimers.

All information we present here is for your education and enjoyment and should not be taken as advice specific to your situation.

This podcast episode was created in support of the Algorithmic Futures Policy Lab – a collaboration between the Australian National University School of Cybernetics, ANU Fenner School of Environment and Society, ANU Centre for European Studies, CNRS LAMSADE and DIMACS at Rutgers University. The Algorithmic Futures Policy Lab receives the support of the Erasmus+ Programme of the European Union. The European Commission support for the Algorithmic Futures Policy Lab does not constitute an endorsement of this podcast episode’s contents, which reflect the views only of the speakers, and the Commission cannot be held responsible for any use which may be made of the information contained therein.

[sound of water trickling]