In early January 1697 a group of Dutch explorers led by Willem de Vlamingh, rowed up the Swan River in Western Australia and found birds that entirely changed the European perspective on certainty and the use of a 1,600 year old expression.

What they had found was the black swan. For over a millennium, Europeans had been so certain about swans being white that the black swan was seen as a way of expressing impossibility. For us, it’s equivalent of going to Mars and finding flying pigs, hens with teeth or perhaps the needle in a haystack.

The association of the black swan with impossibility most likely comes from its mention by the Roman poet Juvenal in his Satires from around 100 AD. The excerpt sits inside Satire VI, a long and fairly offensive (or sarcastic depending on how you read it) tirade against marriage and the difficulty in finding a perfect wife. If we put the misogyny of a Roman poet to one side, what’s notable is that he used the black swan metaphor precisely because it was the most confident example of impossibility his audience would recognise. Every Roman knew there were no black swans, so it created a clear image in readers’ minds.

So when the Dutch expedition sailed up a ‘brackish inlet’ in a Western Australian river in 1697 and found it full of black swans, they ruined a metaphor that had been in use for sixteen centuries. (Somewhat annoyingly for the story, the black swan was probably first seen by Europeans in 1636 when Dutch explorer Antonie Caen in the Banda landed at Bernier Island but ignore that, because it was de Vlamingh’s descriptions and samples sent back that convinced people of their existence)

I grew up along this same river that de Vlamingh visited some 330 years ago and went past black swans daily, so my experience is almost the opposite. I assumed default swans were black, and that white swans were strange things of European fairytales and operas. The first time I saw a white swan on a trip to Europe it was quite a memorable experience; I still take photos of Danish svaner to send to friends and family back home.

And for the Whadjuk people of Western Australia who are the first people of the river and who have lived along the Derbarl Yerrigan for tens of thousands of years, I imagine black swans would very much have been the default. The dreamtime stories about the river and its creation by the Waugal (rainbow serpent) are fascinating. Perhaps there is a lesson here in perspectives on impossibility, lived experience and assumptions, but that’s for another day.

As an aside, if you ever travel to WA, you can still see inscriptions and shipwrecks associated with those early Dutch expeditions, over 80 years prior to the French (1772 - Louis de Saint Aloüarn) and 130 years before the first sanctioned English expeditions (Dampier in 1699, Lockyer - 1826). The Dirk Hartog inscription is now in the Rijksmuseum in Amsterdam as the earliest record of Europeans in Australia, but I digress.

So what does this mean for agtech? For me, this story is about two things:

Predicting the future is difficult, like finding hens teeth, but more importantly, I think, is how we design systems that enable us to minimise the risk that exceptional changes bring; and also to exploit any opportunities that may occur.

This is what I want to dive into in OSA #6, and how open source development for agtech is a critical black swan avoidance tool (they hiss and can bite after all).

Welcome back to the 6th OpenSourceAg newsletter, this time looking at our ability to predict the future. Ironically, I was planning on writing something else this edition, but as we all know det er svært at spå, især om fremtiden (it’s hard to predict, especially about the future - a quote often attributed to Danish writer Storm P). So here we are. Forgive all the Danish references too - Danish lessons are back in full swing.

Thank you as always for reading this and contributing with your emails and comments, it’s fantastic to see the global interest in a more open-source and collaborative future for agtech.

A final comment before diving in; this edition isn’t about judging predictions or trying to pick the winners, but rather about building systems that will uncover black swans the fastest and be most flexible in managing those moments when (not if) they come. For me, open-source development seems like a very logical choice, for reasons I’ll explore directly.

So, let’s read on!

Over the past couple of weeks, I have been hooked on the ‘If you’re listening’ podcast/YouTube show by Matt Bevan at the ABC (Australian Broadcasting Corporation). Besides current affairs and geopolitics, Matt produced a four-part series on the Black Swan Effect described above. It is centered on archival recordings from 1959 of Australians making predictions about the future, which, perhaps unsurprisingly, are quite bleak and also wrong. This is just 14 years post World War Two and only a few years before the 1962 Cuban Missile Crisis. For context, 14 years ago was the 2012 London Olympics, my first year of University, when ‘Call me maybe’ was released and when ‘Thrift Shop’ won the Triple J Hottest 100.

Predictably (?) almost every person was mostly wrong in their prophesying. Many thought that overpopulation would be catastrophic, that planes would get us from London to New York in just minutes or that we would be moving across cities and through time in seconds. There were also very dark predictions around if us modern people would even be there to listen to these recordings, given the concern over nuclear proliferation and the recency of catastrophic war.

I don’t want to paraphrase Matt Bevan and am only doing him an injustice, so I would highly recommend just watching the series. Each episode is only about 25 minutes.

My point is: they were all wrong about the major future trends, so why aren’t we in agtech? Spoiler alert - we are and almost certainly will be wrong.

Spend some time on LinkedIn or X these days and it’s common to see bold claims about the future of agtech. If you want a list of AI-written buzz words, the 2 minute video by Gary Tan of YCombinator is a good watch. In saying this, I have absolutely and hypocritically made hype-y statements.

The biggest hype trends/assumptions I have seen are generally centered on:

This is not a list of things I think are wrong or ending or even something that is close to complete - for that I would recommend reading Shane Thomas’, Sarah Nolet’s/Matthew Pryor/Tenacious Ventures or Rhishi Pethe’s newsletters for much better and more informed insights on trends. Plus, there have famously been predictions about the end of Moore’s Law since it was first coined as a thing. It’s just this is a list of assumptions around which I feel lots of agtech is being built and a lot of content created.

The anti-hype/regen perspective also has some well trodden claims:

If we look at the 80s and 90s and the development of pesticides as perhaps one ‘grey swan’ example, particularly Roundup Ready crops, there was a view among many in the industry that it was highly unlikely for the evolution of glyphosate resistance to occur naturally in weeds at an extent and speed possible to cause problems. Yet, 30 years later glyphosate resistance is endemic. While there were many dissenting views, the future of large scale crop production seemed to quite clearly be stacking herbicide resistant traits and pest control genes. We started with Roundup Ready, then Roundup Ready 2 Xtend, then Xtend Flex and now Vytonic with five-way resistance to glyphosate, glufosinate, dicamba, 2,4-D and mesotrione.

It reminds me of the highway lanes meme: ‘just one more resistance stack bro, just one more, palmer amaranth won’t develop multi-resistance to five herbicides’. A great article on Upstream Ag recently on adaptation of weeds too.

The people in the ABC recordings from 1959 predicted that transport would continue to get faster and the Concorde only seemed like a logical next step. But the 1970s oil crisis and the introduction of the 747 for slower, but more accessible mass transit put an end to the Concorde (wildly oversimplified). Overpopulation and mass famine were also common predictions, but the Green Revolution spearheaded by the work of Norman Borlaug helped save billions of people.

1990s Monsanto and other chemical companies predicted the future of agriculture was stacking herbicide traits and using non-selective chemicals such as glyphosate for control. And to be fair, this technology has been very effective and allowed many environmental wins in places like Australia with no-till, residue retention systems. The issue is, the assumption that we can outcompete weeds this way is in some ways a ‘grey swan’ event. In that it was predicted by some in the industry, but the system required us to continue to pursue herbicide resistance traits or more increasing herbicide use and really offered no flexibility once down that path. Weeds will continue to develop resistance (or whatever you want to call it) to whatever we throw at them. Nevertheless, 30 years ago, I’m not sure it would have been quite so clear that we would have six way resistant palmer amaranth.

So how do we avoid locking ourselves into the next treadmill of stacking resistance traits? How do we avoid being the Concorde of the 2020s or the black swan of the 1600s?

The New York University Professor, mathematician, philosopher and statistician, Nassim Nicholas Taleb wrote a fantastic book on this called The Black Swan: The Impact of the Highly Improbable, describing the importance of building for resilience and robustness instead of for prediction. Taleb has a slightly different take on Black Swan Events, where they should be:

One key difference with Taleb’s take is that the surprise of the event is determined by the observer and should not be predictable. He argues the 2020 pandemic was not a black swan event, because pandemics were entirely predicted to occur in the future.

However, the release of ChatGPT in November 2022 is largely considered a black swan event because of the surprise by which it took the large majority of the world, the immediate impact it has had and the retrospective lens of predictability that has been applied since.

Open-sourcing strategic aspects of your business can be a big win here because it has three main advantages.

For example, because of the AI black swan event, the software moat of many agtech startups is approaching Aral Sea levels. In 2019 a solid ML and CV stack in your camera-based weed detection startup was a huge moat because of the expertise and time it took to build. These days, it is orders of magnitude easier to build an efficient pipeline, so that no longer represents fundamental value. On the other hand, people can now take the OpenWeedLocator Github link, paste it into ChatGPT and have it build them a bespoke interface and weed control system. Not to say that complex production systems are equivalent to the OWL - but the differences are diminishing as LLMs improve.

Open source allows the AI black swan event to be a net positive event because it gives many more people the keys to use it, and improves the base user group of the system. On the other hand, for a closed/proprietary system or shut off asset, it represents an existential threat because now anyone can replicate what your software does or AI models will simply not interact with what you have produced.

In OSA #5, I explained how the correct advice for most startups was to focus. Focus on a region, crop and specific problem. Yet, specificity leaves you open to catastrophic black swan events within that niche. Instead, leaving your tools open and allowing users outside of your focus area to work on them gives you an easier pivot when the time comes, because you already have in-field experience, feedback and connections without ever needing to invest in that market.

It’s why public funding agencies should always invest in open-sourcing research and datasets, because it largely derisks the work becoming either stranded or irrelevant. Same for scientists and researchers. Agencies should pursue a strategy of funding the rising tide, where the open tools and shared datasets ‘lift all boats’. The approach is not just an altruistic option but rather the only strategy that still works when the black swan events inevitably show up.

One of the biggest risks farmers face in adopting agtech from startups is the risk of the company disappearing (often through no fault of the founders). Startups are hard, and harder in agtech still. So the chances of a company making it are slim. It is a much easier sell to the farmer if you can articulate that they will always have access to and can control and work with the system, even if the company fails. Startup failures are hardly black swan events, but the principle stands for unpredictable failures too. If a farmer can purchase the device or software and know that in the worst case, they can pay a local to help maintain or build with it, the decision to invest is easier.

On a more personal note, having had my own brush with cancer and seeing your ‘probability of survival’ written on a piece of paper in front of you is jarring to say the least. But what I know for sure is that anything I do open-source will always live on. Fortunately, I’m very likely not going anywhere for a long time and very thankful for a good outcome.

Uncertainty is certain. That is one thing we can count on. And the truly positive side of this is that open-source development and strategies provide genuine opportunities to capitalise on any upside from these transformational events. Plus, black swans don’t always mean negativity. From my perspective at least, AI has had a net positive effect on many aspects of my life. The Green Revolution as also clearly positive and transformational work that we still benefit from today.

The latest versions of the OWL3.0 are on the way with new stainless steel shrouds and mounting surfaces. You can follow along on the Community Page.

We’ve just crossed 300.000 hosted images on Noktura mostly from previous farmer-led data collection efforts on a SwarmFarm robot and an ATV. I have about 800.000 images I will slowly be uploading to the platform, so keep an eye out. If you’re interested in using the platform, let me know. It’s still in Beta, so you’ll need a custom code to join. Can’t thank Brad Plant and Terry Antonio enough for their efforts collecting so much data over the last couple of years.

Some exciting developments though in the last couple of weeks. It now supports annotation upload and download and export in multiple formats. So you can now share annotated datasets with neighbours, colleagues and friends as needed. Custom download selections are also up and running, so you can create cross-dataset downloads to suit the gap in your data.

You can now also search across all datasets and images based on the metadata filters, and hopefully in the future, image embeddings.

While talking about black swans and the rise of AI, I just wanted to mention my own AI policy I follow when writing these newsletters: if I don’t write it myself, then I won’t send a newsletter.

I of course use a lot of AI in doing research, coding, brainstorming ideas and challenging my assumptions here so that I can hopefully provide a more holistic perspective on the topic. But a major reason I write these articles is to explore ideas myself, and regurgitating AI outputs doesn’t help me learn, and probably doesn’t provide you with anything particularly interesting. So I promise these will always be my own typo-ridden, awkward-syntax abundant and fairly blunt thoughts on a range of open-source topics.

What about that list of trends I mentioned at the start? Well, they really don’t fundamentally matter. I am no better at predicting the future than anyone else and we can take solace in the fact that we are all equally bad.

Yet, what does matter is how we face this uncertainty. You can't predict what the black swan is, when it will arrive, and if it will be the one that hisses at you and bites your hand, or the one that is just chill and poses for a photo. The only thing you actually get to decide, is whether you are the kind of system that gains from surprise or the kind that gets bitten.

After years of driving past black swans on the way to school, who would have thought I’d be writing about them in a newsletter on open-source agtech? Hard to predict, I guess.