Upstream Ag Insights - May 7th 2023
Essential news and analysis for agribusiness leaders
Welcome to the 167th Edition of Upstream Ag Insights!
Index for the week:
CCAs or LLMs?
22 Mental Hacks for Agribusiness Leaders
UPL Wants Half of its Revenue to Come from Biopesticides
Digital Enablement: What’s Holding Us Back?
Australian Farmers Use Starlink Satellite Internet Kits to Access Agtech for Grain Sowing Program
Clocking Up a Week’s Worth of Work in One Day
Small Robot Company Launches Community Bridging Raise
Sentera Announces Series C Funding Expansion
Explaining Tech's Notion of Talent Scarcity
Thanks for reading, sharing, and subscribing!
1. CCAs or LLMs? - Software is Feeding the World
The CCA certification (Certified Crop Advisor) is a standard in agriculture. Passing the exam asserts that an individual has a base competence in soil science, crop physiology, and overall crop production.
My friend Rhishi Pethe asked a great question in his recent newsletter:
So can a large language model clear the CCA certification exam?
Recently, Visual Capitalist published the below image illustrating the percentile that ChatGPT fell on the distribution curve of various placement and professional exams:
It’s safe to say there are areas where ChatGPT is highly capable.
I passed the CCA exams in 2015 and dug into some old practice tests to see how ChatGPT would do.
Of the ten questions I asked, ChatGPT got seven right or 70%, which would typically be enough to pass the exam in any given year.
Any CCA will tell you the exam establishes that you understand a baseline. There is however extreme nuance when making real recommendations to farmers.
For that reason, I asked ChatGPT follow-up questions to dig deeper into some of the questions it got right.
In this example, the number it suggests is relatively low compared to generally accepted numbers, and I found the same with other follow-up questions. The answer fell short of the needs a farmer would have. This illustrates that when it comes to making specific recommendations, it still has room for improvement.
As an interesting aside, I asked the same questions to Norm from FBN, an LLM based on ChatGPT with an augmentation of FBN specific data. Norm had the same results— seven questions correct and three wrong (same outcome on each question). However, when I asked the more specific questions, like in the image above, where I expected Norm to be an improvement, it performed worse—giving no answer and suggesting to consult a label or not understanding the question. I suspect this is a feature, not a bug, for now as FBN would want to ensure the default is “no answer” vs. a “wrong answer.”
In agronomy recommendations, what separates good agronomists from average ones is not answering the broad and basic questions accurately— it’s being able to position good recommendations effectively and correctly navigating challenges that occur on the margins, the edge cases for problems or for farmers wanting to manage their operation at a higher level. This is where artificial intelligence and, more specifically, LLMs like ChatGPT and Norm will be less effective!
This reality reinforces some comments and the image in the April 30th 2023 Edition about the combinations of machines and humans:
Because of this, tools like ChatGPT will commoditize the more straightforward actions and tasks that industry professionals tackle, specifically agronomists. It won’t replace humans; it will re-organize tasks and actions and allow agronomists to reallocate time to different tasks.
In the November 21st 2021 Edition of Upstream Ag Insights, I shared a base framework I use to think about the dynamics of artificial intelligence as it pertains to jobs:
For example, an agronomist is a job that has responsibilities (e.g.: sell products/services, supporting relationships, solving problems etc) that are executed via tasks (e.g.: scouting fields, make fertilizer and chemistry recommendations, etc.) and are made up of actions (e.g., drive to the field to acquire information to inform a recommendation).
Tools like ChatGPT streamline actions and tasks. This gives agronomists and professionals a point of leverage to free-up time and resources to differentiate themselves.
As Rhishi points out in his article, humans augmented with technology opens a new frontier of opportunities. Humans constantly adjust themselves around innovation, as I discussed in February regarding ChatGPT.
If LLMs augment humans, that means that companies that effectively use them will give their teams a competitive edge over others that don’t. The two basic areas in my mind that will give companies an edge are the following:
Which companies have novel data and more of it to make their models superior?
Who can integrate an LLM interface seamlessly into current digital workflows?
With any technology that is valuable and used effectively, it becomes a point of differentiation and that becomes a competitive advantage. Over time that competitive advantage becomes a profit driver, that eventually lowers the costs of capital leading to a virtuous cycle:
LLMs and AI are unlikely to replace agronomists or any agribusiness professional. They have limitations and I’m certain have been over hyped by many. They do, however, present an opportunity for professionals and organizations to gain an upper hand for an array of tasks and actions that can benefit them long term.
For more on LLMs, check out the article I wrote last month on ChatGPT Implications for Agriculture.
2. 22 Mental Hacks for Agribusiness Leaders - Upstream Ag Insights
Over the past few years of Upstream, I have published a “professionals tips” based article for new University and College grads entering the agriculture industry. With it being graduation season, I thought I would re-share it again as it is one of the most popular articles I have written.
One of the most commonly cited pieces of feedback I received around it originally is that it shouldn’t be emphasized to just new grads but all agribusiness professionals.
Taking that into account, in 2022 I adjusted the focus toward agribusiness professionals of all experiences. I emphasized mental models that are useful on a day-to-day basis within the industry.
I won’t claim the principles to be novel for many or all-encompassing. Still, I know the 22 principles act as a constant reminder to help improve my output, critical thinking, professional development, and discipline.
These 22 mental models stem from what I've learned thus far in my career and are the aspects that have had a disproportionate impact on me.
3. UPL Wants Half of its Revenue to Come from Biopesticides - AgroPages
The title of this article is misleading, and it stems from the fact that “sustainable” gets interchanged with “biopesticides."
About 7% of UPL’s business today comes from their Natural Plant Protection (NPP) product line, which generally contains biological-based products. This works out to around USD 400 million in 2022 with USD 6.2 billion in revenue.
What UPL is actually targeting is for 50% of their revenue to come from differentiated and sustainable products by 2027, which 29% comes from today.