“Software is Feeding the World” is a weekly newsletter about technology trends for Food/AgTech leaders.

Greetings from the San Francisco Bay Area after a long’ish break. Due to a technical issue, today’s edition is coming out later than normal. I hope to go back to normal operations starting from next week.

Now onto this week’s edition.

There has been significant talk about Large Language Models (LLMs) like Bard and ChatGPT recently. My friend Shane Thomas did a fantastic primer on the implications of Large Language Models like ChatGP within agriculture.

In today’s edition, I want to briefly explore them in the context of a trend towards automation and advances in technology in general.

Note: CCA stands for Certified Crop Advisor in North America.

I.

Farmers often use the services of an agronomist or many times someone who is certified to provide them with agronomic advice. It is called a CCA (Certfied Crop Advisor) in North America.

​The CCA certification was established in 1992 to provide a benchmark for practicing agronomy professionals in the United States and Canada.

To quality as a CPAg or CCA, one must have the following:

• B.S. degree which includes a minimum of 6 to 9 hours in each of the professional core categories; crop management, pest management/crop protection, and soil science. (Here are the requirements for a Bachelor's Degree program at University of California, Davis for reference)
• 6 to 9 additional semester hours that relate to the three professional core areas and must have a minimum of 30 semester hours of coursework in agronomic related courses.
• Pass a series of international/provincial exams.
• Signed code of ethics. Every certified crop advisor must abide by a strict code of ethics and exercise due diligence when making recommendations.
• Continue education. Because agriculture changes so rapidly, crop advisors must take 40 hours of continued education every two years.

The certification serves as a signaling mechanism to the CCA’s knowledge, commitment, and experience about their experience. If farmers trust the CCA brand, the presence of a certification reduces the search costs for farmers. It provides some minimum level of confidence to the farm, when they encounter a CCA certification.

(Not in this case, but oftentimes required certification acts as a protection mechanism for the incumbents and entrenched players. I will leave my rant against random professional certifications used as requirements for doing certain jobs for some other day.)

II.

Business Insider recently published a list of exams Large Language Models have passed. The list includes the following so far (and I am sure counting)

• Uniform bar exam for lawyers: Scored in the 90th percentile of students.
• SAT reading and writing section: Scored in the 93rd percentile.
• GRE: 99th percentile for verbal, 80th percentile quantitative, 54% writing text.
• USA Biology Olympiad semifinal exam.
• Advanced Placement exams for college level courses.
• Sommelier (wine making) exams.
• Wharton MBA exam.
• US medical licensing exam.
• Stanford medical school reasoning final.

The list goes on.

It is quite clear.

Large Language Models continue to get better and they can clear (or ace) many written exams.

III.

So can a large language model clear the CCA certification exam? I don’t know if anyone has tried this or not, but it doesn’t seem like a stretch that an LLM could clear the pen and paper section of the certification.

So an obvious question one might ask is whether the Latest Language Models can replace CCAs?

Before we try to answer this question, what kind of evidence are we seeing in a similarly (and probably more) complex field like medicine?

In a recent paper published in The Journal of the American Medical Association (JAMA), a peer-reviewed medical journal published by the American Medical Association, the authors published the following question:

The findings from the study showed, (emphasis mine)

​
​

Does this mean chatbots will replace doctors? (There are many problems with how this study was conducted, but even if I leave them aside, this is a fascinating study!)

I don’t think so.

Every patient is unique and would not trust a chatbot for important and oftentime complex decisions involving multiple parts of the medical sciences, and the complex interactions within the human body.

But a technology like this can unbundle some of the lower level decisions made by a doctor or help doctors make faster, better, and more nuanced decisions with the help of technology.

This is not something new because as patients we often route ourselves to the internet for answers to basic questions or get routed to a nurse line, if we want to talk to a human (at least in the US). Doctors will look at X-Rays or blood reports to make decisions and prescribe a particular course of action.

If I go back to Shane’s primer (emphasis from Shane),

Similar to the doctor's example, farming happens in the real physical world, and there are thousands of signals which come from the physical world, which have to be put together before making a material farming decision.

This is all to say, technologies like LLMs can augment human expertise, intuition, and knowledge of local conditions to help make better decisions. It can make the current structure much more efficient, by outsourcing basic questions to tech, and taking help from the tech to augment human experience and intuition.

It might change the way a particular job is done today.

This has always been true in every profession, whether it is farming, or medicine.

IV.

We see a similar trend in robotics and automation in agriculture. Certain types of automation and robotics, take on repetitive, tiresome, and dangerous jobs from humans. It is often found that augmenting the expertise and intuition of humans is well served by technology.

Agriculture is complex, with an open system, unpredictable weather patterns, variety of soil characteristics, localized decision making, and decision making based on experience. Due to these factors, it is believed that a human can never be taken out of the equation.

But it changes the work and the job done by humans.

I had explored these themes in edition 122. Moravec's Paradox for Ag Robotics, and edition 112. Agriculture Robotics is difficult AF​

Should robotics, AI, and ML focus exclusively on automation and replacement of human tasks as performed today? It is tempting to think automation is the holy grail to solve problems associated with human labor, and efficiency.

Augmenting humans with technology opens an endless frontier of new abilities and opportunities. The set of tasks that humans and machines can do together is undoubtedly much larger than those humans can do alone.

V.

In edition 69. The unbundling of humans, I had explored how technology constantly bundles and unbundles human capability to make human beings more productive, change the definition of jobs, and ultimately provide more time for creativity, and leisure, at a higher overall economic level.

Within agriculture, it is not too much of a stretch to imagine a machine learning model, which chooses the right crop type and seed variety based on your past history and agro-ecological conditions. It creates a plan to plant the right amount of seed in the right parts of your field at the right time. The grower or the agronomist can review (or not) the plan.

It will search for the best price for the seed (or a relevant substitute) and place an order. The seed will be delivered in time for your planting operation. The seed is loaded into your autonomous planter. The planter has an accurate map of your field and will plant accurately to the plan created by the model.

If the model can do this consistently, and deliver results, the trust in such models will continue to improve.

You will still need an agronomist or an advisor to ratify decisions or make complex decisions, but their role will have changed dramatically.

Depending on the rate of technology adoption, and what gets unbundled, what it means to “farm” might change in the future.

Mark Young, ex-CTO of The Climate Corporation (digital farming subsidiary of Bayer Crop Science) and I discussed this in edition 52 of the newsletter.​

VI.

So what is the right question to ask?

CCAs or LLMs?

No, the right question to ask is,

How might technologies like robotics, automation, and LLMs augment human intelligence, intuition, and experience and how will it change what it means to be involved with agriculture and farming?

It is CCAs and LLMs not CCAs or LLMs.

What do you think?

💗 If you like “Software is feeding the world”, please share with a friend.

🙏 If you don’t mind answering 3 questions anonymously (2 are optional), I would love to get your feedback.

About me

My name is Rhishi Pethe. I lead the product management and technology delivery teams at Mineral, an Alphabet company. The views expressed in this newsletter are my personal opinions.