So we’re currently seeing a Hudson-River-Moment of A.I. in 2022, mainly driven by applications such as Midjourney and Dalle-2 in the field of image generation, and OpenAI’s ChatGPT in the field of natural language processing. Like in other many industries, language models like ChatGPT have the potential to bring about quite some changes in the agricultural industry and open up new opportunities, businesses, and job prospects.
The question I ask myself: How can OpenAI and ChatGPT language models be utilized in agriculture to help and in general? In this article, I will elaborate on:
1. The Age of the generalists
2. How language models can help in agriculture
3. How GPT4 & image description can help in agriculture
4. How autoGPT will help agriculture: Combining AI agents with the web
5. Limitations: Inability to Abstract and Understand Cause-and-Effect Relationships
6. Actual examples of how farmers try to use chatGPT
Do you want to test how chatgpt and agriculture works? After some time of contemplation, I have decided to combine agriculture and a large language model, resulting in agriGPT.
The Age of the generalists
This is a powerful tool for generalists, as it helps them to quickly acquire and understand new information, and to rapidly develop their understanding of complex topics. Furthermore, language models, such as chatGPT, provide generalists with the ability to learn faster and more efficiently from the vast amount of knowledge and information online.
This is especially beneficial for farmers, who often need to be well-informed about a wide range of topics, from mechanics and finances, to biology and marketing and sales. Using language models, such as chatGPT, farmers can quickly learn new information in a more efficient manner. This enables them to gain knowledge and understanding of the topics they need to know, which is essential in their daily work. For example, they can use chatGPT to quickly understand the complex workings of machinery, or to quickly gain an understanding of the financial side of farming.
The most obvious answer to that will be in a few years, for sure: Ask me anything. Farmers will probably be THE group that will benefit from AI and Robotics, as they need to cover a variety of challenges in different types of biology, technology and business.
We’re ready to look into the future now: OpenAI is a leading research organization in the field of artificial intelligence. One of their most notable achievements is the development of a state-of-the-art language model called GPT-3 (Generative Pretrained Transformer 3), which has the ability to generate human-like text.
In the realm of agriculture, GPT-3, also known as ChatGPT, has the potential to revolutionize the industry by providing farmers with valuable insights and assistance. Here are just a few examples of how ChatGPT can be used in agriculture.
How language models can help in agriculture
Increase general output of farmers and agribusiness professionals: ChatGPT can simplify and speed up repetitive mental tasks, leading to increased efficiency for farmers and resulting in overall increased productivity. With its advanced language processing capabilities, ChatGPT can automate many tasks, freeing up time and allowing farmers to focus on more important aspects of their work. Whether it’s streamlining record-keeping, generating reports, or assisting with decision-making, ChatGPT can help farmers be more productive and effective in their work. Agribusiness professionals will thankfully receive help when it comes to tasks such as content creation, email templates, streamlining emails, training materials, self-serve intelligence tools, and more.
Providing expert advice: ChatGPT can be trained on vast amounts of agricultural data, including information about soil conditions, weather patterns, and pest control. This allows it to provide farmers with personalized recommendations and advice on how to optimize their crops.
ChatGPT can be trained to recognize the specific soil and climate conditions of a particular farm. This allows it to provide tailored recommendations on the most suitable crops for that farm, taking into account factors such as the soil’s nutrient levels and the local weather patterns.
Improving crop yields: ChatGPT can help farmers identify the most suitable crops for their specific soil and climate conditions. This can increase the chances of successful harvests and ultimately improve the yield.
Pest control: ChatGPT can provide farmers with valuable information on how to control pests, such as by identifying the most effective pesticides and providing advice on how to apply them. This can help farmers protect their crops and increase their profitability. Pest control is an important aspect of agriculture, as pests can cause significant damage to crops and reduce their yield.
Language models such as provided by OpenAI, have the potential to revolutionize pest control in the agriculture industry. One of the key ways in which models can assist with pest control is by providing farmers with valuable information on the most effective pesticides to use.
This can include information on the specific pests that the pesticides are effective against, as well as how to apply the pesticides in the most effective manner. Tools like ChatGPT can provide real-time updates on pest activity in a particular area. This can allow farmers to take preventive measures and protect their crops before pests have a chance to cause significant damage.
Additionally, ChatGPT can provide farmers with personalized recommendations on the best pest control strategies for their specific crops and climate conditions. This can help farmers optimize their pest control efforts and ultimately increase their productivity. provide valuable insights and assistance to farmers.
Data analysis and prediction: The artificial intelligence language model developed by OpenAI has the potential to greatly aid in data analysis and prediction in agriculture. The tool can process large amounts of data and make predictions based on that information. To achieve this, it is essential to provide the model with adequate datasets and clear guidelines on how to interpret that data. This will enable the AI language model to make accurate predictions and provide valuable insights to farmers. However, if the data is inadequate or the interpretation rules are not clear, the results produced by the AI language model may be inaccurate or unreliable. Therefore, it’s crucial to ensure that the AI language model is trained on high-quality, relevant data and that the interpretation rules are well-defined to ensure its predictions are as accurate as possible. If a large agribusiness has an API connected to their internal systems, the language models could be trained on their internal data to enhance its abilities.
Identifying diseases: Being part of the above mentioned data analysis, ChatGPT can be trained to recognize the symptoms of various plant diseases. This can allow farmers to identify and treat diseases before they spread, reducing crop loss and increasing productivity.
(The potentially more interesting approach would be to train image models to directly use photographs and other visual data, possibly by combining them. For instance, I have experimented with training Microsoft’s LOBE to predict vine diseases, and the results were remarkable. However, this topic deserves its own separate discussion in a different blog article.)
Neverthelesse, ChatGPT can be trained to recognize the specific symptoms of a particular plant disease. This could include visual cues, such as changes in the color or texture of the plant, as well as behavioral changes, such as the plant’s reduced growth rate or decreased resistance to pests. Once a plant disease has been identified, the model can provide farmers with personalized recommendations on the best treatment options.
This could include information on the most effective pesticides or other methods for controlling the disease, as well as advice on how to apply the treatment in the most effective manner. Additionally, a model like ChatGPT can provide real-time updates on the spread of diseases in a particular area. This can allow farmers to take preventive measures and protect their crops before the disease has a chance to cause significant damage. Overall, the integration of language models into the identification and treatment of plant diseases has the potential to provide valuable insights and assistance to farmers. This can help prevent the spread of diseases and ultimately increase the productivity of the agriculture industry.
Any question that comes up in the context of a farm: As you may know, farmers are multi-talented managers that need to kind of know everything and be maximalist generalists.
Optimizing operations and reducing costs: With all the available options, it’s highly likely that cost reductions will occur. However, it remains to be seen which specific examples will materialize and where language models can make a substantial impact. That’s why I will examine some examples from December 2022 to February 2023 that I am aware of further below.
How GPT4 and its image description can be useful in agriculture
The latest version of ChatGPT, GPT4 , comes with advanced image description capabilities that can be particularly useful in the field of agriculture. With its ability to describe images in natural language, ChatGPT 4.0 can be a powerful tool for analyzing crop health, identifying pests, and analyzing detailed crop and field imagery.
Crop health analysis is essential for ensuring healthy and productive crops. By analyzing images of crops, ChatGPT 4.0 can help farmers identify any signs of distress, such as discoloration or wilting, that may indicate a potential problem. This can enable farmers to take timely action to address the issue and prevent crop damage or loss.
Pest analysis is also crucial for maintaining healthy crops. By analyzing images of crops, GPT4 can help identify any signs of pest infestation, such as bite marks or webbing. This can enable farmers to take appropriate measures to control pests and prevent damage to their crops.
In addition to detailed crop imagery, GPT4 can also analyze field imagery to provide a comprehensive view of the agricultural landscape. By analyzing images of fields, ChatGPT 4.0 can help identify potential issues such as soil erosion, waterlogging, or nutrient deficiencies. This can help farmers make informed decisions about crop rotation, irrigation, and fertilization.
Overall, technology like the image description capabilities of GPT4 can be a game-changer for the agricultural industry. By providing farmers with detailed insights into crop health, pest analysis, and field imagery, the language model can help them make more informed decisions and take timely actions that can improve crop yields and profitability
How autoGPT can help in agriculture and farmers
So I tried to use autoGPT as an assistant to help me with agricultural topics, specifically to help with finding subsidies. An autonomous GPT agent that searches and scrapes the internet could be pretty useful. So I called this AI agent “subsidyAI” and asked to find some interesting current programs. While the AI managed to access the right websites to research, it failed gathering the right information (and let’s not even talk about output). I believe that autoGPT for agriculture could become huge, especially when connected with image detection. autoagriGPT here we come, let’s give us 2-3 more months and I will try this again. Read below what my program tried to do here.
The program is a Python script that uses the AutoGPT API to automate certain tasks related to investigating agricultural subsidies in the Poitou-Charentes region of France. The program starts by searching for and collecting information on relevant regional and environmental indicators, such as data on demographics, labor markets, and social statistics. It then uses this information to identify potential sources of agricultural subsidies in the region, specifically related to crops such as apples, vines AOC cognac, and alfalfa.
The program also plans to start a GPT agent to help generate ideas on how to qualify for and use these agricultural subsidies. The agent is provided with a clear and detailed prompt to focus its responses, and is expected to provide at least five different ideas for how a farmer could qualify for agricultural subsidies in France, and five different ideas on how to make the best use of these subsidies.
The program then aims to research how to determine the value of subsidies and how farmers can apply for subsidies they are eligible for in Poitou-Charentes. It plans to conduct a thorough search of relevant government and organization websites, check with farmers in the region, and consult with agricultural advisors to gain insights into the application process for these subsidies.
Throughout the program’s execution, it is mindful of the potential limitations of the information it collects, and takes care to double-check any steps or recommendations suggested by the GPT agent or other sources. Well, this was not successful.
But first let’s look at what language models can’t do (right now):
Limitations: ChatGPT’s Inability to Abstract and Understand Cause-and-Effect Relationships
AI and machine learning technologies have provided farmers with valuable data and insights to e.g. optimize crop yields. However, the limitations of language models such as ChatGPT have become increasingly evident. Language models lack the ability to abstract and understand cause-and-effect relationships, making them an insufficient replacement for the cognitive understanding of experienced farmers.
To succeed in agriculture, farmers must have a deep understanding of their environment and the ability to adapt their practices based on ever-changing conditions: Think abstractly and make informed decisions based on incomplete or uncertain information. While AI can assist in these decision-making processes, it cannot replace the critical thinking skills and experience of farmers. (Well that’s kind of obvious, at least to me.)
Therefore, it is important for farmers to continue relying on their own cognitive knowledge and expertise, while also utilizing AI technologies as a tool to enhance their decision-making and improve agricultural outcomes. It’s all about the right balance between the benefits of AI technologies and the (for now) irreplaceable value of human understanding and interaction.
Examples of how farmers & growers try to use chatGPT
- After observing, testing and reading about chatGPT and farming, I decided to build my own agriculture GPT version.
I built an AI assistant for agriculture: I call it agriGPT
Let me know what you’re thinking at twitter.com/agtecher_com or send feedback
- This side project, named Farmer GPT, uses ChatGPT and Whisper to provide valuable assistance to farmers in India. It helps them improve farming practices, yield, and profitability by providing real-time weather information, personalized crop recommendations, and peer-to-peer communication with experts and other farmers. This kind of AI assistant has the potential to replace agronomists, revolutionize farming practies in India by leveraging the power of AI and secure messaging to help farmers succeed in a competitive industry.
The “Farmer GPT” has a drawback as it currently provides similar results to the standard GPT without any finetuning model or connected database. Furthermore, users may receive off-topic responses, such as an answer to a question like “how to bake a potato” even though it is not related to farming.
- Yield estimation & the Mango case: Vishnu Sankar asked Feb 11: Got varying answers for the genetic yield potential of mango and the yield gap from the actual. So, checked with ChatGPT. Mangoes #Mango cultivation #UHDP #Farming #Agriculture #Fruits #Orchard
The language model predicts that a mango tree can produce between 50-300 mangoes per year per tree, with high-yielding trees that can produce up to 1,000 mangoes per tree per year.
- Another nice example I found is on regenerative agriculture in vineyards: wine thinker aka post.news/robertjoseph @robertjoseph writes on Feb 5 2023: How many of you know about the use of chicken feathers in viticulture? @SpierWineFarm uses them And #Chatgpt explains some reasons why I asked it to “Write an essay about the role of chicken feathers in regenerative agriculture”
I have asked the Spier Wine Farm via twitter if the essay is more or less accurate, so let’s see about that.
- Agricultural Education, A very specific example by an Agriculture Teacher: Cassie Joiner @Cass_Joiner wrote on Feb 1 2023: “The first task I gave ChatGPT was to write a lesson plan for what I’m teaching tomorrow, “Genetics in Agriculture” and my teacher life is forever changed”
- Tomato pest control, both question as text and python code for image detection: Another very specific question in the field avoid powdery mildew with Tomatoes: Shailendra singh @srj0408 writes on Dec 10, 2022: Tried ChatGPT today with questions around #agritech and the results were awesome. #agriculture #chatgpt3 #ChatGPT #OpenAI #OpenAIChatGPT #startup
The user was testing chatGPT as a coding support (coding python with chatGPT works in an impressive way already even though it was not trained to do so) to implement a support vector machine SVM model in python for image analysis of plant leaves.
To avoid powdery mildew in tomato plants, you can take the following steps: 1. Plant resistant varieties: Some varieties of tomato plants are more resistant to powdery mildew "than others. Choose a variety that is known to have good resistance to the disease, such as 'Mountain Magic' or 'Legend'." 2. Provide good air circulation: Powdery mildew thrives in damp,stagnant air.To prevent the" "disease,ensure that your tomato plants have plenty of space betweenthem, and provide good air circulation around the plants by pruning them regularly and avoiding over-watering." 3. Avoid overhead watering:Watering your tomato plants from above can cause the leaves to become damp,which creates a favorable environment for the growth of powdery mildew." "Instead,water the plants at the base of the stem,or use a drip irrigation system.. 4. Use a sulfur-based fungicide: If your tomato plants do develop powdery mildew,you can use a sulfur-based fungicide to treat the disease. Apply the fungicide according to the instructions on the label,and repeat the application at regular intervals to prevent the disease from spreading. 5. Practice crop rotation: Powdery milde…"
- Not a specific example, but interesting that it was brought up by greenhouse & data specialist John Beauford: How chatGPT may impact the horticulture industry: ChatGPT can be of great assistance to horticulture growers by providing an advanced and integrated technology system for their greenhouses.
By integrating with grower’s ERP software, sensors, weather forecast and sales analytics, this language model can help growers better understand their operation, create and execute better plans, optimize profits, and improve replenishment performance. It can automate repetitive tasks and complex data analysis, freeing up time for more important tasks. This technology will be a key part of the next wave of business automation for growers and has potential applications for real-time crop diagnostics and cost analysis. The integration of a language model with the data and systems specific to a grower’s operation can provide a more human-like interface to their greenhouse business, making it easier for growers to access and analyze critical information.
A real-life example that was relevant on our farm
This is an actual question I had in summer 2022, and I tested it against openAI’s chatGPT:
Question: "I have 1,5 hectares of vineyard close to Bordeaux and i need to install irrigation, what do i need in terms of equipment and what kind of PSI is needed to cover 1,5 hectares?" Answer by openAI: "Installing irrigation in a vineyard is an important step in ensuring the health and productivity of the vines. The specific equipment you will need will depend on a number of factors, including the size of your vineyard, the type of irrigation system you choose, and your budget. In general, a vineyard of 1.5 hectares will require a fairly large irrigation system. This will likely include a main water supply line, as well as a network of lateral lines and sprinklers or drip emitters to distribute water throughout the vineyard. You may also need to install a pump to increase the water pressure and ensure adequate coverage. As for the water pressure, a typical vineyard will require a pressure of at least 40 PSI (pounds per square inch) to ensure adequate coverage. However, this can vary depending on factors such as the type of irrigation system and the layout of the vineyard. It is best to consult with an irrigation specialist to determine the specific water pressure requirements for your vineyard. In conclusion, installing irrigation in a vineyard of 1.5 hectares will require a fairly large and complex system. You will need a main water supply line, a network of lateral lines and sprinklers or drip emitters, and possibly a pump to increase the water pressure. The specific equipment you need will depend on factors such as the layout of the vineyard and your budget. It is best to consult with an irrigation specialist to determine the exact equipment and water pressure requirements for your vineyard."