Exploring AI and How Ag Businesses Can Use It

As agriculture continues to transform in the face of economic, regulatory, and labor constraints, technology offers farms and agricultural businesses potential solutions to the numerous and varied challenges faced in production, human resources, and marketing.  Artificial Intelligence (AI) is just one more tool in the technological toolbox and it offers numerous opportunities to augment and improve productivity, sustainability, and profitability. This article provides an overview of artificial intelligence and explores some of the potential uses of AI by ag businesses as well as sharing cautionary notes and tips for use.

Artificial intelligence is described as "the ability of a digital computer or computer-controlled robot to perform tasks commonly associated with intelligent beings" (Britannica).  AI has existed as a field of study for several decades, tracing its roots to the work of Alan Turing.  You are probably familiar with the following, all of which are examples of products that many of us use in our everyday lives and that use AI.

• Siri, Alexa, and other smart assistants
• Self-driving cars
• Google searches
• Conversational bots on websites
• Email spam filters
• Netflix recommendations

The discussion about, and interest in, artificial intelligence tools exploded in the mainstream with the release of ChatGPT version 2 in November 2022.  Upon release, people immediately began testing the capabilities of this tool, finding both valuable uses and identifying flaws.  While AI technologies have existed for some time, continuing advances in technology leading to tools including ChatGPT and others are opening new possibilities.  These tools may either be integrated into other platforms being used or exist as standalone applications.  For instance, ChatGPT is a standalone application accessed via browser or mobile device app while Microsoft offers "AutoComplete" as a function in Word.

Types of AI

There are four types of AI that become progressively more sophisticated.  The type is determined by the functionality.

Reactive - a type of AI that operates solely based on the present input without the ability to learn from past experiences or store historical data. It is designed to respond to specific requests in real time and make decisions based on the current environment. Examples: spam filters, customer support chat, computer-generated "players" in video games.

Limited Memory - refers to AI systems that have the capability to use historical data and past experiences to inform their decision-making processes. Unlike reactive AI, which only uses current inputs, limited memory AI can retain information for a short period to improve its performance. Examples: speech recognition, recommendation systems, and self-driving cars.

Theory of Mind – this is the current state of AI and is still under development. It is a type of artificial intelligence that is capable of understanding and simulating humans. This level of AI aims to grasp emotions, beliefs, intentions, and thoughts of other entities (humans or other AI agents), enabling it to interact more naturally and effectively with people. Examples: Google assistant, emotionally intelligent robots (that read and respond to your tone).

Self-aware - also known as self-conscious AI or artificial general intelligence (AGI), represents the most advanced stage of AI development. This form of AI would possess a level of consciousness and self-awareness comparable to that of humans. It would have the ability to not only understand and process information about the external world but also possess an awareness of its own existence, internal states, and the ability to reflect on itself. No current examples exist.

Five Subfields of AI

There are also five subfields of artificial intelligence with which we are likely more familiar. These subfields categorize how the AI application or tool is used.

Image recognition - also known as computer vision, is a field of AI that focuses on enabling computers to interpret and understand visual information from the world, such as images and videos. It involves the development of algorithms and models that can process, analyze, and make sense of visual data like human vision. Examples: Object detection, facial recognition image classification.

Speech recognition - also known as automatic speech recognition (ASR) or voice recognition- enables machines to understand and process human speech. This technology converts spoken language into text or commands that computers and applications can interpret and act upon. Examples: Predictive language, phonetic dictionary, transcription services, voice typing, voice response systems.

Chatbots and ChatOps - applications designed to simulate human conversation through text or voice interactions. They can understand and respond to user inputs, performing various tasks from answering questions to facilitating transactions. Examples: customer support chats, sales, and marketing pop-ups.

Natural language generation - focuses on generating human-like text from data. It involves creating coherent, contextually relevant, grammatically correct sentences and paragraphs. NLG systems transform structured data into readable and meaningful narratives, mimicking human writing. Examples: automated content generation, personalized communication, writing assistance

Sentiment analysis. - is a subfield of natural language processing (NLP) that uses algorithms to determine a text's emotional tone or sentiment. This technology can categorize text as positive, negative, neutral, or more nuanced emotions like happiness, anger, or sadness. Examples: customer feedback and brand analysis.

AI Tools

Within each of the types and subfields outlined are numerous AI tools and applications.  Several of these you may already be familiar with. The following diagram presents another way to categorize these tools, with some tools falling into more than one category.

Current State of AI

A tremendous amount of AI research is being done with major research initiatives in many fields such as medical, engineering, and agriculture.  The value of AI comes with knowing and appreciating how it can be appropriately applied. Consumers, therefore, must understand how it works, and where AI technologies can be used in businesses and where they cannot. The current state of AI is marked by significant advancements across various domains, driven by rapid progress in machine learning, deep learning, natural language processing, computer vision, and reinforcement learning. AI technologies are increasingly integrated into industries.  However, challenges such as data privacy, bias mitigation, and AI ethics remain significant concerns, highlighting the need for responsible AI practices. Ongoing research in AI will continue to advance while addressing societal and ethical considerations.

Uses for AI in Your Ag Business

Artificial intelligence can be used and integrated into virtually all areas of an ag business, including:

• Production operations
• Marketing
• Human resource management

Certainly, just because AI can be used does not mean that it must or should be and this article will later discuss some limitations and considerations.  However, a basic understanding of the potential opportunities and benefits that AI offers is useful. AS with any tool, the decision to use an AI tool or application should be made based on individual circumstances and goals with the recognition that these may change over time.

Production

Agriculture can leverage the power of AI to support and enhance farm production. Precision agriculture is a popular term that utilizes data from sensors, satellites, and drones to provide detailed insights into soil conditions, weather patterns, and crop health. This allows farmers to make informed decisions about when and where to plant, irrigate, and apply fertilizers or pesticides, leading to more efficient resource use and higher yields. Technological advances are taking steps toward automated machinery and crop monitoring that do tasks such as planting and weeding, can follow crop growth, and identify nutrient deficiencies and pests.

In livestock production, AI is also making advances in multiple areas. Health monitoring and disease prevention can be incorporated through wearable devices and sensors that follow the animal’s health and behavior in real time. Automated feeding systems and nutritional analysis can be tracked through AI systems to optimize schedules and rations.

Marketing

Marketing is both an art and a science.  A certain degree of creativity is needed to develop marketing content that resonates and engages with the desired consumer segment and analytical skills are needed to study marketing trends and determine the effectiveness of marketing strategies.  For small and farm businesses with only a few people responsible for operating and managing all aspects of the business, dedicating sufficient time to marketing can be demanding.

There are AI tools to assist with all aspects of marketing – content creation (text/copy, image, video, and music), scheduling, data analysis, and communication to name a few. For anyone whose strength is not in marketing, having an AI tool at hand can be very useful.  For instance, with artificial intelligence (AI) tools it is possible to quickly generate content in a matter of seconds, freeing business owners or employees to focus on other tasks.  Additionally, AI tools can easily generate content for use in various marketing formats – from long-form articles and blog posts to concise text for tweets or as part of a Facebook post. For cash-strapped businesses, utilizing AI can be a cost-effective option compared to hiring marketing consultants or developing an in-house marketing department.

Human Resource Management

While perhaps less evident than how AI can be applied in production or marketing, AI can also be used within human resource (HR) management.  For instance, AI can improve recruiting efforts by analyzing the avenues (job boards, social media, etc.) used to advertise job openings, position presentation (e.g. description, language, information shared, etc.), and the corresponding responses.  This allows employers to determine the most effective and successful recruiting strategies. Recruitment is only part of the hiring process.  AI can also be used to improve the interview process by generating questions that match job descriptions and improve the ability to assess a candidate's qualifications and fit for the role. Using AI in these ways for recruitment and hiring can result in finding higher quality candidates and employees and more efficient resources allocated to these purposes.

HR management can also use AI for aspects of onboarding, training, and promotion depending on the ag business's needs.  It is worth noting that while there are several acceptable uses of AI in HR management (e.g. data organization and analysis), AI should not be the sole tool used for evaluation of a candidate or employee; final decisions should be made by humans.

Examples of AI for Marketing Applications

Since it can be helpful to see examples of how AI can be used, let's look at two marketing scenarios and the use of ChatGPT.  ChatGPT, a language model developed by OpenAI, is among the most popular platforms currently available.  It is designed to understand and generate human-like text based on the input it receives.  There are both free and paid versions available.

To share an example what ChatGPT can generate, consider wanting to create a press release that conveys the following message:

"Sunnyside orchards is planning its family fun day for Saturday, September 7, 2024 from 9am-3pm. There will be activities for kids and adults, including hayrides, face painting, and lots for sale at the bakery. Hope to see you there!"

Asking ChatGPT (2024) to create a press release using that information as the input generates the following:

For Immediate Release

Sunnyside Orchards Invites Families to a Fun-filled Day on September 7, 2024

Sunnyside Orchards is thrilled to announce its upcoming Family Fun Day, scheduled for Saturday, September 7, 2024, from 9 am to 3 pm. This exciting event promises a day of joy and entertainment for both kids and adults alike.

Event Highlights:

Date: Saturday, September 7, 2024

Time: 9:00 am - 3:00 pm

Activities for All Ages:

Sunnyside Orchards has curated a delightful array of activities to ensure a memorable experience for everyone in the family. Highlights include enchanting hayrides, creative face painting for the kids, and a visit to our bakery where you can explore and purchase a variety of delightful treats.

Bakery Extravaganza:

Indulge your taste buds at our bakery, where an assortment of delectable goodies will be available for purchase. From mouth-watering pastries to freshly baked delights, there's something for every palate.

Save the Date:

Mark your calendars and join us for a day filled with laughter, joy, and delightful moments. Sunnyside Orchards' Family Fun Day is the perfect opportunity for families to come together, enjoy quality time, and create lasting memories.

Event Details:

Date: Saturday, September 7, 2024

Time: 9:00 am - 3:00 pm

Location: Sunnyside Orchards

We look forward to welcoming you to Sunnyside Orchards on September 7th for a day of wholesome family fun. See you there!

For media inquiries, please contact:

[Your Contact Information]

Note: This press release is for informational purposes only and is not intended for publication without prior approval.

If you felt this suggestion was too formal, you could ask for the output to be made in a more casual tone or simply utilize a mix of the original input and the AI-generated output to create a final product in your voice sharing the information you want relayed.

Here is another example of using ChatGPT (2024) for a marketing task, this time to generate ideas for social media content.

Prompt: Give me 10 ideas for Facebook posts to promote a u-pick strawberry patch

The generated results provide ideas on topics to post about, as well as suggested language for posts which could be used as is or edited to match the ‘voice’ of the business and suggestions for other marketing content (e.g. the vague reference to a blog or webpage in #8).  In addition, while this example was performed for an imagined strawberry operation, the suggestions could easily be adapted and applied by other types of farm businesses.

AI Limitations and Considerations in Use

While the potential uses for artificial intelligence are vast, there are several limitations that anyone considering its use should be aware of.

• AI can generate incorrect or cite inexistent data and/or sources. Artificial intelligence tools are "trained" by feeding them datasets from which the model used by the AI tool finds patterns. The patterns are then used when responding to inquiries. If the AI hasn’t been trained on data appropriate or relevant to your inquiry AI output may not be factually correct. It is critical to check the accuracy of generated content.  It can be risky, for example, to rely upon AI to provide correct information on current regulations pertaining to the licensing and sales of food products in your geographic area.  Thus, a degree of caution should be used when asking for increasingly detailed information. Further, since AI tools may not disclose the sources of information for the content generated, there is the potential that output includes copyrighted or plagiarized material.  While AI tools have begun to incorporate citations, these should be verified as well.
• AI results can be biased and/or unethical. Bias in AI results can be the consequence of programmer bias and/or data bias. If the human responsible for training the tool has a bias that impacted the choice of data used in development, AI tool output may also be biased.  Additionally, the bias is imbedded in much of the real-world data used to train AI tools.
• Overreliance. It is important to not become dependent on AI. While there are many benefits to its use, it is crucial to review outputs and results that are generated to determine accuracy and appropriateness.  Overreliance on AI can also result in the loss of job skills and empathy when faced with in-person interactions. 
• Privacy concerns. The use of AI can raise concerns in privacy as it collects large amounts of personal data, such as location, social media activity, information on expenditures, etc.  This content remains accessible to the tool for continued training, therefore caution should be used before entering private and personal data.
• Brand dilution. Savvy consumers and some of your target market will recognize AI-generated language and augmented photos. It’s important to not simply use the text output as generated verbatim. Your marketing needs to feel authentic to the consumer so ensuring that the "voice" of your marketing content matches real life and your business’s brand is important. 

Tips for AI Use

If you decide to try integrating AI into business functions and activities, here are some tips for success:

• Understand the tool. It is important to understand the functionality AI tool being used, including how it processes information. This will assist in determining the appropriateness of its application to each instance of use and the veracity of output. 
• Provide specific prompts. When using language-generating AI, provide specific prompts to improve output.  Numerous guides and articles online provide examples of how to best word prompts to generate the desired type of output.  As you gain experience using an AI tool you will also learn how to phrase prompts to generate desired output.
• Be prepared to work through iterations of requests to generate a final product. The accuracy and appropriateness of results improve with the level of information provided to the AI tool through prompts. Providing the "right" amount and type of information typically takes multiple tries, particularly if the user wants a highly detailed response specific to their operation.
• Ensure accuracy of information. AI output is only as good as the data on which it is trained. Therefore, ensure that the output generated in response to your prompts fits your understanding and knowledge of the topic.
• Personalize output. Even with prompts that are specific to your business and situation, responses should be tailored to fit your personal voice and brand before using in your marketing. Remember that consumers want to connect with businesses on a personal level, something that AI can’t replace.
• Transparency and citation. There continues to be a discussion on when and how to cite AI-generated tools.  Consider whether and under what circumstances you will indicate that AI is/was used. In situations where consumers or employees may expect a personalized experience, not sharing that AI is/was used may be viewed as a breach of trust.

Artificial intelligence can be a useful tool, just as any number of tools.  However, it should be used with full awareness of its capabilities and limitations.  Ag business owners should view AI as an additional tool in their toolbox, not as a replacement for human experience, knowledge, and personal touch.

Where trade names appear, no discrimination is intended, and no endorsement by Penn State Extension is implied.

Resources

Gillis, A. August 7, 2023. 4 main types of artificial intelligence: Explained. TechTarget.

OpenAI. (2024). ChatGPT (June 5 version) [Large language model].

Spence, C. March 31, 2021. Revolution In The Field: Artificial Intelligence leads the way in a sea change for agriculture. University of Florida.

Siocon, G. November 20, 2023. Ways AI is Changing HR Departments. Business News Daily.

Tucci, L. January 25, 2024. A guide to artificial intelligence in the enterprise. TechTarget.

Acknowledgement: This article reviewed by Mona Thompson, Ed.M., Senior Education Development Specialist, Teaching & Learning Transformation Center, University of Maryland.