The Power of AI in Plant Breeding: Leveraging AI Assistants like ChatGPT

Hello, dedicated plant breeders!

As we explore the dynamic field of plant breeding, it’s crucial to stay on the cutting edge of technology and utilize resources that can enhance our efforts. In this discussion, we’ll explore the remarkable advantages of AI, particularly AI assistants like ChatGPT, in assisting plant breeders with vital aspects of their work.

  1. Comprehensive Plant Information:
    AI, such as ChatGPT, provides a wealth of knowledge on a wide range of plant species, including details like chromosome count, habitat, cold hardiness, related species, natural distribution, and germination requirements. This wealth of information can be invaluable when selecting parent plants for breeding programs.

  2. Data Organization and Presentation:
    AI can process bulk information and present it in easily digestible formats, such as tables. This feature is a time-saver for plant breeders who need to manage extensive datasets related to breeding projects and plant characteristics.

  3. Plant Breeding Insights:
    AI can provide valuable information on the history of plant breeding work on specific varieties, including the methods used and the outcomes achieved. This historical context can inform our breeding strategies and help us avoid repeating past mistakes.

  4. Accelerating Crossbreeding:
    AI can simulate and optimize crossbreeding scenarios based on the data it has accumulated. This ability to predict outcomes aids breeders in making informed decisions about which plant varieties to cross for desired traits.

  5. Disease and Pest Management:
    AI’s predictive capabilities are not limited to breeding but extend to analyzing and forecasting disease and pest outbreaks, enabling us to take proactive measures to protect our crops and plan for disease-resistant plant varieties.

  6. Optimal Plant Selection:
    AI can assist in selecting the most promising parent plants for breeding by analyzing their genetic makeup and compatibility. This helps in ensuring that the desired traits are more likely to be passed on to the offspring.

  7. Genetic Marker Identification:
    AI can aid in the identification of specific genetic markers associated with desirable traits. This can streamline the breeding process by allowing breeders to focus on plants with the highest potential for desired outcomes.

  8. Global Collaboration:
    AI-powered platforms like ChatGPT facilitate global collaboration among plant breeders. This enables the sharing of knowledge and experiences, leading to more rapid advancements in plant breeding techniques.

  9. Access to the Latest Research:
    AI can provide real-time access to the latest research and developments in the field of plant breeding. This ensures that breeders are always up-to-date with the most current techniques and discoveries.

  10. Educational Resource:
    AI assistants like ChatGPT can serve as valuable educational resources for both novice and experienced plant breeders. They can answer questions, provide guidance, and offer insights into best practices.

As we embrace the potential of AI in plant breeding, we can harness these AI-powered assistants to improve the precision, efficiency, and success of our breeding programs. Let’s discuss how we can leverage AI for the benefit of our field!

1 Like

FYI: The write-up above was generated using ChatGPT.

While not 100% accurate, I have found ChatGPT to be an incredible resource for not only researching plants but also processing data. It’s worth noting that ChatGPT can assist in creating tables that compare various plant species and even export them into CSV format.

For instance, I recently needed to compare species within the Prunus genus, specifically those rated as zone 3 cold hardy. ChatGPT efficiently compiled a list of all the relevant species. I then requested ChatGPT to create a table with essential features for each of these species, such as chromosome count, known cultivars, and fruit quality. In just a few minutes, a list of 10 or more species with their important features was readily generated. It’s truly remarkable!

Have you been utilizing this fantastic, freely available technology?

https://chat.openai.com

AI can be useful if you understand that AI has a fundamental underlying hallucinations problem inherent in it’s design. Each implementation has various techniques applied by it’s designers to try and reduce or limit the hallucinations but the risk remains that it just makes things up and presents them as truths during answering at times which is why there is all the disclaimers that are given before giving AI generated results to limit legal liability of the AI providers. You have to have a human experienced in the subject matter area of the questions and answers proofread all of the AI output and manually correct hallucinations and factual inaccuracies. It’s a time saver but not a fire off a question and accept answer solution right now.

2 Likes

Thanks Zach, very thought provoking. I’ve been using chat gpt for grant writing. (Luckily good editors in this group :)) I’d be interested in reading one of the conversations you referenced about plums and plant species tables. I tried to generate a table of 100 species including things like rate of cross pollination but didn’t get that far.

1 Like

One of my yoga classes uses AI generated content. It feels ghastly. Mostly because of inane repetition that a human would never do.

The one plant breeding AI article that someone sent me contained many factual inaccuracies.

I like thinking about the hallucinations inside the machine…

1 Like

I guess AI might be useful to a breeder wanting something very specific but I can’t really see it being of benefit in my own efforts at landrace development. I grow plants together. They either cross or they don’t. If I’m really keen on the crop I’ll hand pollinate to effect some crosses. Tomatoes come to mind here. Other than that, they’re on their own.

1 Like

Absolutely, you’ve raised a valid point. AI can sometimes generate nonsense or present contradictions, and it’s not immune to inaccuracies. However, as you rightly mentioned, these challenges can be found in various sources of information.

What I’m suggesting is that it’s worth giving AI a try. Understanding how to utilize this technology effectively is crucial. It’s a learning process, and we can think of it as an iterative one.

Over time, AI technology will undoubtedly improve. The datasets it relies on are continuously expanding and being upgraded, which bodes well for its accuracy and reliability.

AI won’t replace the actual work of creating the perfect yoga session or breeding the ideal plant variety, but it can indeed serve as a valuable tool in the process. Its potential to assist and augment our efforts is something we should explore and harness.

I haven’t, but I’ve been meaning to for the last year.

So I just went and played with it for a bit, coming up with this thread of conversation: https://chat.openai.com/share/40255d5c-4f54-4daa-a9e8-1ee7aa8232a1

There’s some summary info that might be superior to browsing Wikipedia, but it seemed too vague to be really useful. Also, there’s a lot of “might” involved that I find off-putting. If you, or anyone, wanted to create and share a thread of Q&A like that you think demonstrates good ways to use the tool (refining inputs and drilling into better answers, specifically), I’d enjoy looking at one.

1 Like

I’ve prepared a sample conversation to use with ChatGPT. Please copy and paste each line from my sample conversation, and let ChatGPT respond. You can adjust the location and property size as needed.

Hello ChatGPT, I am working on a project.

I am considering the purchase of a 20-acre parcel of land in Salcha, AK. We intend to transform it into a food forest park.

Can you provide me with relevant information about the weather and climate?

Could you please create a table with the following information for Salcha, AK or the closest know area:

  • Annual precipitation
  • Average temperature for each season
  • Number of frost-free days
  • Hardiness zone rating
  • Length of day for winter and summer solstices
  • Longitude and latitude
  • Any other vital data for crop cultivation

(If ChatGPT says Data Required)
What is the closest town that you have data for?
Please use that location.

Please export the table to a CSV format.

Thank you. Now, I would like a list of fruit varieties that have the potential to be cultivated.

Could you also create a table with the following details for each fruit species:

  • Cultivars
  • Hardiness rating
  • Preferred growing environment
  • Growth structure
  • Age of maturity
  • Longevity
  • Any other pertinent information for cultivating the crops.

Please export the table to a CSV format.

Are there any other locations with a similar climate?

Please list culturally important food crops for each region.

Please list the traditional herbs and spices grown in each region.

What are some customary food dishes for each region? Could you provide descriptions of their ingredients?

That was an engaging conversation. Thank you!!

The important thing to remember about AI is that it is a learning tool–meaning, it learns over time.

A human being has hundreds of millions of data points used in making any decision, many of them subconscious. The decision making process is highly tuned and carefully designed to come up with a correct answer or appropriate response in (hopefully) every case. The human brain is capable of putting together seemingly unrelated data from thousands of different topics in this process.

An AI does not have the capacity to determine if its answers are logical or accurate, or know where to look to get additional information if necessary. It’s accuracy is limited to the accuracy of its programmed data pool, and in most cases that data pool is strictly limited.

In the future developers may find ways to fix these problems, but at the moment anything AI tells you has to be gone over for factual problems and logical fallacies.

In the meantime we can laugh at the AI bots putting 5 arms on a picture of a child, or telling us that Saskatchewan is the capital of China.

I think there is a lot of potential from AI assistants. Technical writing has been a big part of my life, and I find the output of chat-gpt to be more not unlike my own work when I’m writing about a domain where I haven’t worked much myself.

Here are some of the prompts that have produced results I enjoyed. Looking at them, it seems like one of my approaches is to ask the AI to cross-reference two or more domains, and then prioritize the results.

please list all north american native culinary plants that could be used as a substitute for salvia officinalis. note whether they are known to grow in habitats like floyd county kentucky

Please make a list of as many culinary herbs as possible that feature foliage that is not green. For example, foliage that is purple, blue, black, red. Please order the list in approximate order of global popularity across all ethnic cuisines. Each item in the list should have at least a common name, scientific binomial name, and a full sentence or two about its attributes and use

in the united states, the land grant universities operate county agricultural extension programs in order to provide agronomic services and training to rural farmers. what are some equivalents to the agricultural extension program that exist within european countries?

please compile a list of foods and agricultural products that can be "forest farmed" in southeastern kentucky, along with brief notes on the kinds of recipes that they can be used in

Compare the oil content of the flesh of winter squash and pumpkin to the oil content of the flesh of avocado

3 Likes

I’m a little blown away by Google’s Notebook LM today.

The capacity to turn my original, pretty low quality french audio files into something readable might be expected,

but check out this podcast I just made, one click, for example.

Next podcast idea, anyone?
But also, cheers to this very interesting work by Antoine Emmanual, going where I haven’t seen anybody else venture, yet.

Do you know Julia that, by ““coincidence””", that happened:

…i.e.: 2,5hours train from Paris to Rennes 2 days ago.
Many told me about Emmanuel over the past few weeks, including days in Antibes, I didn’t see who he was… but when seating in the train, smiling, he introduced himself… 2 hours of translations to English and to French for me, very enjoyable!

I heard this story, amazing. I’m looking forward to seeing where this all goes.

it’s funny one night during a meal in Antibes I found myself next to him and we talked well…a very nice guy who is still obviously crossing the road GTS :grin:

1 Like