1138

World

AI Tools May Not Transform Agriculture, but Will Help Farmers, Experts Predict

ChatGPT lacks reasoning abilities and cannot predict the future, but it may prove useful for farmers in other ways.
"Olive farmer on a laptop by Vermeer" created by DALL·E OpenAI
By Daniel Dawson
Mar. 1, 2023 14:37 UTC

In the three months since it was first unveiled, OpenAI’s arti­fi­cial intel­li­gence-pow­ered chat­bot has cap­tured the gen­eral pub­lic’s imag­i­na­tion.

OpenAI no longer pub­lishes offi­cial use fig­ures but said 1 mil­lion users signed up for the free ser­vice in the first five days.

We should not really think of this as a mag­i­cal tool that can solve all the prob­lems- Heng Ji, com­puter sci­ence pro­fes­sor, University of Illinois – Urbana-Champaign

The large lan­guage model-based gen­er­a­tor – fed on bil­lions of data points, includ­ing books, news arti­cles and other web con­tent – can instantly answer ques­tions, though not always accu­rately, and mimic cre­ative writ­ing styles.

Since its inau­gu­ra­tion, peo­ple from a wide range of sec­tors have been try­ing to fig­ure out how best to har­ness the unprece­dented power of this arti­fi­cial intel­li­gence to make their busi­nesses more effi­cient.

See Also:Researchers Use AI to Identify EVOO Provenance

ChatGPT is unlikely to rev­o­lu­tion­ize agri­cul­ture, but experts sug­gest that it could help farm­ers with research and tasks, such as writ­ing web­site con­tent and mar­ket­ing mate­r­ial.

Heng Ji, a pro­fes­sor of com­puter sci­ence at the University of Illinois – Urbana-Champaign, told Olive Oil Times that large lan­guage model-based gen­er­a­tors could help farm­ers per­form tasks where human judg­ment can be used to assess the accu­racy of the infor­ma­tion.

Large lan­guage model-based gen­er­a­tor

A large lan­guage model-based gen­er­a­tor is an AI tool that uses machine learn­ing to gen­er­ate text or other types of con­tent based on input prompts. It is trained on vast amounts of text data and uses this knowl­edge to cre­ate new con­tent that resem­bles the input it was given.

But we should not really think of this as a mag­i­cal tool that can solve all the prob­lems,” she said.

For exam­ple, Ji said farm­ers could ask ChatGPT how to plant spe­cific crops in cer­tain regions or cli­matic con­di­tions and expect to receive accu­rate advice.

ChatGPT could also sum­ma­rize the find­ings of sci­en­tific stud­ies or pro­vide prac­ti­cal infor­ma­tion, such as how to repair a piece of equip­ment.

A large lan­guage model is just mem­o­riz­ing a sequence of tokens,” she said. It is about the aggre­ga­tion of human expe­ri­ences from the past. It’s basi­cally gath­ered all the data from the whole web before 2021 and mem­o­rized it.”

Sequence of tokens

A sequence of tokens” means a sequence of smaller units of mean­ing, called tokens,” that make up a larger piece of text or code. Tokens are usu­ally words or punc­tu­a­tion marks. Breaking down text or code into sequences of tokens makes it eas­ier to ana­lyze and process, and is used as inputs for machine learn­ing mod­els to gen­er­ate new text or clas­sify it into dif­fer­ent cat­e­gories.

Ji warned that large lan­guage model-based gen­er­a­tors – despite what they claim when asked – can­not engage in the process known as sci­en­tific dis­cov­ery.

For exam­ple, ChatGPT says it can help farm­ers by pre­dict­ing future cli­matic con­di­tions and mar­ket prices. However, Ji explained why this is not pos­si­ble.

Advertisement
Advertisement

It does not have deep rea­son­ing abil­i­ties,” she said. It can­not pre­dict the future. If you are ask­ing the model to dis­cover some­thing about a new con­di­tion or some tasks that it’s never observed before, it’s not going to do that [accu­rately].”

It will gen­er­ate an answer in flu­ent English, but there’s no knowl­edge ground­ing capa­bil­ity, and there’s no guar­an­tee about the truth­ful­ness of the answer,” Ji added. The sys­tem was not designed for rea­son­ing or dis­cov­er­ies.”

Todd Janzen, an attor­ney spe­cial­iz­ing in agri­cul­tural law, agrees. He asked ChatGPT: What are the top five ways that ChatGTP will rev­o­lu­tion­ize agri­cul­ture in the United States?” Among the responses, ChatGPT said it could use its abil­i­ties for data analy­sis and pre­dic­tions.

If you knew noth­ing about agri­cul­ture, you would think ChatGPT’s pre­dic­tions were pretty amaz­ing,” Janzen wrote in Successful Farming mag­a­zine. These pre­dic­tions come across as very author­i­ta­tive and knowl­edgable.”

But peel back the veneer, and you find a kind of word salad that sounds impres­sive but lacks much depth or mean­ing,” he added. Most of these five points are just regur­gi­tat­ing the same con­cepts: data analy­sis and pre­dic­tion.”

Before ChatGPT can become a rev­o­lu­tion­ary tool for farm­ers, Ji believes that the prob­lem around the accu­racy of its answers must first be fixed.

She added that researchers are already work­ing on their own large lan­guage model-based gen­er­a­tors. One specif­i­cally geared toward agri­cul­ture could be cre­ated, where all the responses would be based on con­firmed knowl­edge. However, even this would still be unable to pre­dict the future.

While ChatGPT and other sim­i­lar AI chat­bots are unlikely to change how farm­ers do their jobs, the rapid advance of the tech­nol­ogy may still have some pro­found con­se­quences for the sec­tor.

My expe­ri­ence is that the agri­food sec­tor is way more frag­ile than oth­ers in terms of poten­tial biases intro­duced by the AI tools such as ChatGPT,” Yu Jiang, an agritech researcher at Cornell University, told Olive Oil Times.

It’s impor­tant not only to con­sider how to let grow­ers use ChatGPT but also how to develop dig­i­tal con­tent strate­gi­cally (in col­lab­o­ra­tion with ChatGPT) to influ­ence the mod­els,” he added.

For exam­ple, Jiang’s research with the AI chat­bot has demon­strated that its answers to ques­tions about wine rec­om­men­da­tions are biased toward com­pa­nies with exten­sive dig­i­tal foot­prints.

If some­one uses ChatGPT APIs [a tool that allows dif­fer­ent soft­ware or appli­ca­tions to com­mu­ni­cate] to cre­ate a rec­om­men­da­tion app, espe­cially for new cus­tomers, these small winer­ies would lose their mar­ket share very fast, result­ing in fur­ther con­sol­i­da­tion of the agri­food com­mu­nity,” he said.

With the inte­gra­tion of ChatGPT into Bing, a search engine, the shift to AI-based search looks increas­ingly inevitable. Jiang warned that small farm­ers must adapt quickly or risk being left behind.

If diver­sity is a part of resilience, we should encour­age the grow­ers to start to think actively with these new AI tools that are not designed for agri­cul­ture and food appli­ca­tions and build dig­i­tal lit­er­acy and train­ing mate­ri­als for grow­ers to embrace the change,” he con­cluded.



Advertisement
Advertisement

Related Articles