Farmer ends up killing 25 acres of his crops after trusting AI on herbicide advice

A 67-year-old farmer in Chuzhou, China used an AI to find a solution to deal with weeds and pests The chatbot provided incorrect information which the farmer did not verify 24.7 acres of sesame seedlings were inadvertently destroyed A farmer in the Chinese city area of Chuzhou has learned the expens

<![CDATA[ <article> <ul><li><strong>A 67-year-old farmer in Chuzhou, China used an AI to find a solution to deal with weeds and pests</strong></li><li><strong>The chatbot provided incorrect information which the farmer did not verify</strong></li><li><strong>24.7 acres of sesame seedlings were inadvertently destroyed</strong></li></ul><p>A farmer in the Chinese city area of Chuzhou has learned the expensive cost of relying on unverified AI responses to problems, after its advice destroyed almost 25 acres of crop. The 67-year-old, who had previously expressed misgivings about using AI, was provided with incorrect information by an unnamed chatbot, and failed to verify it before proceeding.</p><p>The response to the query – concerning the composition of a pest control agent – was taken at face value, with expensive consequences for the farmer, known as Wu.</p><p>A destructive mix of chemicals was recommended by the chatbot, destroying crops as well as weeds and pests, raising questions once more about whether people should blindly trust the responses from AI chatbots.</p><h2 id="an-ai-instructs">An AI instructs</h2><p>Wu’s problem concerned weeds and pests threatening his sesame crop, planted across 24.7 acres. </p><p>Having being initially wary of AI chatbots, his recent experiences had given him some confidence in the technology, he told a local reporter, so asked for help. </p><p>The response from the chatbot was to mix “high-efficiency flupyrimethalin” and “flusulfasulfaether” to kill the weeds, and mix it with “thiamethoxazine” and “methyl salt” – a combination that destroyed the sesame seedlings overnight. </p><p>Simply verifying the response with an AI-free search, or contacting agricultural technicians, would have avoided the destruction of the crop. </p><p>Speaking to local press, Wu said: “If you spray it, the next day the seedlings won’t survive. Both the grass and the seedlings will die, and the seedlings will die even faster.”</p><p>Wu returned to the AI to ask what had gone wrong. The main problem with the chemical mix (a practice that should never be attempted without knowledge) was the inclusion of flusulfasulfaether, something that experts have confirmed. </p><p>The chemical is used against broadleaf weeds, but sesame is widely classified as a broadleaf plant. Targeted spraying is also recommended for the chemical, but Wu’s approach was to spray the entire field.</p><h2 id="verification-please">Verification, please</h2><p>Errors, mistakes, incorrect information, <a href="https://www.techradar.com/pro/pwc-left-red-faced-after-being-caught-using-ai-hallucinations-and-fake-citations-in-multiple-reports">vibe citations</a>, and hallucinations are all risks of using a chatbot for answers or generating material. LLM and chatbot developers know this, and spend a lot of time fixing these issues. As a result, AI tools invariably come with a warning.</p><p>In this case, Wu’s chatbot displayed the warning: “AI generation may be incorrect, please verify.” Wu’s failure to verify has lost him almost 25 acres of crop, and potentially the loss of the land’s use for at least a season, if not longer.</p><p>All material produced by an AI should be verified, whether a simple answer or a solution to a workflow. </p> </article> ]]>