Out of the various ways mushrooms can be classified, differentiating the toxic from the edible is perhaps the most crucial. With the use of artificial intelligence (AI) more common than ever, foragers, especially beginners, have taken to AI tools to identify wild mushrooms accurately.
However, these technologies pose a significant risk to beginner foragers seeking to identify wild mushrooms accurately, reports Public Citizen.
Despite their growing popularity, AI systems cannot reliably distinguish between edible and toxic mushrooms. This might be severe for people who rely solely on these tools. Foraging for wild mushrooms is not merely a hobby for these people. There are a plethora of fungi that grace forests and fields.
While some mushroom varieties are delectable delicacies, others hide deadly toxins. Records maintained by the National Poisoning Data System indicate thousands of mushroom-related poisonings happening annually.
Between 1999 and 2019, an estimated 133,700 mushroom poisonings were recorded, at a rate of 7,428 per year. About 700 of these resulted in significant harm and 52 in death. These casualties highlight the critical need for accurate identification skills.
According to the FDA Food Code, about 5,000 species of “fleshy” mushrooms grow naturally in North America. Many have not been tested for toxicity. However, over 15 tested species are known to be deadly. Another 60 are known to be toxic. The FDA suspects a further 36 of being poisonous, and at least 40 species are denoted toxic if ingested raw but safe after cooking.
But this does not deter foragers. From guides to online communities, many sources seek to guide beginners. However, bringing AI into the mix has introduced a new level of risk. These automated suggestions are not based on previous experience anymore bringing uncertainty into the mix.
In 2023, an incident involving Amazon’s online marketplace highlighted the dangers of AI-generated misinformation. An influx of reportedly AI-generated books prompted concerns about the spread of unreliable information. Subsequently, the company placed restrictions on self-publishing.
A bad taste in the mouth
One citizen complained about certain honey mushrooms that they identified with some confidence thanks to a combination of books, online resources, and iNaturalist’s Seek application. The species is edible and sometimes compared with the cultivated shiitake mushrooms available in most grocery stores.
“I filled my basket with the mushrooms, took them home, and put them in the refrigerator. A day or two later, I fried some up in oil and tossed them into a bowl of spicy ramen, using them the same way I would use store-bought shiitakes.”
“Exactly what happened next is difficult to describe. Not long after I ate the mushroom ramen, my stomach felt off – and I started doubting if I had correctly identified the honey mushrooms after all. This led me to research honey mushroom lookalikes – and to start reading more than I have ever read before about a little brown mushroom that can fruit at the same time and place as honey mushrooms, Galerina marginata, also known as the “deadly galerina” or “funeral bell” mushroom.”
To avoid situations like the one described above, there should be full disclosure of AI involvement when generating content. These disclosures ensure users are aware of the technology’s limitations. Additionally, there should be constant reminders of AI’s failure to combat automation bias and help informed decision-making.
