Written by 10:56 am AI problems, Discussions

### Leveraging AI for Mushroom Identification: Mitigating Harmful Consequences

A new report warns that AI can’t be relied upon for mushroom identification, which is way more comp…

Don’t rely on AI mushroom identifiers for your safety, consumer advocates warn. Illustration.

If you have a passion for exotic mushroom hunting and foraging in the great outdoors, go for it. However, when it comes to deciding whether the mushrooms you’ve collected are safe to eat, don’t put your trust in AI.

The Fujifilm X100VI stands out as the best lens I’ve ever possessed.

A recent detailed report from Public Citizen, a nonprofit organization dedicated to consumer advocacy, advises amateur mushroom hunters against depending on artificial intelligence for identifying edible fungi. The report stresses that recognizing mushrooms is a risky endeavor that necessitates practical skills. Public Citizen issued this warning in response to the growing interest and online searches for mushroom identification software and applications, as indicated by data from Google Trends.

While Rick Claypool, the research director at Public Citizen, acknowledges the potential usefulness of AI-powered apps for educational purposes, he points out their limitations when it comes to accurately identifying mushrooms. Merely taking a photo of a mushroom and running it through an app is inadequate.

In the report, Claypool emphasizes that seasoned local foragers understand the irreplaceable value of firsthand experience in finding, observing, smelling, touching, and sometimes even tasting wild mushrooms in their natural habitat. Local knowledge is crucial due to the significant variations in mushroom species across different geographical regions.

For instance, according to Claypool, relying solely on a photo of a mushroom’s cap is rarely sufficient for a confident species identification. It is essential to examine features on the underside of the cap, the length of the stem, and the base of the fungus. Additionally, foragers need to take note of the mushroom’s habitat, whether it grows on the ground or on hardwood, and the types of trees it associates with.

The report highlights that North America is home to over 5,000 species of fleshy mushrooms, with much of their characteristics still unknown. Out of these, 75 species are confirmed to be poisonous or lethal, while 36 are suspected to be harmful. Moreover, 40 species are known to be toxic if consumed raw but safe for consumption after cooking.

Between 1999 and 2016, there were at least 133,000 reported cases of mushroom poisoning, resulting in 704 severe injuries and 52 deaths.

Claypool also mentions that the research on the accuracy of mushroom identification apps is limited and not very promising. A study conducted in 2022 by Australian poison experts evaluated three such AI applications and found that they correctly identified wild mushrooms only around 50% of the time. In some cases, toxic mushrooms were mistakenly identified as edible.

However, the issues with AI’s mushroom identification capabilities extend beyond apps. Claypool notes that in 2023, Amazon was flooded with AI-generated mushroom foraging books, many of which contained incorrect and potentially hazardous information. Some of these books lacked clear evidence of being AI-generated, as reported by 404 Media at the time.

In a blog post in August 2023, the New York Mycological Society raised concerns about the proliferation of AI-generated foraging and identification books on Amazon and other retail platforms. They emphasized the critical importance of purchasing books authored by reputable experts and foragers to avoid potentially life-threatening misinformation.

According to the Public Citizen review, the integration of AI into research poses additional risks. For instance, Google provided a user with detailed instructions on how to prepare Amanita ocreata, a poisonous mushroom, in the following year. Furthermore, generative AI has been known to produce inaccurate images of specific mushroom species, potentially leading amateur foragers astray, as highlighted by Claypool.

Claypool argues that mushroom identification serves as a prime example of the dangers associated with excessive reliance on artificial intelligence.

“No matter the promises made by companies promoting AI technology, the automation of individual expertise is currently unattainable—and may remain so,” he concludes.

Visited 3 times, 1 visit(s) today
Tags: , Last modified: March 22, 2024
Close Search Window