Mushroom hunting with LLMs: what can go wrong?

stared

Repo for reproducible science: https://github.com/stared/mushroom-hunting-llm-bench

Insanity

Trusting an LLM with a decision that can potentially be fatal is a good Darwin test I suppose. Sadly I don’t think the general populace understands that LLMs are unreliable. (And even people in tech can vastly overestimate the capabilities.. at least judging by the insane spending on them).

atopia

the interesting question is: what is the barrier we need to cross to trust LLMs with stuff like this? we'll probably cross the barrier but I'd have a hard time setting a true percentage success boundary

khazhoux

I think the more important question is: have people really run out of interesting flavors at their local stores and restaurants, that they are so obsessed with eating things they find growing on the ground? I’ll never understand mushroom culture.

_whiteCaps_

Considering that some mushrooms need a spore print to tell them apart, I'd say that you can't ever trust an LLM.

Legend2440

Is it possible to reliably (>99%) identify mushrooms from a single image, with any method? All image datasets have an intrinsic minimum error that comes from the limit of information in the data, regardless of what model or method you use to analyze it. You might not be able to tell apart two similar-looking mushroom species just by looking.

deadbabe

Giving the impossibly high standards people here have for identifying a mushroom, it’s a wonder that they would eat a mushroom at all. You may not trust an LLM, but how would you even trust yourself? Who would you ever trust?

calldacopsidgaf

Love this idea. We should encourage every AI enthusiast to eat wild mushrooms using their sycophantic bullshit machine to determine each one's safety. Who benefits? Society

zikduruqe

There are old mushroom hunters, and there are bold mushroom hunters. There are no old bold mushroom hunters.

RattlesnakeJake

Gotta' love it when a possible answer to an LLM-related question is "Death."

WarmWash

I've used Gemini quiet a bit for foraging. It's definitely a "trust, but verify" situation. It's really good at at least getting you in the right lane to verify, rather than having to comb hundreds of pictures. So even if it's wrong, a short verification usually reveals why it was wrong. I haven't had a situation yet where it was catastrophically wrong (i.e. the mushroom it called out looks nothing like the mushroom imaged). It also will generally warn of lookalikes. I also use it a lot for general plant ID, and it's really impressive there too, with much lower stakes (I'm not eating those).

asveikau

I actually find AI-based plant and mushroom identifiers very interesting. They make being curious about such things as a beginner or layperson more accessible, and you learn some patterns. However it's very obvious that you shouldn't rely on them to decide what to put in your body. All those apps that I have seen (I guess the responsible ones?) will warn you not to do it. That's not what it's for.

julienchastang

Ha! I just had a very interesting conversation with ChatGPT about a mushroom that I found in my front lawn. I am 99% sure it was an edible oyster mushroom which included analysis the of the size, growing substrate (dead ash tree root which turns out to be a useful identifying characteristic -- I learned a lot about mycology in a short time). Plus I don't use fertilizers, herbicide, pesticides, fungicides (obviously) on my lawn so I think it would have been quite tasty. Did I eat it? Hell, no. But then I got to thinking, would I trust a human to correctly identify it? I am actually not sure I would either. Penalty is just too great for misidentification LLM or not.

kwarcode

Nobody ever mentions the fox piss.

theseamusjames

There's been a spike in cases of people eating death cap mushrooms in California this year, and I would bet almost anything that LLM adoption has something to do with it.

tk781

What are the error rates for a human experienced/trained in spotting mushrooms? (isn't that the standard proponents use for autonomous driving?)

ortusdux

Chanterelles and false chanterelles are especially tricky. They fruit at the same time in the same areas. I have to do a 2nd pass after I have washed and dried my haul. https://www.mushroom-appreciation.com/wp-content/uploads/202...

behole

I have been using "new" Siri on iOS similarly but less dangerously. Pointing my camera at dogs at the dog park and seeing how well Siri identifies them. Pretty impressed so far. I am not eating them tho.

soricus

The check can't be just a fast glance because an incorrect answer for sure looks exactly the same as a correct one. From my experience a separate pass with verification of the source is required.

dataviz1000

The single best animated data visualization to demonstrate the stochastic nature of LLM models: an animation of the probability of solving a long multiplication problem over several runs. [0] [0] https://adamsohn.com/reasoning-grid/#walk-the-surface