this post was submitted on 19 Mar 2024

338 points (97.5% liked)

Technology

59311 readers

6308 users here now

This is a most excellent place for technology news and articles.

Our Rules

Follow the lemmy.world rules.
Only tech related content.
Be excellent to each another!
Mod approved content bots can post up to 10 articles per day.
Threads asking for personal tech support may be deleted.
Politics threads may be removed.
No memes allowed as posts, OK to post as comments.
Only approved bots from the list below, to ask if your bot can be added please contact us.
Check for duplicates before posting, duplicates may be removed

Approved Bots

founded 1 year ago

MODERATORS

338

Using AI to spot edible mushrooms could kill you | AI tools are good for some things, but don’t trust your health to apps that make frequent mistakes (wapo.st)

submitted 7 months ago by silence7@slrpnk.net to c/technology@lemmy.world

57 comments fedilink hide all child comments

In particular, know how to identify the common and deadly species (eg: much of the genus Amanita) yourself, and get multiple trustworthy field guides for your part of the world.

you are viewing a single comment's thread
view the rest of the comments

[–] deranger@sh.itjust.works 66 points 7 months ago (6 children)

Of all the things to use machine learning for, identifying poisonous fucking mushrooms seems like a poor choice. I’m sure it sounds very confident in its wrong answer, though.

[–] hedgehog@ttrpg.network 16 points 7 months ago (3 children)

Identifying mushrooms with an ML-based algorithm is a fine idea if you properly design the application to leverage that. As a hedgehog, this is what I would do:

Train my model on a variety of mushrooms, particularly poisonous ones.
When testing the model, test as many mushrooms as possible and take note of what’s frequently mis-identified.
When testing the model, make sure to get a variety of different kinds of lighting.
In addition to the mis-identifications noted while testing the app, maintain a list of commonly misidentified mushrooms - like the hedgehog mushroom and its counterparts - particularly the ones a forager should be most concerned with (meaning the most poisonous ones).
When identifying a mushroom to the user, err on the side of calling it a poisonous mushroom. Consider providing a list of possible matches, with the worst case scenario ones up top.
Include pictures and other information about the mushrooms, as well as regional mushroom lookups for mushrooms that weren’t included.
Don’t include text like “99% confident that this is a hedgehog mushroom” when the 99% figure is an output from your ML model. I know we said earlier to make sure to do a ton of testing and I’m sure you think you did, but you didn’t do enough to be able to say that. At best, reduce your certainty by 25%, then divide that number between the identified mushroom and the lookalikes, making sure to give extra weight to the most poisonous ones. So that 99% certainty becomes at most a more realistic 38% chance that it’s the poisonous lookalike and 37% chance that it’s whatever was identified in the first place.

You might say that this app would be useless for determining if a mushroom is safe to eat, and I agree, but it’s also a better approach than any of the existing apps out there. If you need to use an app to determine if a wild mushroom is safe to eat then the answer is simple: it isn’t. C’mon, I’m a hedgehog and even I know that.

[–] Ultraviolet@lemmy.world 7 points 7 months ago* (last edited 7 months ago) (1 children)

It's probably just a ChatGPT wrapper with a preset prompt. That's all these "AI entrepreneurs" are capable of. Absolute fucking hacks.

[–] Akisamb@programming.dev 3 points 7 months ago

Convolutional neural networks and plant identifying apps came before chat gpt. Beyond both relying on neural networks they don't have much in common.

[–] Knuk@lemmy.world 5 points 7 months ago (1 children)

This comment reads like it was written by AI

[–] hedgehog@ttrpg.network 3 points 7 months ago

This comment reads like it was written by someone who has never designed a mushroom identification app.

[–] KairuByte@lemmy.dbzer0.com 3 points 7 months ago (1 children)

I feel like training on poisonous mushrooms is the wrong direction. You want to err on the side of poisonous, not edible. Anything it can’t identify should be considered poisonous.

[–] hedgehog@ttrpg.network 4 points 7 months ago

Many edible mushrooms have poisonous look-alikes, so your approach would be likely to misidentify those poisonous look-alikes - a potentially deadly mistake.

For example - from https://www.gardeningknowhow.com/edible/vegetables/types-of-edible-mushrooms-their-poisonous-look-alikes

Poisonous Morel Mushroom Look-alikes:

A common fungus, the false morel is almost the spitting image of its edible cousin except it is not hollow inside and contains cottony material.

Big red is similar except it has reddish tones and the cap is more brain-like.

Wrinkled thimble cap truly looks like a morel except its wrinkled cap hangs over the stem.

Bell morel is smaller and the cap, although similar, is much less textured and it has a cottony interior.

It would be easy to train an ML model to confidently identify any of those as morels if you only trained on morels.

The idea is to train on both so it’s less likely to mistake a poisonous mushroom for an edible one, and to then “hedge” your bet anyway, by always presenting the poisonous look-alikes first.

The most dangerous scenario with this app is also the most useful - a user who has some training in mushroom identification uses the app as a quick way to look up a mushroom they think is a particular edible mushroom, notes that the mushroom they think it is is within the list, then reviews the list of poisonous look-alikes, and then applies their training to rule out those look-alikes. Finally they confirm that they cannot rule out the edible mushroom.

The risks here are that

the user’s training is lacking and that they ruled out a poisonous mushroom that the app suggested, or
the app didn’t include the particular poisonous mushroom in the first place and the user was thus unable to consider it.

[–] Imgonnatrythis@sh.itjust.works 8 points 7 months ago (1 children)

This will be how the machines ultimately win.

[–] phdepressed@sh.itjust.works 2 points 7 months ago

Now that is a thought. Instead of AI doing a skynet/terminator thing it gets to the point we trust it and then tells us to eat or do things that kill us.

[–] givesomefucks@lemmy.world 4 points 7 months ago (1 children)

Yeah, I mean there's only going to be a handful of species you want to eat in any location at any time.

You should never just be finding random mushrooms and figuring out if you can eat them.

[–] Drusas@kbin.run 0 points 7 months ago

Finding random mushrooms and learning to identify them (which includes learning if they are edible) is absolutely how you should start in amateur mycology, especially if you don't have any mycology groups nearby that you can join. And if you do, you know what that group will do? Gather random mushrooms to learn/teach identification.

Just don't go around eating random stuff.

[–] Harbinger01173430@lemmy.world 2 points 7 months ago

What about identifying three different types of similar flowers by using the sepal length? 🧐 Very valid machine learning 101 tutorial exercise

[–] Cethin@lemmy.zip 2 points 7 months ago

I would personally trust it if it said it's poisonous, and then do more checks if it said it was edible. It'd be useful for ruling out some options, and maybe to give you a start for further verification. Basically, don't trust it to tell you what you can eat but trust it when it tells you you can't eat something.

[–] acetanilide@lemmy.world 1 points 7 months ago

Probably equally as confident as the people commenting in Facebook groups

Although at least when asking in a group like that you will probably get a bunch of different answers, which should sow some doubt, as opposed to only getting one answer