Applications and services for identifying mushrooms have appeared
recently. The user gives a photo as input and it will tell the
mushroom's name, along with information on whether it is edible or
not. Such programs claim to be driven by "artificial intelligemce."
Experts warn that these tools may make mistakes which can potentially
sicken or even kill a user. The error can happen at two levels.
First, given an input image, the program might wrongly identify the
species. Second, an error in the database may tell users that a
certain kind of mushroom is edible, when in fact it is not. The
latter issue is complicated for several factors influence toxidity:
region, season and method of cooking. Some mushrooms are known to
cause problems when consumed with alcohol.
We can tell people about the hazards of non-free software using these
"artificial intelligence" systems as examples. One may want to modify
the program to correctly warn the user that the mushroom in the photo
looks toxic, or examine the code to determine why it makes mistakes,
but non-free software does not allow that.
Another issue I would like to raise is that many people erroneously
believe that "artificial intelligence" is an improved form of software.
Conventional programs are accurate when properly written, but "AI"
systems work on a different principle which cannot guarantee accuracy.
Many people fail to understand this. They think: "Computers are accurate.
AI makes computers better so one with AI must also be accurate."
We are discussing single-purpose systems here, but sooner or later we
will have general-purpose image identification systems which claim to
be capable of properly identifying most things around us, including wild
mushrooms, herbs, snakes, jellyfish and all else.
Relying on AI to Identify Mushrooms Could Kill You
https://gizmodo.com/ai-mushroom-id-dangerous-consumer-advocates-warn-1851355484
Mushrooming Risk: Unreliable A.I. Tools Generate Mushroom
Misinformation - Public Citizen
(long article)
https://www.citizen.org/article/mushroom-risk-ai-app-misinformation/
_______________________________________________
libreplanet-discuss mailing list
libreplanet-discuss@libreplanet.org
https://lists.libreplanet.org/mailman/listinfo/libreplanet-discuss