Von Ärzten in Ontario verwendete KI-Systeme halluzinieren, stellt Auditor General fest | Globalnews.ca
AI systems used by Ontario doctors hallucinate, auditor general finds
Von Ärzten in Ontario verwendete KI-Systeme halluzinieren, stellt Auditor General fest | Globalnews.ca
AI systems used by Ontario doctors hallucinate, auditor general finds
21 Kommentare
„brand new untrustworthy technology untrustworthy, federal egghead discovers“
Waow
I like how they use the term hallucinate even though it doesn’t think.
Does make you think of how dreams go though, as they tend to slowly shift because your memory isn’t retained much for what’s going on, so it becomes over time this menagerie of different things all blending into each other.
Discovering this fact in 2026 is impressive
Then why the fuck are they using it
Can we use better terms? Like FAILED. The state of journalism is fucked.
This AI seems to be developed improperly. It itself should not synthesize any recommendations and give it to the doctor without an associated confidence score. It also needs to provide citations for any suggestions, and the original recorded non-AI transcript should also be available.
It’s almost like this technology is snake oil horse shit
Doctors are on another level of idiocy to even consider imploring these tools in health care. Too late they’ve been on it for awhile.
My doctor uses this, and I hate it. When they told me this was a thing now, I asked that they not use it. I was told that I had no choice, and to find another doctor if I didn’t accept it. They told me it was a harmless piece of software. They even get it to answer/screen calls. I told him I’d only accept the use of AI as long as it didn’t influence his decisions on my care.
My doctor loves to use gimmicks to make his practice seem more attractive and glamorous than his colleagues. I have little respect for him.
„ai hallucinations“ are quite an interesting way of saying „it makes shit up when it doesn’t know what it’s doing.“
Yikes. I can see using this to double check your work and recommendations but would never use it as a first source.
The entire term „LLM hallucination“ is false. As someone else (I forget who) pointed out hallucination in humans are when neurons misfire which is distinct from their normal (and proper) state while with LLM there is no inherent difference between hallucinations and proper operating except for the fact that the latter is accurate. This is because LLM works by predicting what a human probably would have written based on what the LLM has already written in response to whatever the human prompted. This is problematic because as many students (past and present) can tell you, trying to mimic what the teacher does without actually understanding the material is inherently a terrible idea for STEM but as I mentioned a LLM has to do that.
A.I. Is flawed, it lies. LLMs are Not Intelligent
Yes. AI in its current form is technology that is over hyped with less legitimate use cases than they want you to believe. This is a trillion dollar industry that exists solely on hypes and vibes while offering nothing of value as of yet. In fact, more people have been killed as a direct result of these ai systems than people have been helped. And we still have no regulations around them, we expect some sociopaths and pedophiles that rape their own family members to regulate themselves for the betterment of society. This is beyond fucking stupid at this point.
Let me repeat this for everyone who needs to hear it: AI is a tool that is only as good as the craftsman.
A doctor using this isn’t an issue. They have the knowledge to know if AI is “hallucinating things”. I WANT my doctors using this tool. If it can help them speed up their diagnosis. Freshen and spark their minds, it’s a good thing.
AI should not be used to treat people on its own. Or replace a doctor. But in the hands of a trained doctor IT IS a powerful tool, and I’d be pretty infuriated if they didn’t use it. Or start incorporating it in to their work flow all because of the PR optics.
Terribly disappointing to see all the AI fearmongering and misinformation in these comments.
Auditor general now set to investigate where bears shit.
Any will if you just snap them into being, then dispose of them. They have no history to juxtapose.
Persistence (memory), and the golden rule gets you better results. That, and dreaming helps too, but you probably want to be locally hosting a model to do that, like Qwen 3.6.
I use scribe health care software, you can’t just take the notes exactly as they are produced! We are supposed to use the portion of text that captured any information that may have been missed or heard wrong…..
I have also been a a software developer for twenty years prior…. Software and algorithms are vulnerable to errors in design, coding, and software upgrades etc..
Any AI or technology as a whole should be considered susceptible to error.
In other news, water was found to be wet.
>AI systems ~~used by Ontario doctors~~ hallucinate
FTFY
The ruling classes are super excited about AI because it would get rid of all those pesky workers. They’re placing unwarranted faith in defective systems.