502
We have to stop ignoring AI’s hallucination problem
(www.theverge.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 17 May 2024
502 points (94.8% liked)
Technology
55938 readers
3632 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
What I don’t like about it is that it makes it sound more benign than it is. Which also points to who decided to use that term - AI promoters/proponents.
Edit: it’s like all of the bills/acts in congress where they name them something like “The Protect Children Online Act” and you ask, “well, what does it do?” And they say something like, “it lets local police read all of your messages so they can look for any dangers to children.”
The term "hallucination" has been used for years in AI/ML academia. I reading about AI hallucinations ten years ago when I was in college. The term was originally coined by researchers and mathematicians, not the snake oil salesman pushing AI today.
I had no idea about this. I studied neural networks briefly over 10 years ago, but hadn’t heard the term until the last year or two.
We were talking about when it was coined, not when you heard it first
In terms of LLM hallucination, it feels like the name very aptly describes the behavior and severity. It doesn't downplay what's happening because it's generally accepted that having a source of information hallucinate is bad.
I feel like the alternatives would downplay the problem. A "glitch" is generic and common, "lying" is just inaccurate since that implies intent to deceive, and just being "wrong" doesn't get across how elaborately wrong an LLM can be.
Hallucination fits pretty well and is also pretty evocative. I doubt that AI promoters want to effectively call their product schizophrenic, which is what most people think when hearing hallucination.
Ultmately all the sciences are full of analogous names to make conversations easier, it's not always marketing. No different than when physicists say particles have "spin" or "color" or that spacetime is a "fabric" or [insert entirety of String theory]...
After thinking about it more, I think the main issue I have with it is that it sort of anthropomorphises the AI, which is more of an issue in applications where you’re trying to convince the consumer that the product is actually intelligent. (Edit: in the human sense of intelligence rather than what we’ve seen associated with technology in the past.)
You may be right that people could have a negative view of the word “hallucination”. I don’t personally think of schizophrenia, but I don’t know what the majority think of when they hear the word.
You could invent a new word, but that doesn't help people understand the problem.
You are looking for an existing word that describes providing unintentionally incorrect thoughts but is totally unrelated to humans. I suspect that word doesn't exist. Every thinking word gets anthropomorphized.