Welcome to DU!
The truly grassroots left-of-center political community where regular people, not algorithms, drive the discussions and set the standards.
Join the community:
Create a free account
Support DU (and get rid of ads!):
Become a Star Member
Latest Breaking News
Editorials & Other Articles
General Discussion
The DU Lounge
All Forums
Issue Forums
Culture Forums
Alliance Forums
Region Forums
Support Forums
Help & Search
General Discussion
In reply to the discussion: I was thinking about this "AI" stuff . . . [View all]highplainsdem
(59,323 posts)26. Re hallucinations - see this article:
Gemini 3 Pro tops new AI reliability benchmark, but hallucination rates remain high
https://the-decoder.com/gemini-3-pro-tops-new-ai-reliability-benchmark-but-hallucination-rates-remain-high/
No single model demonstrated consistently strong factual reliability across all six domains. Claude 4.1 Opus led in law, software engineering, and the humanities; GPT‑5.1.1 ranked first in business questions; while Grok 4 performed best in health and science.
According to the study, these domain differences mean that relying solely on overall performance can obscure important gaps.
While larger models tend to achieve higher accuracy, they dont necessarily have lower hallucination rates. Several smaller models - like Nvidias Nemotron Nano 9B V2 and Llama Nemotron Super 49B v1.5 - outperformed much larger competitors on the Omniscience Index.
Artificial Analysis confirmed that accuracy strongly correlates with model size, but hallucination rate does not. That explains why Gemini 3 Pro, despite its high accuracy, still hallucinates frequently.
According to the study, these domain differences mean that relying solely on overall performance can obscure important gaps.
While larger models tend to achieve higher accuracy, they dont necessarily have lower hallucination rates. Several smaller models - like Nvidias Nemotron Nano 9B V2 and Llama Nemotron Super 49B v1.5 - outperformed much larger competitors on the Omniscience Index.
Artificial Analysis confirmed that accuracy strongly correlates with model size, but hallucination rate does not. That explains why Gemini 3 Pro, despite its high accuracy, still hallucinates frequently.
Edit history
Please sign in to view edit histories.
Recommendations
0 members have recommended this reply (displayed in chronological order):
69 replies
= new reply since forum marked as read
Highlight:
NoneDon't highlight anything
5 newestHighlight 5 most recent replies
RecommendedHighlight replies with 5 or more recommendations
False premise. NASA and the other entities involved used computers. They did not only use slide rules.
Celerity
Sunday
#4
Didn't John Glenn ask the women mathmaticians of "Hiden Figures" to do manual calculations
Deminpenn
Sunday
#5
Never saw that movie, but you said 'check the computer calculations' so computers were obviously used to a degree.
Celerity
Sunday
#7
No, the women checked the computers' outputs. Also see comments in this thread confirming that computers were used
Celerity
Monday
#58
Then anyone could "write" such a thesis because it would require minimal knowledge and the AI
highplainsdem
Sunday
#16
You can't enhance creativity with AI, any more than you enhance creativity asking someone else to
highplainsdem
Sunday
#9
Curious about what you mean when you say it inspires you. Do you mean you ask it for ideas?
highplainsdem
Sunday
#18
Okay, I'll give you an A+ for creativity just for writing a poem for a science communication workshop.
highplainsdem
Sunday
#49
Yours is one of the few nuanced takes I've read about one of the major faults with AI...
appmanga
Sunday
#54
Thanks, but I'm just trying to relay some of what I've heard from artists and writers and others
highplainsdem
Monday
#61
It isn't at all cool that AI is being widely used for cheating and students are learning less as a
highplainsdem
Sunday
#19
GenAI is never hallucination-free. I don't know where you got the idea that it is.
highplainsdem
Sunday
#23
It wasn't that long ago that Grok was identifying him as the main source of misinformation on X,
highplainsdem
Sunday
#46
You just contradicted what you said minutes ago about it being hallucination-free.
highplainsdem
Sunday
#31
I specifically said history topics along with other disciplines that don't change and are "set"
WarGamer
Sunday
#33
50 years ago if I told you I could hold a piece of glass and access global knowledge...
WarGamer
Sunday
#21
You don't know if it was "dead accurate" unless you took the time to check that those were the
highplainsdem
Sunday
#24
I find this discussion fascinating. It seems that the algorithm has figured out people are inherently lazy learners.
cayugafalls
Sunday
#38