The Machine That Can't Explain ItselfFrom 2024 to 2025, the Foundation Model Transparency Index published by the Stanford Institute for Human-Centered AI dropped from 58/100 to 40/100. Oops.Why AI adoption is outrunning accountability and what happens when the auditors arrive
In my very 1st anti-LLM rant, I described LLMs as "indecipherable oracles" - I don't see how that's going to change.
This next I find almost unbelievable. LLMs will tell you what's in an image without seeing the image?!?!? They just confabulate what's likely to be in the image?!?!? Holy crap, now that's some bullshit! From Gary Marcus:
The mirage of visual understanding in current frontier modelsMarcus is reporting a research paper just released by Stanford:When a model achieves a “top rank on a standard chest X-ray question-answering benchmark without access to any images” you know something is deeply wrong.
MIRAGE: The Illusion of Visual UnderstandingAnother great new term in the dream world of LLM hallucinations: "mirage reasoning".First, Frontier models readily generate detailed image descriptions and elaborate reasoning traces, including pathology-biased clinical findings, for images never provided; we term this phenomenon mirage reasoning. Second, without any image input, models also attain strikingly high scores across general and medical multimodal benchmarks, bringing into question their utility and design. In the most extreme case, our model achieved the top rank on a standard chest X-ray question-answering benchmark without access to any images. Third, when models were explicitly instructed to guess answers without image access, rather than being implicitly prompted to assume images were present, performance declined markedly.
This tech is clearly NOT ready for prime time, but, $T invested, tough shit. You will take your bullshit sandwich & enjoy eating it, nom-nom!
I've downloaded the PDF from arxiv, reading through it, it's scary but also LOL. I shouldn't laugh, the Bullshit Apocalypse is not funny, but laughter has always been a part of what keeps me sane. So, laugh, dumbass, laugh!
Here's the home page/directory for my posts on Bullshit. This is post #113.