I wonder how well Alpha Geometry would do on this
Vision language models are blind
31–40 of 202 posts
Re: Vision language models are blind
#32I was hoping that someone in the comments talking the paper down would have published a paper or have had relevant publications of their own to point to. You know, meet the lads halfway sort of thing.
So what I’m left with to judge instead is anonymous online commenters vs. the publication of 2 prestigious universities. Whose word do I take on this? Decisions, decisions.
You can swap LM out with Web3 out with NFT out with Crypto in this case.
Re: Vision language models are blind
#33Earlier quoted context omitted.
I don't see how this is "embarrassing" in the slightest. These models are not human brains, and the fact that people equate them with human brains is an embarrassing failure of the humans more than anything about the models. It's entirely unsurprising that there are numerous cases that these models can't handle that are "obvious to humans." Machine learning has had this property since its invention and it's a classic…
It is embarrassingly, shockingly bad, because these models are advertised and sold as being capable of understanding images. Evidently, all these models still fall short.
Not only are they capable of understanding images(the kind people might actually feed into such a system - photographs), but they're pretty good at it.
A modern robot would struggle to fold socks and put them in a drawer, but they're great at making cars.
Re: Vision language models are blind
#34Earlier quoted context omitted.
I don't see how this is "embarrassing" in the slightest. These models are not human brains, and the fact that people equate them with human brains is an embarrassing failure of the humans more than anything about the models. It's entirely unsurprising that there are numerous cases that these models can't handle that are "obvious to humans." Machine learning has had this property since its invention and it's a classic…
These models are marketed as being able to guide the blind or tutoring children using direct camera access. Promoting those use cases and models failing in these ways is irresponsible. So, yeah, maybe the models are not embarrasing but the hype definitely is.
> Promoting those use cases and models failing in these ways is irresponsible.
Yes, exactly.
Re: Vision language models are blind
#35A few comments below talk about how tokenizing images using stuff like CLIP de-facto yields blurry image descriptions, and so these are ‘blind’ by some definitions. Another angle of blurring not much discussed is that the images are rescaled down; different resolutions for different models. I wouldn’t be surprised if Sonnet 3.5 had a higher res base image it feeds in to the model.
Either way, I would guess that we’ll need new model architectures for multimodal to get really good at some of this, and even then some of these tasks are adjacent to things that we know LLMs are already bad at (numeric logic, for instance).
As context lengths get longer, devoting more tokens to the image tokenization should help a bit here as well. Anyway, I’d anticipate next year we’d see 80s and 90s for most of these scores with next gen models.
Re: Vision language models are blind
#36Well, all the models (especially Claude 3.5 Sonnet) seem to perform much better than random, so they are clearly not blind. The only task where Claude Sonnet 3.5 does not perform better than random is the one where you have to follow many different paths (the ones where the answer from A to C is 3), something that would take me several seconds to solve. I have the feeling that they first choose the title of the paper…
OK. They're legally blind.
Re: Vision language models are blind
#37I've been generally frustrated at the lack of analysis of vision LLMs generally. They're clearly a very exciting category of technology, and a pretty recent one - they only got good last October with GPT-4 Vision, but since then we've had more vision models from Anthropic and Google Gemini. There's so much more information about there about text prompting compared to image prompting. I feel starved for useful informa…
Why not use them yourself if you have access? I have been using Claude 3.5 Sonnet for gardening recently, and while it's not perfect(and can be a little blind unless you tell it to focus on a specific thing), it's helped me understand how to keep my plants alive in some challenging conditions(for me; this is my second or third attempt at gardening so it's all challenging lol). But just a experiment with it and see wh…
I've used them a bunch. I want to learn from other people's experiences as well.
Some of my notes so far:
- https://simonwillison.net/2024/Apr/17/ai-for-data-journalism... - my datasette-extract plugin, for structured data from both text and images
- https://simonwillison.net/2024/Apr/17/ai-for-data-journalism... - where they failed to extract data from a handwritten scanned document in various weird ways
- https://simonwillison.net/2024/Feb/21/gemini-pro-video/ talks about video inputs to Gemini Pro (which are actually image inputs, it splits them up to one frame per second)
Re: Vision language models are blind
#38Well, all the models (especially Claude 3.5 Sonnet) seem to perform much better than random, so they are clearly not blind. The only task where Claude Sonnet 3.5 does not perform better than random is the one where you have to follow many different paths (the ones where the answer from A to C is 3), something that would take me several seconds to solve. I have the feeling that they first choose the title of the paper…
OK. They're legally blind.
This really has nothing to do with vision impairment.
Re: Vision language models are blind
#39A few comments below talk about how tokenizing images using stuff like CLIP de-facto yields blurry image descriptions, and so these are ‘blind’ by some definitions. Another angle of blurring not much discussed is that the images are rescaled down; different resolutions for different models. I wouldn’t be surprised if Sonnet 3.5 had a higher res base image it feeds in to the model. Either way, I would guess that we’ll…
The problem with the current crop of projectors such as LLaVA is that as far as I know do not take the previous conversation into account. You only really get zero shot responses. This means that you cannot steer the model towards paying attention to specific instruction related details. The projector simply creates a token representation of the visuals (not necessarily human language tokens) and the LLM just processes that as usual.
Re: Vision language models are blind
#40in other news, vision models are bad at things they aren't trained to do