3 ms·
Isn't this outdated already. It's 2 models slapped together rather than natively multimodal.
by KaoruAoiShiho 2y ago
Isn't this outdated already. It's 2 models slapped together rather than natively multimodal.
- maciejgryka 2y agoIsn't "two models slapped together" basically how all of these things work, starting with CLIP? Not sure about GPT4o, obviously, I don't think they released any underlying architecture details?
- whimsicalism 2y agoYour understanding is correct, even GPT4o will have an encoder model.
- tkellogg 2y agowhat even is “a model”? I’m not sure there is a technical definition that corresponds to how it’s used by the tech public - single interconnected neural network (LLM attention layers break this, autoencoders complicate this) - single training pass (LLMs have multiple passes, GANs have a single but produce multiple models)
- whimsicalism 2y ago> - single training pass (LLMs have multiple passes, GANs have a single but produce multiple models) LLMs have multiple passes? wdym?
- whimsicalism 2y agoNo, that's how all of these models work.
- JeremyHerrman 2y agoFrom TFA, PaliGemma is competitive to GPT-4o and even beats it in terms of speed and OCR accuracy. It also can do object detection (bounding boxes) and segmentation which GPT-4V/o and Claude 3 Opus can't do at all. Not to mention it's built to be fine tuned and commercially permissive!