4 ms·
I have difficulty trusting this. There are plenty of videos online of LLMs making up stuff like "I just ate a hot dog, is there mustard around my mouth?" "No, e
by Almondsetat 5mo ago
I have difficulty trusting this. There are plenty of videos online of LLMs making up stuff like "I just ate a hot dog, is there mustard around my mouth?" "No, everything is clean" while there is a big yellow stain om the guy's face
- postalrat 5mo agoLike coding, creating images or text, maybe the alternative of doing it yourself is too easy or enjoyable for you. Don't expect that will be true for everyone.
- Almondsetat 5mo agoDid you reply to the wrong person? What are you even trying to say here?
- postalrat 5mo agoYou say you don't trust it but whats your alternative assuming you lost your vision?
- WarmWash 5mo agoThe problem is using a language model to assess images. Probably 80% of "LLM's are below expectation" complaints (from the general population) involves some form of image analyses. Image tokenization is hard because unlike language tokenization, where every token is extremely dense with meaning, image tokens tends to be meaningless or irrelevant but are processed all the same. Give an SOTA LLM a picture of toothpicks and ask it to move one to make a square, and it will probably struggle and fumble it. But give a mid-size LLM from 2 years ago the same problem in verbal form, and it will nail it almost every time. That takeaway is, do everything you can to avoid having the LLM need to rely on images for the answer.
- gruez 5mo agoI thought all the recent models are "multimodal"? Is the image part just sticking an image recognizer in front of the text model?
- RobMurray 5mo agoMost of those videos are chatGPT voice mode, which still used gpt 4o last time I checked. it is far from SOTA.
- Almondsetat 5mo agoI mean, if you want AI to help you real time... you'll need slimmer models. How is what Apple is doing any more SOTA than ChatGPT?