4 ms·
No, it does not work like that, it actually can process the image itself there is not an intermediate image to text step
by weird-eye-issue 3mo ago
No, it does not work like that, it actually can process the image itself there is not an intermediate image to text step
- voidUpdate 3mo agoHow does a Large Language Model process images then?
- jappgar 3mo agoIt can only deal in tokens, so you're essentially right that it creates a textual description before describing it back to you. This process is obviously incredibly lossy and details are easily missed