Visual input.
WHAT THEY CAN DO
Describe an image Read text from a photograph Answer questions about a picture Extract data from a screenshot or a document photograph Explain a diagram or chart
PRACTICAL USES
Photographing a receipt or invoice and extracting the figures Understanding an error message from a screenshot Reading a handwritten note Explaining a chart
WHAT THEY DO BADLY
Precise counting Exact measurement Poor quality images Anything requiring specialist visual judgement
WHAT TO NEVER RELY ON
Medical images Engineering measurements Financial figures from a poor scan Identification of anything where being wrong matters
WHAT TO VERIFY
Every extracted figure, against the original.
FOR ACCESSIBILITY
Describing images for someone who cannot see them is a genuinely valuable use.
Descriptions may be wrong. That is a limitation worth knowing.
PRIVACY
Photographs contain metadata including location. Consider before uploading.
