Everything is changing so fast that it's hard to keep up. But my biggest concerns with artificial intelligence and photography remain the same:

  • Generative AI, used for things like noise reduction, creates information in the pixels that didn't exist in the original file. Therefore, in my opinion, a photo processed this way isn't the original photo but an artificial creation (e.g., Topaz Photo).

  • Generative AI models create graphics, photos, and videos that didn't exist before, that is, artificially. (e.g., nano banana)

    So, it turns out that AI can be useful for certain things where the original pixels aren't affected, and that leads me to the next point, which is one of my biggest concerns:

  • Data centers used for AI have proven to be a real environmental and social disaster. Therefore, I don't recommend using AI that depends on the internet and these data centers.

    But there are alternatives. There are AI modes and tools that run locally on your computer without needing the internet, much less a subscription.

    There are AI models that aren't generative; instead, they're created to perform specific functions like noise reduction and information enhancement without creating information that didn't exist in the original file. (Example: RAW FORGE)

    There are functions that don't affect the original pixels of the original file. For example, using AI to create raster masks. (Example: SAM2)

    So, it turns out there are AI tools that are useful and don't pose any risk because they don't generate artificial information and run locally on your computer.

    I recently started using one of these tools in my workflow: SAM2.

    I think it's really important that we can identify which AI tools are safe and useful and which are just factories of artificial information.