Old Dogs, New Tricks
As all of you should know, Ann and I are not exactly what folks would call Luddites. We love our gear and we do not shun technology. And as much as we are both extremely opposed to AI making our images for us, face it, we shoot with digital cameras instead of shooting film, and are happy that Lightroom’s de-noising, cloning and masking tools have significantly improved thanks to AI. Which brings us to today’s blog post.
Say hello to Claude.
It didn’t take long after Anthropic released Claude Cowork that I started looking into ways it could help me in my billable hour work-flow (proceeding cautiously as one might expect given the number of malpractice suits that have flowed from the stupid use of AI by lawyers), particularly given the reasonable cost for a first-tier subscription. Ann, being Ann, thought of much better uses for Claude and, given she has more time on her hands than I do (see billable hour note above), she set out seeing how it could help her photography in a number of ways.
She started out with Claude advising her on how to best set up and use her Fjorden iPhone camera grip. Why search, find and read dozens and dozens of reviews for a good setup when Claude could do the searching and make recommendations for you in a matter of minutes.
It was, to say the least, a good learning experience. Why? Because it very quickly revealed the shortcomings of current levels of AI. I’m sure you’re aware of what folks call AI “hallucinations” (read: lies), well Ann came across plenty of those. “Claude, I don’t see that setting.” “Sorry Ann, I made that setting up. It doesn’t exist.” (see malpractice note above). Yeah, that sort of stuff. They made for good stories at the dinner table.
Still, in the end, Ann worked through potential configurations of the Fjorden grip and its app and . . . after a few discussions with me as well, settled on a relatively straightforward set-up for a couple of different situations. Like so much in our technological world, it’s hard to keep things straight when the program or settings are complicated (read: have a zillion setting) and you rarely use them. KISS still applies these days.
Ann then turned Claude onto how to make iPhone photographs not so over the top, so they look more like images from cameras. Again, it meant delving into different settings, mostly involving sharpening and saturation, but it was much more productive. She worked on images in-camera (on the phone) and then in programs to determine which settings gave her results that she found . . . better (face it, a small iPhone sensor can’t do what our dedicated cameras can). She then made some fundamental setting changes in the camera app and, hopefully, the photos she takes with her iPhone will be more pleasing in the future. Yeah, the Iceland trip made both of us appreciate the ease of a phone camera, especially for casual pictures, though the output still left us cringing over quality issues more often than not. Hopefully, the setting changes . . . developed through Ann’s conversations with Claude . . . will fix that in the future.
Now it was time for some serious work - Organizing Ann’s photographs! Ann thinks much less than I do about organizing images into categories, galleries, collections, and folios. Just check out the website sections for our images and you’ll see just how many different ways I think of organizing my images than Ann does hers. Neither is better, but Ann wanted some ideas about how to organize images in a way that shows the work but is also maintainable. Face it, all she has to do is look at me to see that I’ve got a lot of work still to do for 2025 images on the website to know that being able to keep up with and maintain a galleries section is something someone should keep in mind when setting it up. Unlike me.
Anyway, it started slowly. Ann started exploring Claude’s taste by doing things like tossing an image tino Claude to see how Claude would critique the image. Satisfied that Claude wasn’t going to do something like say the images suck, Ann went for it. She culled through her images and compiled a group of about 60 images that she liked with the intention of having Claude go through them to see Claude it could discern any themes in her work that she could use for her website galleries.
She then loaded up the first 20 images . . .
. . . and engaged Claude in a discussion about the images. It was a very back and forth process using normal conversational language. What’s interesting is that Claude fully understood different descriptions from earlier conversations about types of photography as well as the work of certain photographers as they engaged in the back and forth. Conversations about why Claude reached the conclusions he did (yes, I know, I’m personifying an AI - but if I can be polite to Siri and say, “Hey Siri, please set a timer for three minutes!”, I can call Claude a he), which images he thought were the best, what were the shortcomings of other particular images.
Then it would be time to dump another set of 20 images, and start the discussion again.
This went on for a couple of weeks, refining what Claude was doing and why Ann might want to do it a particular way. In the end, they came up with seven categories where images would be organized in Lightroom with some, but not all, of them eventually making their way onto website galleries. Then, as you can see below, Ann and Claude moved on to coming up with an image ratings scheme.
Again, it was a back-and-forth, several day process. Often, at the end of my work day, Ann and I would chat about the latest discussions between Ann and Claude, my thoughts, and any decisions Ann had made. It’s interesting to see how an AI can see things very differently than you or I might (why did Claude think highly of an image Ann wasn’t so sure about), and how sometimes it wants to go off in an odd direction and needs to be pulled back on task. Believe me, Ann put a lot of effort into it. In the end, they came up with a rating scheme for images.
Eventually, Ann decided she wanted to document the analysis and they decided to output a working document.
And, of course, onto the next idea that Ann came up with - examining each of the images and having Claude rate it and explain why it gave each image that rating. The work continues.
One of the things I’ve found impressive about Claude/Claude Cowork is its ability to plainly and clearly convey information in the output it produces. Tables, colored headings, clear and concise bullet points . . . it’s rather impressive.
As you can see, the Website Category Reference is several pages long.
Overall Ann thinks she’s benefitted from the back-and-forth she’s had with both Claude and examining her images. That can only be a good thing.
Now for me to get off my butt and . . . .