After publishing my AI Is Causing Problems article, two people reached out to me to express their disagreement with my paragraph on AI vibe coded camera firmwares. Basically, people are using AI to hack and rewrite firmware to add new features and functions that the camera doesn’t actually have. It seems like the general consensus is that this is good, basically supercharging your camera, and ushering a new era of tech. Any camera can be anything you want it to be, as long as you have the “right” firmware uploaded. Personally, I think it’s really bad, and should be avoided. Let me explain.
“When it comes to AI vibe coded camera firmware,” I stated in the article, “that’s super risky. It’s not so much if someone’s camera will become a permanent paperweight, it’s a matter of when and how many, and whether it will be yours. Personally, I wouldn’t load any unofficial third-party firmware into any of my cameras, the risk is much too large. Even if it seems to be working fine, you have no idea if some rarely used setting combination will be what wrecks it, or if by using it the camera will wear out much quicker.”
If a team of programmers created a 3rd party firmware for your camera, there would certainly be a risk to installing that, and you’d hope that it was extremely heavily tested before being released to the public (how confident are you in that being true?). Even the camera companies themselves, who know their technology the best, sometimes mess it up, and on a rare occasion it can even be catastrophic—just ask the Sony photographers who had their camera brick after updating. Even Fujifilm a couple years back temporarily screwed up the AF on some models. That’s why it is always a good idea to wait at least a few weeks, if not a few months, before updating the firmware, and that’s for official updates! Unofficial 3rd party firmwares should be handled with a ton more caution.
Unfortunately, most of the 3rd party firmware hacks are not created by a team of experienced camera programmers, but instead are made by AI vibe coding. AI is better than a novice, but not as good as an expert, so the risk is much higher. It gets things wrong sometimes. What AI is especially good at is doing something quickly. Maybe it would take a team of programmers several months to complete a task that AI can do in a matter of minutes, but I can assure you that the team would have done a better job of it. It’s just not financially worthwhile to assemble a team and devote months to something like 3rd party camera firmware; however, AI can basically do it for free (sort of) and very fast, so it becomes worthwhile, even though not as good. In a sense, you get what you pay for.
It’s inevitable that unofficial 3rd party firmware will brick people’s cameras. It’s going to destroy the motherboard and turn it into a paperweight. This is going to happen with 100% certainty. Like I previously said, it’s not a matter of if, but when and how many, and whether it will be your camera or not. It might seem to be working great for awhile, but you don’t know if some rarely selected (and never tested) setting combination will do it. Because these firmwares often require the processor to do things it wasn’t designed for, the camera will wear out much quicker than if it had the official firmware installed, in all likelihood. Maybe it will break in eight months instead of eight years. You’ll have to consider if the new features are worth the shorter lifespan.
But I think there’s another aspect we should also consider. What does it mean to use some camera make and model? For example, some people really love the Fujifilm X-Pro2, even though it is 10 years old now. If that camera can be hacked with AI vibe coded 3rd party firmware to change its functions and features, is it still the same camera? What will it mean to use an X-Pro2? If AI vibe coding can turn a Sony into a faux Fujifilm, or a Fujifilm into a faux Nikon, what does it mean to use those brands? If a camera can be whatever you want it to be, what are the consequences of that? I don’t have the answers, but I think the answers are generally going to be pessimistic, and potentially really bad for camera makers.
One option that camera brands have is to embrace AI vibe coding, and put everything into every model. Yeah, it’s going to be less stable and less good than their actual team of coders could have made it, but they’ll beat the 3rd parties to the punch—basically, the first party becomes the third party. And they can lay off some of their staff, which is good for profits. Another option is to keep the status quo, but why invest so much into something that their users are just going to throw away anyway? What are the incentives? Is this even a viable choice in the long run? Which brings us to the third option which is to stop innovating and just let the AI vibe coders take over firmware. Basically, what the official firmware is in ten years will be nearly the same as it is today, and you pick your 3rd party firmware of choice after buying the camera so it becomes whatever you want it to be. The only other option that I see is for photographers to reject AI vibe coded 3rd party firmware, but it seems that I am in the minority with this opinion. So which of the three do you want? Camera makers embracing AI vibe coding? Keeping the status quo—but will they really with fewer incentives to do so? Or just let the 3rd party AI firmwares take over?
I don’t envy the camera brands. I think this will be especially tough times for them to navigate, and most likely some brands will get hit pretty hard. Obviously the tactile experience—usability of knobs and buttons and such, as well as ergonomics—will have a larger roll in a camera’s success than what the camera can or can’t do out of the factory. If it can’t do something you want it to do, just download some file and install it on your model (hopefully it doesn’t break it!). So, again, what will it mean to use a certain make and model? It will be more about which 3rd party AI firmware you are using than the anything else. And I think this is sad.
Something I’ve talked about a few times are the benefits of limitations. Less is more, oftentimes. “You must always work not just within but below your means,” Pablo Picasso famously said. “If you can handle three elements, handle only two. If you can handle ten, then handle five. In that way the ones you do handle, you handle with more ease, more mastery and you create a feeling of strength in reserve.” In other words, there is beauty in restrictions. In my opinion, the Fujifilm X-Pro2 is special because of what it is as it is; adding more features does not inherently make it better. I fully understand that this is an unpopular take, but I hope it makes sense. And I hope that you pause for a moment before installing a 3rd party firmware onto your camera, and consider if it is actually worthwhile.
My two cents – people have been jailbreaking and mod’ing their firmware forever. I remember hacking my PS2 back in college 25 years ago to be able to play games on burned CDs. I don’t blame people for doing this, especially given the glacial pace that Fujifilm has been adding and amending things in their firmware of late. Yes, I’m sure people will brick their cameras, but just like you’ve said elsewhere on the recipe upload from the phone – buyer beware!
Granted, this type of behavior won’t be from the average user; it’s the techy geeks and people who just can’t wait for Fuji to update something. I’m with you Ritchie – why would you purposely Brick your camera. That’s plain dumb.
And now we’ll have to add to the list of questions we ask sellers on the second hand market. Not just dumb things like shutter count, but also “Does it have Fuji firmware or is it some AI coded trash?”
Interesting questions.
What I think is, that some people are either not very good at thinking or they just don’t think a lot. AI doe not make this better in any way.
What are the potential consequences? You can brick your camera, just as you could do that with your iPhone and a jailbrea., Now, if a camera owner wants his camera fixed, it’s up to the camera company to either say no, or to charge them with a heavy amount. That will potentially make people learn from their mistakes, it worked in the past already.
Some other things could happen as well with that AI coded and non tested firmware. It could introduce security vulnerabilities in your camera and personal files or server, it could introduce malicious code in the output files of your camera. Imagine a fake jpeg that introduces code to delete all your photo files, or the software such as Capture One or Lightroom. Would that be fun? I guess not.
In the end, I share your rejection of vibe coded or non original software, that has not been sufficiently tested, but for slightly other reasons.
Thank you for your thoughts.
Not a straightforward one. Well, leave the AI out, that is a given. They always make mistakes and they of course, take no liability when they do. Firmware hacking in principle has been going on as long as cameras have had chips. Was given a pair of Canons, many years ago, ex corporate ID pic cameras, and hacked those to facilitate phone controlled tethering. The everlasting battle against vibrations. It worked. Whether they would still work, I don’t care. They were free and did the job I needed doing at the time. Canon, so did not fit in my quiver.
Distorting a camera’s character is another matter, or is it. I suppose in the large format game, picture character is influenced by choice on lens. Manufacturers tend to aim for a style across a suite of lenses. Even with just Fujinons, A, T and SWD all have a physical and colour rendering look, even across my all EBC, multi-coated glass.
When it comes to the great lenses for the lessor formats, Companies are proud of their image look. Of course, they generally also make task focussed glass, particularly for technical or portrait purposes. You wold be pushed to separate an EOS 180 macro from a Nikon D 200 macro. Both have obscene resolving power, geometric accuracy from their film heritage and as close to colour neutral as possible. Also means, they don’t require much support from their bodies magic sauce table. Makes them first class elements for Frakinrigs. If your fully focussing adapter cannot spoof the body, never mind, corrections aren’t needed. Lessor creations however, need such support to compensate for geometric and colour idiosyncrasies and of course the body does it on the fly.
You should now see where I am going. Fuji and now others have introduced say Nikon’s extreme Vivid, on the back of work already done to make a cheap plastic Nikon or Canon lens produce results, way beyond the traditional price point. Just don’t ding it without a metal stepping ring or a robustly framed filter.
That in itself derives from the first specialised sports lenses introduced by Canon. The affordable, replaceable Cream lenses. Optically terrific and not even all glass but with time limited elements.
So, not heritage glass such as Nikon or Leitz. Market case based on the following:-
Individual guys hire to do a job. Newsrooms and Paps expect high attrition with long lenses. Lost, dropped, stolen, confiscated and what have you. A mere 3 years of optical stability is plenty. Cherry on top. The previous ‘go to’ rig for sports, surveillance and what have you, the Novoflex Follow Focus rig was still the best but expensive and getting people shot in places like Ulster. Carrying anything long, black and with a pistol grip in the ‘Bad Lands’ of County Fermanagh, not good for the life expectancy.
Summary. We all hack our kit to give us results we like. I have a weakness for Voigtlander lenses. I just love the crisp, accurate resolution, the for me beautiful colour balance and the sensuously silky controls. Fujifilm give us a wonderful high resolution starting point. I still hate the shutter button and power switch. I also hate the problems with CamRanger2 that makes work with my Nikons so easy. Would I like better firmware? Absolutely. Would I like less issues from the Clarity setting? Gertcha! Would I spend what is left of my 1,000 months, potentially bricking my grandson’s legacy. Do I really need to answer?