"Qualcomm exec thinks AI will help smartphones eclipse dedicated cameras soon" --reasonably accurate headline repeated by all the news scrapers who scraped the Android Authority interview. [Disclosure: I own Qualcomm stock.]
(I'll have more to say about AI in an upcoming article.)
The premise of Sony's original statements about smartphones surpassing dedicated cameras and now Qualcomm's echo of them lies in a very simple notion: that AI can make up for physics. In essence, what's been happening is that the image sensor/image processor chain in the smartphones has been getting tighter and faster and more "intelligent", to the point where the images produced overcome the obstacles of underlying physics of such a small sensor.
"The processing in Snapdragon [Qualcomm's image processor for smartphones] is 10 times better than what you can find on the biggest and baddest Nikon and Canon cameras." Notice the word used wasn't "faster," but rather "better." The Qualcomm exec does go on to say they're doing "many times more processing", though. I'm not 100% sure that is true given the current dedicated camera processors. It's mostly just different processing.
Quite obviously, both Sony's and Qualcomm's remarks about smartphones surpassing dedicated cameras are self-serving. Sony wants to sell more small image sensors, and Qualcomm wants to sell more sophisticated image processors. The huge volume of the smartphone business means that there's more return on R&D in that space than there is in the far smaller dedicated camera market. But the phone market is starting to top out and peak just like the camera market did. Once you have a recent, high-end phone, how often would you really need to upgrade? 5G was supposed to kick off a whole new wave of buying, but it was really the discontinuation of 2G that did that.
I'm a little surprised that the camera makers haven't snapped up some of the machine learning software that's been percolating in the photo processing realm. Yes, using a smaller process size and adding cores gives you more processing power, and all the camera makers have been doing that. But ultimately it's coding algorithms into the hardware that gives the bigger bang for the buck. You need to own the right algorithms to get that advantage, which is the Apple/Google/Qualcomm thrust at the moment.
Nikon and Sony seem to understand what's going on a little better than the rest of the dedicated camera crowd. Nikon sees the advantage of dedicated cameras coming in optics, particularly long optics. Sports and wildlife photography is not a strength of smartphones, after all. [Of course, I have to ask if Nikon sees this, why the heck isn't there a Z90 yet?] Sony sees the social media slide towards a video preference being key, and many of the processing techniques that the smartphones do with stills are still out of their reach with top-end video. So Nikon has introduced a Z9 and lot of telephoto lenses in the Z System lately, and Sony is introducing a lot of video products and lenses. Canon doesn't seem to be focused (pardon the pun) on any one thing.
Ultimately, the issue isn't whether smartphones are going to render dedicated cameras useless. I don't see that happening any time soon. On the other hand, dedicated cameras are going more niche, which I don't see smartphones attempting to match (gives up their spread of R&D over unit volume). This is basically a replay of the CD/High-Fidelity story.
Bonus: you can still buy specialized Hi-Fi equipment, and that industry has moved on from a CD-caused bottom. But when was the last time you bought a CD player or CDs? That's right. Smartphones as cameras are likely short-lived. (Since I know you're going to ask, what do I see replacing them? Multi-camera AR scene formation.)
"Time for Sony to Change Their FF Body Design?" --headline in a forum post on dpreview
This is a relative to the "if it isn't broke, fix it" idea.
There's a reason why tools all tend towards a common design. That eventual design is the most effective to make use of the tool, basically. Cameras are highly configurable tools (lots of settings), which further lends themselves towards a common need in their design: the ability to change settings without taking your eye from the viewfinder or your index finger from the shutter release.
Viewfinders end up in the center of a camera for a reason, too: particularly with long lenses the view is on axis with the rotation. Yes, I know that some of you are left eyed or big nosed and prefer the rangefinder style of off-rotation-axis viewing, but a right-eyed user also then loses some of the bracing effect that keeps the camera steady. And yes, I know you can turn on VR and get steadiness back, but generally I don't want "fixes" to something that wasn't broke. That's not the right design goal to pursue.
Then we have the "muscle memory" aspect of cameras. As I've documented for 30 years now, Nikon can be their own worst enemy at this, such as when they decided on the N80 to move the exposure compensation button that had been in the same position for two decades (and has since returned to its "proper position" for over two more). Once you learn a system, whether it be Canon's or Nikon's or Sony's or whomever's, you don't want to be relearning the basic things you do every time you get a new camera. New car? Oh, we moved the turn signal to the door panel and the horn to a button on the dash. Yeah, we don't want that in cameras, either.
I actually applaud Sony for what they've done. The original A7 bodies had a lot of issues, some major, some subtle. Over time, Sony has directly addressed each of those without actually breaking their overall designs. For example, the AF-ON button got a modest raised platform to make it easier to find than the small, flat, back-panel-level one that had preceded it. It used to be that you couldn't find the AF-ON button with even thin gloves on; now you can.
Moreover, one thing that made the Sony A7 bodies popular was size. Specifically, smaller-than-full-frame DSLR bodies, and frankly, smaller than most APS-C DSLR bodies. Start tinkering with new body design, and size could suddenly become in jeopardy. Within the current size you tend towards optimal-for-size and smaller iterations, which I believe the A1 and its followups are fairly close to.
In product management teams an old adage was often repeated: "don't let customers design products." To some degree, that's correct, in that any specific customer isn't seeing the broader and more difficult problems that need to be balanced. Even architects designing one-off custom homes have issues with customer designs ("they don't know the ramifications of what they're asking for").
On the other hand, product management needs to listen to customers to figure out what is working and what is not. Leave the things that work, fix the things that aren't.
The body of the headline poster's forum query got to the reason why they wanted Sony to change their body design: style. "[The Hasselblad X2D is] so alluring that you simply want to hold one in your hands." Ah, consumers. So fickle. They won't buy the Subaru because it's so pedestrian, but oooh, look at that Porsche.
Style is not something I want in my camera, it's something I want in my photos. The camera I want in my hands is something that fits, whose controls are optimized and accessible, and that holds up to continued use. I think I've been very clear about that for 30+ years. The day you see me with an "alluring" accessory is the day I jumped the shark.
"Fujifilm just broke all the rules for APS-C camera sensors." — digitalcameraworld.com review headline
Let's leave out the question of whether or not the X-H2 is a good camera or not for a moment. When I see hyperbole such as the quoted line, my radar goes off. First, there are no such APS-C sensor "rules" that I know of, and second, breaking rules generally has both positive and negative consequences, depending on your viewpoint and how carefully you look.
Moreover, the publication in question purports to use Imatest to provide results, though they don't show actual Imatest results ;~). My own Imatest results have never aligned with what digitalcameraworld reports in their graphs, so I would say this: show me your work.
I'm also worried about lines such as "We thought APS-C sensors had reached their resolution limit, especially with the lack of any significant increase in real-world resolution from Canon's 32.5mp sensor." Um, no. Apples-to-apples comparisons please. Because of AA filters, lenses, and a host of other factors, it's real easy for "more pixels" to not show any "significant increase." We've had this problem since forever in digital. I've written it many, many times: more sampling is always better. How much better, however, is always the question these days, as we're dealing with incremental increases that don't always show obvious visual change to most people.
Physics and math don't change just because someone created a new image sensor. Perhaps someday a completely different sensor technology (photon sensing JOTs, GaAs instead of Si, maybe a quantum sensor) may reveal that there was information possible to capture that adds clearcut ability to our imaging, but we're in a world right now where we're making small gains via incremental iteration on a fairly fixed, known path.
I get it. As a reviewer you want to try and add relevance to your review as opposed to simple test results. It's real easy, however, to fall into the trap that all the High Fidelity magazines did at the peak of that era, and to overstate something purely because you're trying to prove your own relevance. (Disclosure: I once received a press association "award" for being the first person to use the word "screams" to reference the speed of a computer. So, yes, I get it. You don't see me writing "the Z9 in High Efficiency raw just screams and renders buffers superfluous." )
The problem, of course, is that hyperbole gets picked up and transmitted over the Internet so fast that it becomes "reality" for most people. That, of course, is one reason why I write these "strange things said" articles. Marketing messaging has become so pervasive and amplified in our society as to mask actual need and/or performance. As I was considering this article, an email came in that was asking about gear, but seemed to be valuing f/2.8 over f/4 for no apparent reason. It's real easy to hear about "better" and then lock into that as actually meaning "I need that."
I haven't evaluated the X-H2 yet, and it will be a bit before I get to it, so I can't say how good it is. My expectations would be that more sampling is better. But I'd also expect other aspects to be relevant, too: with phase detect on sensor being based on photosites, smaller photosites get less light. So I'd want to understand how more sampling impacts other aspects of the camera.
Use Manual Exposure Mode with Auto ISO — a common email statement I receive and observation of user setups
The problem with Manual exposure mode coupled with Automatic ISO is simple: you're altering your pixel integrity to preserve a specific depth of field and subject motion constraint. I run into this a lot with wildlife and sports photography users. They want maximum aperture of the lens, but 1/1000 or 1/2000 of a second shutter speeds to stop action. The problem is that with a lens like the 500mm f/5.6 and 1/2000 is that's not a lot of "exposure." Remember, EXPOSURE = LIGHT filtered by APERTURE filtered by SHUTTER SPEED. Indeed, we're already two stops below Sunny 16 in bright sunlit light for an ISO 64 camera. As the light goes down, we get into pixel integrity jeopardy pretty fast.
We sports and wildlife photographers usually want to keep aperture fixed at maximum both for exposure and background isolation. I'm fine with that. Where I start to depart is shutter speed. At times I need 1/4000 (bird wing tips) while at others 1/500 will suffice (bird on stick). In the former case I'm still not getting what I want with Manual exposure+Auto ISO (because I set a slower shutter speed), and in the latter case I'm giving up two stops more exposure on those photosites (because I could have used a slower shutter speed). You might quickly say, you're in manual exposure, so why not just change your shutter speed? Sure, but then why do I need Auto ISO? ;~)
That said, I'm okay with using Manual exposure with Auto ISO if you simply don't have time to be making shutter speed adjustments and if you've carefully figured out at which point the pixel integrity will break down too much. I say that last bit because I keep finding that the Maximum ISO people are setting with Auto ISO is indeed the maximum ISO the camera can set. Even with the best of the best noise reduction routines out there I'm not going to use my cameras at their maximum ISO unless there's absolutely no chance that I can avoid it.
Which brings me back to this: I find it faster to just change shutter speed than I do to take the camera out of Auto ISO and change shutter speed. Duh. Thus I don't tend to use Auto ISO. I instead settle for underexposure in a few cases and will push my shutter speed downward whenever possible.
