New thread on DXO Photolab 4

[Adrian]

Re AI: I have a somewhat different view of this than you do, as I am trying to maximize control over my edits rather than turning them over to an algorithm based on someone else's machine-learning exercise. I find some of the AI-based algorithms useful but have no use for others.

Dan, agree with you on this score.............

AI is without doubt the future in many areas, as we move faster into the machine led world. From a developers point of view there is no money in the enthusiast amateur photography market, which has shrunk a great deal in the last decade. There is a big market for AI though because it is essential to sell phone upgrades and has a huge social media market where super fast templated enhancement software will grab market share. Vast numbers of users. For of whom are interested in any curves, let alone a learning curve.

I think many pro photographers can benefit from AI workflow, especially those who shoot video as well. Even if only for You Tube content. Amateurs will get the crumbs from the table, which is fine if AI doesn't matter to you.

In my view it os fake AI anyway in the photo market. It's not "clever" in the sense of learning your preferences. Really it is just a set of moderately sophisticated templates at the moment. However, much more sophisticated AI is around. For example my eldest son is currently using it to adapt gas turbine blade engineering to take account of machine learned stress variables. It's very interesting.

[pschlute]

AI is without doubt the future in many areas, as we move faster into the machine led world. From a developers point of view there is no money in the enthusiast amateur photography market, which has shrunk a great deal in the last decade. There is a big market for AI though because it is essential to sell phone upgrades and has a huge social media market where super fast templated enhancement software will grab market share. Vast numbers of users. For of whom are interested in any curves, let alone a learning curve.

I think many pro photographers can benefit from AI workflow, especially those who shoot video as well. Even if only for You Tube content. Amateurs will get the crumbs from the table, which is fine if AI doesn't matter to you.

In my view it os fake AI anyway in the photo market. It's not "clever" in the sense of learning your preferences. Really it is just a set of moderately sophisticated templates at the moment. However, much more sophisticated AI is around. For example my eldest son is currently using it to adapt gas turbine blade engineering to take account of machine learned stress variables. It's very interesting.

Codswallop

[Adrian]

I have no idea what you mean by "out to get me"; the is not at all how I see it. However, I dislike sharp practice (in most fields companies are not allowed to opt people in automatically unless they take action) but it's trivial sums. This is a philosophical matter for me.

[DanK]

Codswallop

I had to look that one up. Not used on this side of the pond. Thank you.

Of course it's not learning your own preferences. For the most part, it's using training sets to learn what someone somewhere deems good and then building a set of algorithms to apply that to whatever you feed it. Jeff Schewe said that he is one of the people Adobe approached to help them amass the images they would use as a training set. Most users have nowhere nearly the computing power to do AI locally, and there isn't much data. It would take a long time to amass data about your own preferences.

That's why I have so little interest in it. I want to get better at creating images that I like, and I hope that some of them will actually be a little bit original.

[Adrian]

Peter you have such a way with words. Very erudite.

The point I am making is that AI, implemented properly, is more than capable of learning your preferences. But camera software developers are not selling that, so it has become just a marketing tag, like 'turbo" on things like vacuum cleaners. When I dipped my toe into the arena by trying out Luminar AI I had expected that it would actually be AI such as I have seen implemented in other industries where it is used cleverly to learn parameter changes. I was quickly put off Luminar because it is not really artificial intelligence being implemented at all.

[DanK]

Peter you have such a way with words. Very erudite.

The point I am making is that AI, implemented properly, is more than capable of learning your preferences. But camera software developers are not selling that, so it has become just a marketing tag, like 'turbo" on things like vacuum cleaners. When I dipped my toe into the arena by trying out Luminar AI I had expected that it would actually be AI such as I have seen implemented in other industries where it is used cleverly to learn parameter changes. I was quickly put off Luminar because it is not really artificial intelligence being implemented at all.

I agree. I think it's mostly advertising hype. They seem to be using machine learning to help them develop algorithms that are more complex and better tailored to certain uses than hand-coded algorithms would be. Adobe's wonderful object selection tool is a great example of this, IMHO. This approach can be very valuable, but it's not what most consumers think they are buying when vendors hype their "AI" software.

[davidedric]

I do sometimes click Lightroom's Auto button. One thing I find interesting is that it often gives a pleasing result, at least as a starting point, but using very different settings from those I would have chosen.