AI based photo editing tools like those from Topaz Labs have been around for a while, but I’ve never found them to be good enough to justify disrupting my otherwise Adobe Lightroom based photo editing workflow.

In the past couple of years, Adobe has added native AI based photo enhancement tools to Lightroom, starting with Super Resolution and more recently Raw Details and Denoise. They’ve also steadily improved these tools while also rebuilding Lightroom to work with Apple silicon for optimal performance.

I have an affection for small, powerful digital cameras but there are fewer of these being made as camera manufacturers seek to differentiate from increasingly photography capable smartphones by investing in larger sensors, image stabilization, articulating screens and high quality electronic viewfinders, all of which make cameras physically larger. These investments have yielded great small mirrorless cameras like the Fuji X-E4, Leica CL (discontinued, sadly) and the Canon R50 (what a camera!), but gone are experiments like the Sony RX1 range, and the development much smaller cameras like Canon GX and SX series, the Panasonic ZS, TZ, LX ranges, Nikon A models, Fujifilm XF10/X70 and Leica TL/CL and compacts seems to be done or close to it. The Ricoh GR III is alone as a larger sensor compact actively being sold (the Leica Q2 is a much bigger camera), and the Sony RX100 range with a 1” sensor lives on but future development seems to be focused toward video oriented cameras in this form factor, such as the ZV-1.

So, I turned to my collection of older small cameras which includes the Panasonic LF1 (2013) a 190g compact with a 1/1.7” 12mp sensor, Panasonic GM1 (2013) a 205g ILC (without lens) with a 16mp micro 4/3 sensor, and Ricoh GR Digital II (2007) a 200g compact with a then large-ish 1/1.75” 10mp sensor.

My hunch was that features like Super Resolution and Denoise would breathe life into the images created with these cameras, in particular RAW files shot under challenging conditions or, at the very least, just increase the resolution to enable greater latitude for cropping and larger prints (if that’s your thing).

I was not disappointed.

Denoise, in particular, delivers impressive results that make unusable images not only acceptable but highly usable.

For example, from the Ricoh GR Digital II, which really struggles with the dynamic range and variety of textures in this scene.

Settings: ISO 100 | f/2.4 | 1/200 sec | Spot metered on the center of the image Light: +1.5 exposure, -70 contrast, -50 highlights, +55 shadows, -30 whites, -5 blacks Color: 5,614 temp, +73 tint, +5 vibrance, +5 saturation, -100 pink, -35 purple Effects: +5 texture, +10 clarity, +30 dehaze Detail: 60 sharpening, 70 masking

This raised the shadows while preserving the highlights but even with ISO 100 there’s a ton of detail destroying noise in the shadows.

Below is with Denoise applied set at 60.

Not only is the noise almost gone, Lightroom accomplished this without smearing the details or introducing any other artefacts which has been the problem of previous noise reduction solutions, especially in-camera software. On the contrary, it appears as though detail has been added to the image, which essentially is what AI-based denoise techniques aim to do.

Because the enhanced output is a new DNG raw file, I tried to further optimize the exposure which re-introduced some noise but still yielded a very usable image with plenty of detail.

This is exciting in a number of ways but primarily for me because it extends the life of older cameras, many of which have both technical and ergonomic features that manufacturers have moved away from.

For example, Leica has stopped making APS-C cameras, many of which are really beautiful and lovely to use, AI based editing tools make it possible to get the best possible images from them.

There are other problems with older cameras – autofocus and low resolution screens/viewfinders are notable examples – but there’s a joy and freedom of being able to take a tiny camera like the Panasonic LF1 everywhere, knowing that I’ll be able to have the control it offers as a photographic device without having to sacrifice image quality.

It also makes photography more accessible than ever for people just getting started as they have a huge range of cameras available to them.

Of course, these tools also make it possible to get even more out of the amazing images produced by the latest modern cameras.

#photography #cameras #editing #tools #ai

Source: Benedict Evans Newsletter No. 472

My preferred order for reading books is audio > ebook > paperback > hardback.

(Except for art and photography books, in which the order is hardback > ebook > paperback.)

Reading an audiobook of course means listening to it and debates abound whether we retain information better or worse if one reads a text or listens to it being read (examples here and here).

But, as ChatGPT says, learning styles vary and, like with many things, practice and repetition build optimal habits to get the most out of an experience.

In 2022 I read 45 audiobooks (full list at the bottom of this post).

The number of books isn’t the thing to anchor on here – listening to articles vs reading them will have a similar effect – rather the habit I’ve found most effective to both parse and retain information is reading several books on the same topic in clusters.

This has two benefits:

1. You listen to variations of the same data, research and anecdotes.

2. You hear different points of view on a given topic.

Several distinct topic clusters emerge in the list of books:

1. Financial speculation and the irrational behavior that drives it. This was a timely topic to go deep on, following the crypto and NFT collapse. Spoiler alert: it’s predictable human behavior that causes these endlessly repeated cycles of boom and bust, divorced from the underlying asset class, which in many cases (see: Beanie Babies, NFTs) have no intrinsic value.

2. Inclusive product design. An important and under invested in area.

3. Product management and design at Apple. For a company that has a reputation for secrecy there’s more information available on this topic about Apple than pretty much any other company.

4. Financial (mis)adventures of ultra high net worth individuals and the structures that enable them. An evergreen topic.

5. Frameworks for thinking. Mental models on how to approach problem solving.

The main way I find subsequent books is via references in a current book. In most cases it leads to books that go deeper into a topic, like branching roots, but sometimes it takes you to whole new tree.

This was the case with Strangers to Ourselves, by Rachel Aviv, which was referred to by Tom Vanderbilt in You May Also Like. It’s not a book I would’ve otherwise found my way to but I’m very glad Tom led me to it as it’s a powerful and moving book.

Looking at this list, another thing that stands out is that there’s only one novel (Red Pill by Hari Kunzru, which is excellent).

Novels are almost never read by their authors. Red Pill is and it’s all the better for it. Voiceover actors tend to act out novels, altering their voices for different characters and I find it very distracting. Hari doesn’t do this (or at least I don’t remember him doing it).

While not all non-fiction books are read by their authors, there’s a greater prevalence and the voiceover actors who read non-fiction books either don’t act them out or when they do, do so subtly.

This is one thing that makes Apple’s announcement about AI audiobooks interesting. Will these AI readers act? If so, will it be subtle? What models were used to train the voices? How do they handle non-Western/Anglo-Saxon names and words? Can an author train an AI to read their books so that it sounds like them?

I’m intrigued to try audiobooks read by AI readers and also think this is bridges a valuable gap by potentially creating audiobooks of the many books that previous didn’t have audio editions, which is a very welcome development.

Full list of books in reverse chronological order (links go to Apple Books or Audibooks.com):

1. Strangers to Ourselves by Rachel Aviv

2. You May Also Like by Tom Vanderbilt

3. A Short History of Financial Euphoria by J.K. Galbraith

4. Money Mania by Bob Swarup

5. Irrational Exuberance by Robert J. Shiller

6. The Great Beanie Baby Bubble by Zac Bissonnette

7. Amusing Ourselves to Death by Neil Postman

8. Move Fast and Break Things by Jonathan Taplin

9. Cinema Speculation by Quentin Tarantino

10. When Women Lead by Julia Boorstin

11. Like, Comment, Subscribe by Mark Bergen

12. Status and Culture by W. David Marx

13. The Laws of Simplicity by John Maeda

14. Weapons of Math Destruction by Cathy O’Neil

15. Dark Horse by Todd Rose & Ogi Ogas

16. Technically Wrong by Sara Wachter-Boettcher

17. Mismatch by Kat Holmes

18. Imaginable by Jane McGonigal

19. How Design Makes the World by Scott Berkun

20. The Paradox of Choice by Barry Schwartz

21. The New Breed by Kate Darling

22. Predictably Irrational by Dan Ariely

23. Loved by Martina Lauchengco

24. Change by Design by Tim Brown

25. Sprint by Jake Knapp, John Zeratsky & Braden Kowitz

26. User Story Mapping by Jeff Patton & Peter Economy

27. Start with Why by Simon Sinek

28. After Steve by Tripp Mickle

29. Inspired by Marty Cagan

30. Insanely Simple by Ken Segall

31. Creative Selection by Ken Kocienda

32. Red Pill by Hari Kunzru

33. Build by Tony Fadell

34. Lucifer’s Banker Uncensored by Bradley C. Birkenfeld

35. Butler to the World by Oliver Bullough

36. Moneyland by Oliver Bullough

37. Don’t Believe a Word by David Shariatmadari

38. Reality+ by David J. Chalmers

39. Thinking in Bets by Annie Duke

40. Thinking in Systems by Donella Meadows

41. The Scout Mindset by Julia Galef

42. How to Turn Down a Billion Dollars by Billy Gallagher

43. Four Thousand Weeks by Oliver Burkeman

44. The Precipice by Toby Ord

45. The Alignment Problem by Brian Christian

#reading #amazon #books #apple #ideas #ai