Chapter 1: The Rise of A.I. in Photography

A.I. has become a hot topic in discussions today, often accompanied by fears that technology may overshadow human creativity. Photographers have mixed feelings about A.I.; some worry it could eliminate the profession, while others see potential benefits, such as enhancing images or improving workflows. However, since 2016, tools utilizing A.I. for image search and keywording have been available, aimed at aiding photographers rather than replacing them.

Imagine having an application that could effortlessly manage your photo catalog, help you find specific images, or generate relevant keywords for your files. My experience with these tools began years ago during their initial development phase, and it’s intriguing to reflect on their journey to mainstream recognition.

In 1993, Apple introduced the Newton, a tablet capable of handwriting recognition and calendar management, yet it saw limited adoption before its discontinuation in 1997. Fast forward to 2010, when the launch of the iPad marked the beginning of the tablet era.

Photo Credit: Apple

Similarly, in 2001, Real Networks debuted Rhapsody, a streaming music service that, despite its vast library, never reached widespread popularity. The true revolution in on-demand music emerged with Spotify in 2010.

In 2016, Excire, a small German firm, released a Lightroom plugin that employed A.I. to automatically generate keywords and allowed users to search based on visual similarity. Although it garnered positive reviews from industry figures like Jim Nix and Peter Forsgård, it failed to attract significant attention. In June 2023, ON1 launched Photo Keyword A.I., ushering A.I. image search and keywording into the photography community.

The tech world is filled with innovations that arrived too soon for their time. Throughout my 25-year career in product management across various media companies, I experienced three major shifts: streaming music at Borders, electronic publishing at Thomson Reuters, and A.I. image search at Shutterstock. Some ventures flopped (remember Borders' dismissal of the Internet?), while others achieved moderate success (like Shutterstock’s Visual Search).

As a passionate photographer, developing the first commercial visual search and keywording tools in 2015 was a highlight of my career. Now, with ON1’s Photo Keyword A.I. hitting the market, it's an excellent moment to revisit those formative days of A.I. image search and reflect on Excire, the tool that preceded ON1 by seven years but struggled to gain traction.

In 2015, I joined Shutterstock as the Director of Product for Search and R&D. The R&D team had already begun experimenting with A.I.-based image search using neural networks. My role involved acting as the customer’s voice, balancing engineering capabilities with users’ needs to drive revenue.

Through extensive customer interactions, I discovered a strong reluctance to trust text-based searches. Users preferred to browse images rather than rely on keywords. One customer expressed, “I prefer to find images by looking at them, not by searching for them. I know the right image when I see it.”

Armed with these insights, we developed three A.I.-driven search tools: a reverse image search that identifies images based on uploaded content, a visually similar images feature that showcases related visuals, and a keyword generation tool for our artists.

To summarize how visual search and keywording function at a high level, it’s essential to have a large dataset of well-labeled images. These images and their labels serve as training data for the computer, which breaks down the images into attributes such as color, shape, and texture.

Using the training data, the computer analyzes the images to identify their contents. More scans lead to better recognition. An image can be compared to a topographical map, where each layer signifies different details, allowing the computer to create a vector that encapsulates all attributes of the image.

When an image is uploaded to the service, it can generate keywords or identify similar images based on the vectors created. Our team at Shutterstock, with its extensive collection of stock images, was able to build a sophisticated computer vision service that could accurately identify specific breeds of dogs, for example.

After a year of developing a visual search service, I sought to manage my own poorly organized Lightroom catalog. I discovered Excire and was amazed by its capabilities. Within seconds of scanning my collection, it could:

• Search by object and color
• Automatically assign keywords
• Identify visually similar images
• Recognize people and facial attributes
• Upload external images for comparison

In 2016, alongside Excire, Apple also introduced a version of visual search for its Photos app. Although it had limitations, it demonstrated potential, continuously improving since its launch.

Despite these advancements in visual search tools, I anticipated that 2016 would mark the year of widespread adoption. However, the response was surprisingly muted. Although Shutterstock received positive press for its Visual Search, the reverse image search feature did not resonate with users. The keywording tool succeeded mainly due to its utility rather than its underlying technology.

After the launch, I analyzed various metrics and engaged with users about Shutterstock’s visual search tools. Many enjoyed the reverse search, but it did not capture widespread usage. As one user noted, “Even if you don’t like searching for images using keywords, after being conditioned to do so for a decade, changing that habit is difficult.”

I have been using ON1’s Photo Keyword A.I. since its recent launch and find it impressive for a first version. While it may not have all the features of Excire, the distinction lies in its standalone nature versus Excire’s plugin format, which I prefer as a dedicated Lightroom user.

Both Excire and ON1 Photo Keyword A.I. are excellent products. Despite my loyalty to Excire, the advent of ON1's tool is likely to make visual search and keywording essential features in photo editing software. Although it may be seven years in the making, I’m excited that A.I. image search has finally arrived.

Chapter 2: Meet the Live View Writers

Live View features a diverse group of authors passionate about photography. You can learn more about them, read their stories, and connect by visiting the Live View About page.