Creativity is (was) thought to be a prerogative of humans and ingenuity the last frontier AI would conquer. But AI has come for both—or so it seems. The brouhaha over the capabilities of AI tools, especially of the generative kinds, is numbing. Listening to the hubbub gives one the impression that an alien has appeared among us
Remove yourself from the pandemonium and you’ll see that AI is not out of this world. AI tools, though adept at many tasks, are generally inept. They are not to be relied upon solely or even primarily for tasks that require judgment and nuanced understanding. Photo editing is one such task. The perks of using AI, at any rate, remain minuscule when there are stellar photo editing services that can be had easily.
Pitfalls of AI-based photo editing
Not only are AI-based tools for photo editing not up to the task, but using them could lead to plenty of pitfalls. Here are a few reasons why AI tools need to be used with caution, particularly for real estate photographs.
Loss of creative control
Creativity is the focal point of photo editing. Enhancement or suppression of certain characteristics in a photograph requires understanding the context and is driven by a goal in mind—an idea of what the outcome is to be. AI systems lack proper context and have no intrinsic goal or desire.
Using AI-based photo editing means ceding the creative control either fully or partially to an algorithm. AI tools do not have the understanding of the photographer’s artistic intent or the specific requirements of a real estate photograph. It may thus apply standardized enhancements, diverging from the unique aesthetic preferences of the photographer or the client. Moreover, with AI systems there is less scope for infusing human creativity into the editing process. This makes it difficult to emphasize important and unique features to create a compelling visual narrative that photographs are meant to convey.
Loss of authenticity
The concept of authenticity is fuzzy. And some may claim that a photograph that has undergone some modifications regardless of whether a human or an AI made the changes is no longer authentic. The argument is that to be truly authentic is to be original. But such a hard stance misses the point. Authenticity is about perception, too. Authenticity can be retained, even enhanced, when we apply changes to a photograph fastidiously.
But that requires deftness and a deeper understanding of what the subject of the photograph is. AI systems lack that crucial awareness. Reliance on them can thus lead to loss of the unique character and atmosphere of a property. They may prioritize generic standards of beauty over showcasing specific features or personalities of the space. And they may remove or alter distinctive characteristics that make a property stand out, rendering it bland and distinctive.
Homogenization
A similar problem to genericity and inauthenticity is homogeneity—all output looking more or less the same. AI systems cannot but be conformists. They learn and adopt conventions and aesthetics that their training data supply. They can only work within bounds—the training data and the design of the algorithms—and cannot extrapolate, only interpolate. If an AI model is trained with, say, data from the West, then it will struggle to apply its learning to samples from the East—or worse, westernized eastern traits. This may homogenize images similar to how social media has contributed to the narrowing of views.
The homogenization tendency of AI systems is problematic, particularly for real estate photos where it is important to highlight certain characteristics and peculiarities that may not be conventional. AI tools may apply standardized edits across a set of images, reducing diversity and uniqueness. AI algorithms may also inadvertently smooth out distinct features that make a property special.
AI systems may also opt for a one-size-fits-all approach applying enhancements without or with little consideration for the individuality of each property. This can add monotony to the photos subtracting their appeal.
Privacy and data security
AI models are trained with millions of images taken from the web, often without consent. Ethically contentious as this is, it is far from the only issue. Consumers have reasons to care, too. AI tools can not only be intrusive but also be a gateway for personal data to escape.
About 9GB worth of user data was leaked, for instance, from a popular AI photo editing tool, Cutout. The data include usernames and images that users queried. Incidents such as this should make one wary of using them. Media intended for private use could get into the hands of malicious agents through AI-based photo editing tools. And since most processing is done via the cloud, the risk of interception and leakage swells.
Another related issue is using consumers’ data without consent to train the model. Lensa, another popular AI portrait app, was accused of using user content to train its algorithm. The information and content that are shared with the tool may not be properly safeguarded—and who can tell what companies do with them besides using them as feedstock for their model?
Erosion of trust and reputational damage
AI systems, however seemingly smart and knowledgeable they are, do not have a complete picture of the world. Their conception of what is good may not therefore align with reality. This may lead them to create undesirable results, which may lead to embarrassment or even reputational damage.
Using AI for photo editing may also lead to potential clients and buyers feeling deceived. Using AI tools without disclosing to clients that images have gone under their knife may lead to a loss of trust. So, consent must be specifically sought especially when using AI for altering photographs, and disclaimer explicitly provided when done so. To preserve and foster trust, ensuring that the visual representation aligns with reality is essential.
Avoiding the pitfalls
Clumsy as the AI tools are, they are here to stay. And crass though they are, their progress is rapid. Thus to be a Luddite and disregard or discard them is unwise. A wiser move would be to use them, perhaps for more trivial works and to use them sparingly. If properly utilized, they can be helpful and make photographs look amazing. But they can also make amazing photos look appalling.
Balancing these extremes is a human in the loop. AI tools are efficient but they do not have a concept of taste nor do they have the story-telling capability that is required of a photograph. A human is required to infuse these, to breathe life into the images. Real estate photo editing services provided by humans or humans working in tandem with machines play a significant role in this life-giving function.
AI may one day progress to such a degree that it may be able to edit images perfectly. Its idea of perfect may however diverge from a human’s. Moreover, the allure of presenting a perfect photo should be weighed against potential ramifications such as inauthenticity, misrepresentation, and ethical concerns. For this and other reasons, it is more sensible to outsource photo editing services to human experts than to delegate the task to insentient machines lacking taste.