Automated Advertising Agencies: Custom Procedural Adverts

Imagine a video advertisement pops up during your favorite media stream. It demonstrates a product you've been interested in. It highlights the options and features, already demonstrating things like your favorite colors and most desired preferences. It features artificially-generated actors that look similar to the people you most associate with. It sounds like you and your friends talking, using the same dialects, slangs, and mannerisms as your closest connections. The advertisement feels incredibly personal and made specifically for you, even talking directly to you by name or featuring your digital likeness. This advertisement is uniquely personal to you and only you. Could the future of advertising be individualized and tailored to each viewer, in real-time?

Intelligent Prodedural Advertising System:

Through the use of aggregated personal meta-data, an algorithm or artificial intelligence system could compile procedurally generated advertising specifically targeted to your most recent interests and interactions. This concept would require an exponentially expanding bank of visual and audio assets to represent every kind of thing imaginable, each broken apart as the most modular components possible and tagged with context labels. This way any representation of any kind of person or thing could be combined with any kind of style, theme, or pattern to create the most appealing thing to that unique person. With the use of a complex web of "context labels" or tags, everything can be pulled from the asset banks and combined to create the "perfect" piece of enticing media for that person.

This system of visual assets would have to be as modular as possible. This could potentially be achieved with the use of realtime rendering technology, similar to the ever-evolving capabilities of video game engines, rather than the pre-rendered and mass-distributed video files of today. This system could use artificially-generated virtual actors in order to have an endless cast of any combination of people imaginable. These people would be completely computer-generated and therefore easily represent any type of person, without exploiting real people of any kind. These artificial-actors would not require contracts, do not have controversial human lives, do not have work schedules, or any other limitations and risks when compared to human actors. These actors are brought to life with a near-limitless modular system of animations, sounds, and visual styles in order to create the perfectly appealing unique cast for each advertisement's unique viewer. The backdrop and setting would be compiled from another bank of modular visual assets to create a complementary setting for that specific advertisement and based upon desired parameters.

Companies would have their content added to the respective asset banks and labeled accordingly for them to be easily embedded in any advertisement that fits their contracted parameters. An advertisement could represent the products of a number of companies simultaneously in complementary ways, or one product exclusively. I have no doubt that companies could purchase various packages for exclusivity or levels of exposure, from background to the main subject matter. With every visual asset tagged it would be relatively simple to track the amount products were shown and to what level of exposure and divide costs or payouts accordingly. Categories could be easily flagged so that companies can choose what sorts of products theirs can/cannot be displayed with. Ultimately allowing the companies purchasing advertising space to better target their markets with advertising the viewers are more likely to enjoy, all translating to less money being wasted on wasting fewer people's time with things they don't actually like.

The system would evaluate the prospective viewer's interests, values, preferences, and most recent media interactions. These evaluations would be used to instantly compile a unique advertising experience catered specifically to that unique viewer. The system of algorithms would compile stories and settings based on those unique combinations of preferences. Overall these productions could be described as unique to each viewer, however, there are likely to be patterns among people with similar combinations of preferences. Therefore in order to be even more efficient, the system would be able to group or even layer some of these previous advertising combinations for another added level of modularity and efficiency to the system.

These advertisements are rendered in realtime and would, theoretically, be far more effective at converting views into purchases than the current way of trying to create general advertisements that capture as many people as possible. This antiquated "shotgun spread" strategy of modern advertising has proven problematic for both companies and the public alike. With society becoming ever polarized in all directions, it would seem that there is no way to create a piece of media that everyone will enjoy and approve of. Companies risk dangerous backlash no matter who they attempt to appease. Instead of trying to force everyone to enjoy the same thing that must be universally appealing, it would potentially be more harmonious if everyone experienced media that was relatively unique to them based on their personal interests and preferences without being forced onto anyone else. It would not magically solve those problems but it may mitigate one area of perpetual issue.

Addressing Monoculture versus Polyculture in Advertising:

With a system of adverts generated specifically toward each unique user, there is a great risk for the emergence of a repetitive monoculture of only viewing content the user already enjoys, with nothing new being added. This would likely seem advantageous at first for maximizing viewer attention, however, in the long run, this would become stale or less appealing to people. To combat this the advertising system could harness meta date to find a statistically appealing balance of directly preferred and tangentially related content. It could also experiment with showing content directly related to unique interests, opposite from interests, and anywhere in between; always monitoring the viewer's interactions with the media in order to learn each person's unique balance. If the system doesn't have such 'goodness of fit" balances built into it, then it would be likely that competing automated systems would emerge with offers of statistically more profitable "preference vs novelty" balance algorithm systems.

Going One Step Further:

With the emergence of augmented/overlayed reality technology, such as the google glass and similar devices, the potential of this advertising system expands exponentially. Every billboard and sign could be displaying uniquely personalized advertising to each person passing by that is automatically evolving with you. Within the growing marketplace of virtual reality, this will be a natural integration. The potential for advertising space within video games and other virtually interactive spaces, a topic worth its own discussion, is relatively untapped and currently filled with fake placeholder advertising. This can and will likely change as this technology becomes more integrated into everyday society.

Other Realms of Possibility:

This form of an automated procedural media creation system could potentially be used to generate any kind of visual content. I could imagine automated film production studios, television producers, and video game developers that function in much the same way. Will this ever be able to rival the creativity output of human artists and designers... maybe just challenge them to innovate?

Could this type of "intelligent" procedural advertising be the next potential evolution in targeted marketing?

Would this further trap us in echo-chambers of our own preferences, or liberate us from the tyranny of mass appeal?

- Jan

#future #automation #advertising #proceduralgeneration #thoughts