Illustration created by Midjourney with the prompt “a second empire home at the end of a long cobbled road, at the far end of which is a simple, unadorned wrought iron gate, which we see with a drone, the road to the gate is cobbled, 3D, great details, not too sharp, velvety lights –ar 3:2 –chaos 23 –v 4” Note: The house is not actually second empire style and there are only a few cobblestones, but the artwork achieves the effect I was going for — which is exactly the point of this article.
Synthetic media, sometimes referred to as “deepfake” technology, is the use of AI and machine learning to generate realistic, digital media such as images, videos, and audio. This technology is already revolutionizing the creative process as both artists and artisans conscript AI-coworkers for production assistance. But in the coming months, we are likely to see this technology evolve into Generative Synthetic Media (GSM) which we will define (for this essay) as data-driven synthetic media (created in near real time and surfaced in place of traditional media). When fully achieved, GSM will revolutionize human communication. This is a bold claim, so it requires bold evidence. Let’s review.
Synthetic Media vs. Generative Synthetic Media
Synthetic Media is produced by human creators who use computers and purpose-built software (including AI) to realize their production goals.
Generative Synthetic Media (as defined for this essay) is media that is produced by AI based only on a description of the desired work product. Importantly, the description for the requested Generative Synthetic Media can also be generated by AI.
Synthetic Media Today
To date, synthetic media has been used in a variety of ways, including the following:
Film and video production – Synthetic media has been used to create digital doubles of actors, allowing for more efficient and cost-effective filming. It has also been used to generate realistic special effects, such as explosions and fire.
Advertising – Companies have used synthetic media to create realistic and engaging digital characters that can be used in their advertising campaigns.
Gaming – Synthetic media has been used to generate realistic in-game characters and environments, making the gaming experience more immersive.
News and journalism – Synthetic media has been used to create realistic virtual news anchors, allowing for faster and more efficient news broadcasts.
Music – Synthetic media has been used to create realistic virtual musicians, such as virtual pop stars, and to generate new music tracks.
E-commerce – Companies have used synthetic media to generate realistic product images and videos, making the product more attractive to customers and increasing conversion rates.
Education – Synthetic media has been used to create virtual teachers and trainers, allowing for more efficient and effective training.
Generative Synthetic Media
In the very near future, we are going to see an explosion of applications built over large language models, such as GPT-4, BLOOM, GLaM, Gopher, Megatron-Turing NLG, Chinchilla, and LaMDA. Let’s imagine how a Generative Synthetic Media application that included workflows and processes might custom-create and deliver an advertisement for a specific target audience in near real time.
Data collection – The platform would gather data on the target audience from a variety of sources. Batch data might come from the sponsor’s CRM system. Real time (streaming) data might come from the behaviors exhibited by the user (clicks, location, etc.). The network might contribute additional data (venue rules, local regulations, etc.). The user might also contribute data by opting-in to a specific experience.
Audience segmentation – The data would be analyzed to segment the target audience into different groups with similar characteristics. This would allow for the creation of tailored advertisements for each segment.
Content generation – Using natural language generation (NLG), the platform would generate a script along with the required prompts tailored to the specific characteristics of each segment.
Hyper personalization – Based on the amount of available data and the prevailing privacy regulations, content could be created for specific individuals (rather than larger audience segments).
Optimization – The platform would use machine learning algorithms to optimize the content for different channels, such as social media, email, or display ads.
Real-time monitoring – The platform would continuously monitor the performance of the content and make adjustments accordingly, based on the data it receives, ensuring that the content remains relevant and effective.
When Will This Happen?
Generative Synthetic Media tools are emerging now. It will not be long until the ideas described in this article are simply table stakes. Don’t expect a photo-realistic, acoustically perfect, “Is that real or is that fake?”, custom-created commercial to appear on your favorite FAST service in the next few weeks. This will start with generative AI creating ad copy and still images (GPT-4 and Midjourney APIs), and then we’ll start to hear voice-overs and music. Next we’ll start to see on-the-fly deepfake videos, and ultimately, all production elements will be fully automated, created in near real time, and delivered.
If I had to put a timeline on this? More than a year, less than three years. Full disclosure: we’re working on all of this right now. So get ready. Generative Synthetic Media is just around the corner.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. I am not a financial advisor. Nothing contained herein should be considered financial advice. If you are considering any type of investment you should conduct your own research and, if necessary, seek the advice of a licensed financial advisor.