Tobias Hann is the CEO of MOSTLY AI. He is a serial entrepreneur and former management consultant. He holds an MBA and PhD in business.
You may not know it yet, but you don’t need to look too far to realize that generative AI is already a big part of our lives. Whether you’ve opted for a cartoon version of yourself as a profile picture, peeked at what you might look like at 90 (gulp), or your personal identity has been transformed into a synthetic version of you somewhere out there in the ether—using a synthetic data generator—that’s generative AI doing its thing. And if the last example sounded like something out of a sci-fi movie, know that this is indeed already happening—and you’ll be happy to learn why this is a good thing for you.
But what exactly is generative AI, and why should we care?
Defining Generative AI
As explained in the Complete Guide to Generative AI in 2022, “Generative AI leverages AI and machine learning algorithms to enable machines to generate artificial content such as text, images, audio and video content based on its training data, in a manner which tricks the user into believing the content is real.”
It’s called training data as the AI learns from existing (raw or production) data to create new content. This may be images, audio files or even structured data sets. Although generative AI may be something almost imaginary in how we perceive it, it’s rooted in reality and based on real data, of which a lot is needed for the training (or generation) to be effective, and accurate.
Better Than Real?
Some might argue that it’s a distortion of reality, but a more progressive perspective is to view generative AI as a new layer of reality. And when you consider what’s already being done with this new technology, it’s arguably better than real. A malleable layer that allows for creativity and experimentation—and sometimes even the greater good. Its unrealized potential is massive.
A good example of generative AI is the video created of David Beckham speaking nine different languages, as part of a drive to raise awareness and help prevent child deaths caused by malaria. The example is straight out of Hollywood and highly convincing. The technology used here is called visual (in this case, face) synthesis and voice cloning, also referred to as “deep fake” technology in the media and entertainment industry. The synthesization process involved the use of raw video to build a 3D model, which was then re-animated in order to change the language and script. The campaign, known as the Malaria Must Die initiative, achieved 400 million impressions shortly after launching.
Another healthcare-related example of the application of generative AI is the early identification of disease. Here, for example, a generative adversarial network (GAN) enables the production of different angles of an X-ray image to visualize the possible expansion of a tumor. This creates more training material for tumor detection algorithms to work more effectively.
Generative AI also enables text-to-image generation. A simple description of what you want pictured, and boom, the world is your oyster! Imagine, for example, something as simple as a comfy avocado armchair. Or what about something more complex like the scenario that The Girl with the Pearl Earring may have found herself in, in line with how you might imagine it.
Both these examples, featured on OpenAI, make use of a neural network called “DALL-E.” The latter example demonstrates a new feature called “Outpainting” which enables users to extend an image beyond its original borders, whether by adding visual elements in the same style or even taking a story in a completely new direction.
This new tech is constantly stretching the limits of our imagination. Sparking new ideas.
A Synthetic You To Protect Your Privacy
Let’s take a step back and look at a less glamorous application of generative AI, equally powerful and important to understand since it involves the protection of your personal identity.
Think about yourself as a customer for a moment, about how many businesses have your personal information housed in their data warehouses. Even if they have your permission to store your details and notify you of relevant promotional offers, this does not guarantee your information will not be leaked at some point.
Data leaks are not going away any time soon, so businesses focused on enhancing personal and relevant customer experiences—while remaining committed to protecting your privacy—are fast waking up to the value of synthesizing their structured data. By structured data, I mean the hundreds/thousands/millions of rows of data that live in places like databases or CSV files. We’re talking about billions of data points, and this number continues to grow.
Here, AI trains on the original data and generates a synthetic version of that data which is privacy safe, with zero links back to any original data points. Not only is it statistically representative, but the data can be modified during the synthesization process; for example, an existing bias can be corrected to produce a more balanced data set.
It’s the latest in data anonymization. A key tool to de-risk innovation and democratize data. And the best thing about it is that while businesses can innovate to the max and offer you better products and services, your privacy remains intact.
Generative AI is here to stay, and it’s changing our reality in thousands of visible, and sometimes invisible, ways. Yes, there are risks, as with all new technologies. But progress, I believe, is about embracing and even mastering newness which includes preempting and managing risks along the way. The applications of generative AI that I’ve touched on merely scratch the surface of what this emerging technology is capable of. We can—and absolutely need to—make the most of it, and continue to work together to realize ways we can use it for the greater good.