You’ve probably read about the recent wave of generative AI startups that have captivated the attention of the public and investors. The venture-capital herd has suddenly migrated from crypto and metaverse to generative AI. But while the mainstream media focuses on the billion-dollar valuations, hype, and fear, they ignore the real revolution.
Generative AI is an emerging technology that creates content rather than just analyzing content. This technology will accelerate the process of leveraging computers to think for us. (To learn more about this process, see our book, Become an AI Company in 90 Days.)
About generative AI
Generative models (also called synthetic media or generative media) fall into a class of AI that allows machines to use elements from media such as audio files, text, and images to produce content. Although the technology is not new, in the past year researchers have achieved stunning results that have captivated the attention of the public and investors alike.
Examples of generative AI projects include GPT-3, DALL-E 2, Stable Diffusion, GitHub Copilot, and Jasper. Researchers are making rapid progress with generative models that can produce digital content such as illustrations, photos, videos, stories, movie scripts, and software.
Although generative AI is evolving at a breathtaking pace, we’re still a long way from AI that can replace artists, software engineers, and writers. The initial applications of generative AI will help people create content more efficiently. Just as the world’s best chess players are now human + AI teams, content creation is rapidly evolving to a more efficient partnership between AI and people.
The first business application: Digital marketing
Over the past 25 years, digital marketing has supplanted all other forms of product promotion. The Mad Men era of advertising campaigns driven by teams of strategists, artists, and copywriters has been replaced by an iterative concept that relies on Google Ads and Facebook. Instead of spending months designing and testing an advertising campaign on traditional media outlets like radio, TV, and newspapers, smaller marketing teams can now rapidly and cheaply test digital ads online.
A process that took years a few decades ago can now be accomplished in weeks by using digital marketing. Generative AI will further condense the timeline for digital marketing projects from days into hours.
Consider a few examples based on my personal experience with sourcing blog illustrations, writing headlines, and advertising online.
Designing blog post images
The most frustrating part of blogging for me is identifying the right images for my blog posts. My available options are often slow, expensive, irrelevant, or visually inconsistent.
My options for illustrations
To add an illustration to my blog, I have a few options:
- Design it myself. This is the worst possible option. It takes me too long and the results are terrible.
- Buy a stock image. Although many services provide high-quality images for a reasonable price, finding a relevant image that also fits with the blog style is tedious. Without continuous effort to keep image styles consistent, a blog can quickly devolve into a hodgepodge of random images.
- Hire a designer to create customized blog images. A professional designer can build custom blog images that align with the content and the visual aesthetic. But the process is expensive and slow because I need to iterate ideas with the designer. And if I use a different designer later on, maintaining a consistent style is practically impossible.
How generative AI could help
Generative AI has the potential to help me rapidly generate striking, consistent, and relevant blog images myself. I can seed a generative model with a bit of language or a particular style and rapidly iterate through ideas. Although the result might not be as good as something created by a professional designer, the technology promises to provide fast, relevant, inexpensive, and consistent results.
As advertising tycoon David Ogilvy famously wrote, “On the average, five times as many people read the headline as read the body copy. When you have written your headline, you have spent eighty cents out of your dollar.”
Whether I’m writing a landing page title, a sales email headline, an ad, or a tweet, assembling the right seven to fifteen words in the right combination is the most important part of the exercise.
How I generate copy headlines
I’ve found that the approach advocated by Joanna Wiebe of Copyhackers is most useful for entrepreneurs like me. Here’s my process:
- Generate a series of headline ideas in a spreadsheet.
- Evaluate and iterate word combinations by using services like Headline Analyzer and Sharethrough.
- Select the best five to ten versions and copy the headlines onto a set of slides.
- Get feedback from my team or others who are familiar with my sales goals.
- Run a series of sample campaigns on Google or Facebook to see results.
- Repeat these steps as necessary.
How generative AI could help
Although effective, my headline-writing process is slow. It also takes my full attention and the most creative part of my day to generate new ideas. Generative text models could help accelerate my process by creating example headlines that are likely to capture my audience’s attention.
Innovations in digital advertising garnered significant credit for Barack Obama’s digital strategy in the 2008 and 2012 campaigns. The campaign team optimized fundraising by continuously iterating through ads (often called A/B testing) as the campaign evolved. Generative AI can significantly improve the efficiency of A/B testing by accelerating the creative parts of the process.
For example, a generative model could write ad copy, create an accompanying image, and test the combination in online venues. The model could then use the test results to adjust the ad copy and image and to test the combination again. The model looks something like the following diagram:
This process won’t be AI-driven magic. Instead it will work like GitHub Copilot, where generative models take much of the grunt work out of coding, making the process more efficient. Large teams of specialists in domains such as digital marketing, design, and copywriting will be replaced by smaller teams that orchestrate the process for the generative models.
The future is clear, but the timing isn’t
Like most disruptive technologies, generative AI will change the way we work. It isn’t hard to imagine how problems in our current world can be handled more efficiently by using generative AI. But it is hard to predict when and how these changes will happen.
The technology is still in its infancy, and the practical applications are still limited. But if this article is interesting, I suggest you sign up for a DALL-E 2 and have some fun generating images. You will quickly realize the possibilities (and current limitations) of this exciting technology.