Much ado about generative AI: The money, the potential, the pitfalls

Part of the reasoning behind this momentum was explained in a recent A16Z blog post by Matt Bornstein, Guido Appenzeller, and Martin Casado. As they wrote, “Models like Stable Diffusion and ChatGPT are setting historical records for user growth, and several applications have reached $100 million of annualized revenue less than a year after launch. Side-by-side comparisons show AI models outperforming humans in some tasks by multiple orders of magnitude.” Put more simply, generative AI can be big business.

Generative AI is also seeing significant traction globally, particularly in the APAC region. In fact, Asia Pacific is projected to grow even faster than the U.S., with a CAGR greater than 35% from 2022-2028. The market is being driven by significant government initiatives as well as the widespread adoption of AI-based applications. This will further increase innovation and demand for new offerings throughout the world, as we are inhibited more by imagination than by technology.

But it’s not just the techies getting excited about generative AI. Everyday people are fascinated too. From school classrooms to local gyms to dinner parties, folks are bringing generative AI into the mainstream by playing around with ChatGPT or creating images with Lensa AI.

In fact, Lensa AI, a paid app, is seeing heavy traction among consumers. Lensa AI has made $16.2 million revenue in 2022, with $8 million in December 2022 alone. The app has been downloaded more than 25 million times, with 1.1 million active users at the start of December 2022. ChatGTP’s numbers are likely to be even more impressive with its paid product. After all, ChatGPT crossed 1 million users within a week of its launch and continues to captivate users and rapidly gain popularity.

So what does it all mean, and where is generative AI going?

The new creators

Among the most prolific use cases for generative AI are content creation and entertainment. People take joy in seeing what they can command an AI model to put together — be it with text, voice, static images or even video — which can then be shared with friends, family or the world at large. The potential for building new breeds of avatars could be particularly enticing if the metaverse rises to its potential. A new world is opening.

For example, generative AI can help people with no relevant skills or expertise use or create art. Previously, designs such as those now developed by the layperson required skills like handling Unity 3D, Unreal Engine, Adobe creator tools, and so on. Even the most limited compilations demanded a certain type of deep knowledge, training or special equipment.

Today, we live in an age where everyone is a creator but not everyone has the skill set to make a great piece of art. Generative AI levels the playing field and democratizes art.

Not only are people using generative AI to construct visual works or produce content, like a script or blog post. There are also advances that make it possible to bring multiple components together, like avatars that can be used in videos or chat.

Why would something like this matter beyond being something fun to show friends or a different way to create a school project? Generative AI can be so much bigger than that! It actually helps people of all ages and in any field bring their imagination into the real world.

If we’re looking for “big money” cases, though, think about when a casting director is trying to consider the right actor for a movie role or someone has an idea for a YouTube video that has the potential to get millions of views. It is now possible to generate an “ideal” actor using an avatar tool that can display emotions, change the tone and expression of the voice (think creating voice identities that do not exist in this world or applying a licensed, distinctive voice, like Homer Simpson or James Earl Jones), and so much more — all before seeking out a human actor.

Such a use case removes risk and saves time by identifying the right characteristics and testing one’s vision before a project moves forward. It is much more precise and efficient. They can see what works — and what doesn’t. Ultimately, making the right decisions from the onset can save a project up to several million dollars, depending on its scope.

Drawbacks of generative AI

But like any new or evolving technology, there have been issues. On the one hand, there is the potential for someone’s likeness and views to be co-opted or misinterpreted. Fortunately, sophisticated technical solutions are being developed rapidly to guard against misappropriation. If someone creates an avatar and addresses a sensitive topic, an example-based filtering algorithm can be applied, leading to an AI system understanding what their avatar would actually say or how it would answer. These capabilities are far beyond traditional filtering algorithms, which only reject things like swearing. 

Creation tools can present the AI system with an example of a political issue or article, and the user can provide feedback (good or bad). Based on that feedback, the avatar knows what to say or what not to say (details of the approach can be found in the research paper User-defined Content Detection Framework).

Similarly, OpenAI recently released a detection tool to discriminate AI writing from writing that was done by a human. A word of caution is that it is not yet 100% reliable. Mistakes remain prevalent, as this is very new territory, but progress is being made. Social consensus and ethics about using such technology are becoming increasingly important as complex issues of copyright, IP and plagiarism continue to emerge, with greater adoption and new use cases popping up daily.

Artists, writers and other creatives have spoken out about how generative AI is a type of cheating, one that devalues actual art. New lawsuits are working their way through the courts. One recent suit against Stability AI, image-generator startup Midjourney and online gallery DeviantArt brought by three working artists claims that AI image-generators are nothing more than “21st-century collage tools that violate the rights of millions of artists.”

Another suit filed by Getty Images is related to copying images without a license and infringing on intellectual property. The suits highlight two key issues with respect to generative AI.

First, the definition of art. What constitutes art? Who decides what it is and how it must be created? New generative AI creations are still the product of someone’s mind and experience; they are still expressions intended to provoke emotions and various responses. It’s simply that the method changes. Rather than creating with brushstrokes, the artist creates with questions and commands.

Second: Who owns what in a product created by generative AI? This is a bit stickier, but there is still plenty of safe space. Images, voices, text, etc., that are appropriately licensed and/or cited can be used, as well as work in the public domain. Additionally, generative AI can pull from characteristics of a voice print, facial expressions, etc., to build something that does not exist — and that has never existed — in the real world. 

In the future, expect to see far more licensing deals that reflect a changing world that incorporates generative AI, as well as changing laws related to intellectual property and copyright. This is inevitable as generative AI takes off.

Generative AI is just getting started, and there is so much yet to discover. The potential is there to open a whole new creative universe, but only if we ask the right questions and give the right commands in creation. Just as AI systems are getting smarter, so will the people that use them. And with the right tools, the results will astound us all.

Taesu Kim is the CEO of Neosapience.