The last few years have seen explosive growth in power and popularity for generative AI tools like Dall-E and ChatGPT. While other types of AI tech are also popular, gen AI leads the charge. GenAI tools are trained by human-created content to produce new text or images based on word or phrase prompts. We’ve seen tech companies make major investments to introduce gen AI onto their platforms. These investments come with the promise that the real AI revolution is just getting started, that the power of these tools will keep accelerating upwards.
Modern day marketers are curious about these new technologies, often because of the growing pressure to “do more with less.” Budgets shrink, timelines shift, and goalposts move faster and further away. You might be starting to dip your toes in the water or you’re already a pro user. At Catch+Release, we are generally supportive of AI technology. For example, our own platform features an AI-enhanced semantic search engine. It’s AI technology, but it’s not a GPT.
GenAI has taken on many controversies and received much backlash that other AI technologies like ours don't face, due to how it’s created. GenAI requires human-made creations to train from, and many creators are demanding that this practice stops or is severely regulated. Having the actual human artists and writers on the opposite side of the issue has created a major stigma against using gen AI. Marketers using generative AI for art or copy within their text have gotten major pushback in many cases, leading many brands to ban the use of gen AI in their campaigns.
On top of this growing stigma, AI companies may be hitting limits in what the tools are actually capable of. If your hope is that generative AI will someday be so realistic that detractors won’t even detect it, don’t be so certain. The fact is, we don’t know what the future of AI will bring. Will AI companies’ promise of continued acceleration of improvement all the way to superhuman “general AI” prove true? Or will we end up in another AI winter as hype fizzles out?
Are Generative AI Companies Running Out of Training Data?
Generative AI, like ChatGPT and Dall-E, works by building connections between huge amounts of classified data – a process known as “training”. One of the most simple and reliable ways to improve the accuracy and output diversity of generative AI is to train it on even more data. But this raises an obvious question: how much data is there to work with?
The internet is huge, unfathomably huge, and grows every second. But it isn’t infinite. Moreover, the subset of internet data suitable for training (labeled meaningfully, in the required language, etc.) is much smaller, and grows much slower than the internet as a whole. The easiest to access and most useful data has already been incorporated, and expanding training sets further gets harder and harder for less impactful returns.
But that isn’t the only issue AI companies are facing with training data. Currently, copyright laws are ambiguous regarding whether training on copyrighted data is a violation. Thus, unsurprisingly, most commercial models do include copyrighted data. This could all change depending on the outcome of upcoming bills and court cases. AI models could have a huge chunk of valuable data removed from them, or be required to pay major sums to license them.
All of this is to say that humans are still the essential core of generative AI. When we say that training data is valuable, what makes it valuable is the human creative energy put into it. When someone says that a generative AI’s output is “good”, they mean that it replicates something they appreciate in human-created content. It’s our human judgment that gives AI output any worth. In essence, the power of AI is really the power of human created content, and any advancements in AI means an advancement in human creativity. So why not just work with human creativity at the source?
What’s a Trustworthy Alternative to Generative AI?
So let’s say you’re a marketer looking to use some content to enhance your marketing campaigns. You want something that’s specific to your campaign’s needs, you want it fast, and you don’t wanna break the bank. This sounds like a job that only generative AI could solve, right? But there’s another category of content that fits the bill: content made by real humans. Content that’s raw, real-time, and honestly reflects our world.
As the world gets more online and apps like TikTok grow and grow, the amount of user-generated content on the internet is ramping up at a rate akin to the acceleration of AI power. Granted, we may be seeing a boom of user-created content right now as it hits the mainstream. But unlike the inevitable slowdown of generative AI, there’s also good reason to expect steady growth after that: the population keeps increasing! If more people keep coming online and sharing their lives, the pool of user-created content will keep growing too.
Not only is user-created content immediately available and diverse enough to meet your specific needs, it provides a sense of cultural authenticity and relevance that AI content can never hope to match. At Catch+Release, we believe in empowering marketers to make the most of user-created content, and empowering creators to be credited and compensated for usage they consent to. Learn more by exploring our marketplace of community content!
Addendum: Is Generative AI Running out of Stuff to be Good At?
There’s reason to be cautious around even relying on gen AI as it currently stands to remain available. AI companies have invested billions into developing features into their products, and spend millions on keeping them afloat – a report in 2023 showed ChatGPT costing $700,000 per day to run, which is likely even higher today.
At the same time, some of the most high profile rollouts of generative AI, like Google’s AI search results summaries, have been met with criticism and controversy. Recent surveys have found that not that many people are actually regularly using the tools. As the novelty of generative AI fades, and impressive improvements stagnate, the day may come when keeping things like ChatGPT online simply isn’t worth it.
Are there that many things left for AI to do? If there is another major frontier for AI to tackle, it’s in the physical world: autonomous vehicles, factory line robots, delivery drones, and more. We’re seeing tentative steps into this domain, but before this tech can be deployed en masse, it needs to navigate a minefield of legal regulation across many industries, far tougher than the regulations currently catching up to online AI applications.
AI has actually been around much longer than most people think. It was first introduced in the mid 1900s and has been embedded in a lot of the technology that we use regularly as everyday consumers. It’s just been hidden from center-stage. For example, Google Search has used AI for decades to populate the best results for its users. Turns out we’re not all actually as good as we thought we were at boolean searches.
All these tools are great, and we are hopeful AI will continue to enhance the lives of humans across the globe. That said, we see its limitations. AI is not a replacement for humanity, the essence of who we are, our connections, our creativity, and our emotions. As marketers in a creative field, it’s important to consider how we leverage AI technology in our work. Using it as a helpful tool to reduce toil makes sense, but completely replacing our creative intelligence doesn’t. Remember, authenticity is what performs, because humans are drawn to humanity.
P.S. Like the blog’s thumbnail? Shoutout to @rambosphotos from our Creator Community!
