As the AI revolution continues apace, one thing is clear.
Thoughtful human-made content will remain the most important source of information for consumers online. And our appreciation for human expression will only grow as we experience more diluted and derivative outputs created by AI models.
The allure of generative AI also presents interesting considerations for Google, which confronts the fate of the content once elevated by its search results.
All of us, Google included, should recognize the irrevocable value of the individual point of view as fundamental to our culture. The commercial alternative would be reducing an iconic search engine to just another in a rapidly growing line of AI response machines.
Different modes of expression
The centrality of human participation in quality content creation endures for two reasons. Generative AI can’t replace the human mind as the indispensable engine for new ideas. And the refined knowledge of one trusted contributor is more valuable than consensus outputs generated by AI.
You might argue: “Isn’t all human creation the reflection of prior works, making AI-generated content as ‘new’ as any other?” Yes and no.
Writers, musicians and artists typically acknowledge the influences that inspire them. The Rolling Stones don’t exist without Muddy Waters; there is no Claude Monet without Edouard Manet. That’s the nature of cultural evolution.
But it’s not how generative AI works. Large language models process immense content pools and derive aggregate signals to form outputs: papers, analyses, research reports, studies, business processes and recommendations of all kinds. Those products can be valuable and time-saving. But an unrestrained AI-generated future could lock us into the status quo and recycle it forever.
When training AI, individual acts of creativity are reduced to LLM inputs, plunging us into an abyss of averageness. We risk losing truly remarkable expression.
The distinction between human thought and AI mechanization is even recognized in legal contracts. Nondisclosure agreements prohibit all forms of information copying and reproduction but often permit the use of “residuals.” Residuals are loosely defined as the impact that content has on the “unaided memory” of a human reader. Information enters the mysterious human consciousness, some amorphous impact is made and that person evolves or changes; from that influence, something truly original and inspiring can arise. But supercomputers cannot recreate unaided memory.
Quality content vs. noise
Another reason consumers will continue prioritizing human-made content: It’s of a higher quality than the AI-generated alternative.
When people seek advice, they want to know where it’s coming from. Articles about travel destinations and logistical nuances are best written by real people who have navigated those complexities and have the skills and comprehension to share their experiences. Stories about food (and recipes), wine, parenting, leisure, fashion and style, home and hobbies, how to live one’s best life and how to cope with struggles are all better told by specific individuals with a hard-earned point of view.
People are drawn to content they trust, and large language models are plagued by bias and trust concerns. When a model is “trained,” will consumers ever know how? And can we really overlook AI’s tendency toward obsequiousness and flattery in its responses?
When it comes to matters of media quality in the AI era, publishers have much to consider. The smart ones will use generative AI in ways that advance the human condition within a framework that preserves its finest human elements.
They’ll use LLMs to better understand how content gets consumed. This includes learning what types of content people need and want, how content consumption correlates with consumer intent and how that information can improve user experiences online. Those data points inform content creation and advertising strategies, making publisher assets more valuable and expanding overall reach, influence and relevance in a competitive and tumultuous digital environment.
Low-quality AI-generated content is just noise that’s easily ignored. It takes a human touch to elevate content and make it sing.
Consider these lyrics from “Tombstone Blues,” written by poet-musician and Nobel literature laureate Bob Dylan in 1965:
The ghost of Belle Star, she hands down her wits
To Jezebel the nun, she violently knits
A bald wig for Jack the Ripper, who sits
At the head of the Chamber of Commerce
What do Dylan’s words above mean? No idea. But sixty years later, people are still speculating about his meaning online. That’s the kind of engagement with the human spirit AI could never generate. And if AI wrote a song like that, it would seem cheap, pandering, nonsensical – computer-random as opposed to cool-random.
We may never know the exact meaning of Dylan’s lyrics. But knowing he meant to express something actually means a lot.
“The Sell Sider” is a column written by the sell side of the digital media community.
Follow People Inc. and AdExchanger on LinkedIn.
For more articles featuring Jim Lawson, click here.
