What do a Shakespearean sonnet, a Drake song, and a professor’s lecture notes all have in common? All three have been replicated to some degree by generative AI, one of the latest forms of artificial intelligence technology to disrupt human content creation.
At an event organized by Columbia Business School’s Digital Future Initiative (DFI), Professor Stephan Meier moderated a conversation between Professor Olivier Toubia and Golnar Khosrowshahi ’97, founder and CEO of Reservoir Media and DFI advisory board member, on what the advent of GenAI means for the future of the human creative process.
The panelists emphasized that GenAI, such as Claude — a popular Anthropic-powered AI assistant — and ChatGPT cannot be viewed through a binary lens. Rather, AI-augmented human creativity, the collaboration between machine learning technology and humans, may soon become the norm.
“We need to frame this narrative to not be ‘AI or,’ because that’s how we’re speaking about this all the time — AI or human,” said Khosrowshahi, who is also a classically trained pianist.
To argue that human creatives must work separately from AI is binary and outdated, according to Khosrowshahi. Instead, human-AI collaboration can often produce work that is superior to non-AI augmented works, while still maintaining human creativity at its core.
“We need to reframe [the conversation] as ‘AI and,’” Khosrowshahi said. “Human plus AI is a more powerful creative engine because of the efficiencies that can be enabled, because of the knowledge that we can have about music, how we can develop and understand sounds, how we can understand blending cultures and creating new sounds. That is, in effect, what happens in the music business day in and day out.”
Reservoir is an independent music company, publicly traded on Nasdaq, that is contending with the role GenAI plays in shaping music creation, intellectual property laws, and the rights of human music creators.
To illustrate GenAI’s limitations, Khosrowshahi shared an anecdote from her recent experience asking Claude to write her an original folk song that included “country roads” as a subject. Instead of a wholly original song, the AI created one with a chorus that was identical to the popular American folk song “Take Me Home, Country Roads” by John Denver.
“Claude is very good at making art, but Claude is also very good at making art learned from the craft, the skill, and the expertise of people like John Denver. So I don’t think our days are numbered,” Khosrowshahi said.
Below are three additional takeaways from the panel’s conversation:
1. When it comes to prompt engineering, the onus may be on technology to get better.
Structuring text so that it can be understood and processed by GenAI can be a challenge for professionals looking to explore AI-augmented creation, according to the panelists. This process, known as prompt engineering, can make all the difference in improving the quality of output from an AI system. However, it may be too early to expect GenAI to produce the best creative output at all times.
“Part of that burden is on the technology, and then the other part of it is for us to actually learn from trial and error,” Khosrowshahi said. “Maybe at some point that becomes a streamlined process, but this [technology] is just going to evolve and get better.”
2. AI’s role in the human creative process is secondary.
While much of the conversation surrounding the viability of GenAI’s creativity positions the technology as the most prominent creator — a songwriter, for example — it’s GenAI’s background applications that should be explored, according to the panel.
An audience may not connect with GenAI’s lyrics or even scoff at its tendency to copy human-made songs word for word, but a commercially successful song takes more than just a singer-songwriter.
“I think that there are a lot of applications in our business that people are not talking about,” said Khosrowshahi.
For example, she noted that GenAI could soon be applied to how the music community might “license music, assess the metadata around our creative product, and understand why people love certain trends and why they're embracing that.”
Additional GenAI applications may include sound engineering and mastering, accelerating the time at which the music industry can get a product to market.
3. The demand for human-made products isn’t going away.
Toubia, the Glaubinger Professor of Business at CBS, posited that no matter how good GenAI gets at creating content and no matter how much a business incentive there is to use cheaper GenAI over its human counterpart, there will always be a demand for human-crafted products.
“Just like people put a premium on human interaction in the music industry, I think that also in the innovation space there is a premium for human-crafted, human-created products,” said Toubia, whose research revolves around idea generation and other aspects of innovation. “Knowing that this was made by a real person would probably have a premium on the market.”
He added that GenAI still can have a place in bringing new products to market, by summarizing market research and new consumer insights, for example. But humans still have a major role to play in the creative market.
“The inspiration for creativity comes from life, from things that happen to us,” said Toubia. “But AI doesn’t have a life, so it doesn’t have these random things happening to it that are going to inspire creativity.”