‘Datasets Are Backward-Looking’: Grand Theft Auto Maker’s CEO On Why AI Can’t Replace Creativity

Strauss Zelnick, chief executive of Take-Two Interactive, drew a sharp distinction between artificial intelligence as a tool of efficiency and the nature of true creative work, arguing that AI systems are fundamentally rooted in past data while creativity must look forward.

Speaking about AI’s role in technology and business, Strauss Zelnick said he is broadly supportive of tools that improve efficiency, describing himself as “all in on them” when it comes to technologies that streamline production and operations but he was more cautious when it comes to claims that AI can meaningfully replace human creativity.

At the core of his argument is how AI systems are built as Strauss Zelnick described AI as “big datasets, lots of computers and a large

language model mushed up together,” emphasising that these systems are trained on existing information.

In his view, that makes AI inherently backward-looking.

View this post on Instagram

A post shared by Metav3rse is for the Culture (@metav3rse)

“Data sets by their very nature are backward looking,” he said, arguing that AI reflects and recombines what has already been created rather than originating something truly new.

He contrasted this with creativity itself, which he described as inherently forward-looking.

“Creativity by its very nature is forward looking,” Strauss Zelnick said.

Creativity, he suggested, is not simply the reproduction of prior ideas but the generation of something that breaks from them.

While acknowledging that all human thinking is shaped by prior exposure- books read, podcasts heard, experiences accumulated- he argued that creative output cannot succeed if it is merely derivative.

“Creativity is informed by data,” he said, adding, “Like you are informed by the hundreds of books that you have read or you are informed by the podcasts that you have listened to. How could you not be?”

But he warned that over-reliance on imitation undermines originality.

“If a podcast was just a highly informed clone of another, who would watch that,” he said, using the example to stress that audiences ultimately reject derivative work.

Strauss Zelnick also pushed back against the idea that AI alone can meaningfully generate original creative properties at scale, warning against assumptions that efficiency automatically translates into creative success.

“And the thesis that wow with AI we can more efficiently we can create a completely derivative property like derivative properties don’t work,” he said.

Instead, he framed AI as a tool that can support and accelerate production, but not replace the forward-driven nature of human imagination.

“So AI will make it easier and quicker and better,” he added, emphasizing its role as an enabler rather than a substitute.

He also stressed that embracing technological progress does not mean dismissing its value.

“So I don’t take the benefits of technology lightly at all,” he said, reinforcing that he is not skeptical of AI’s practical benefits.

However, he maintained that creativity remains a distinct domain- one that depends on originality, judgment and the ability to move beyond existing datasets.

“So that’s where the thread has been lost that AI is really great at asset creation,” he suggested, drawing a line between efficiency tools and true creative authorship.