When Take-Two Interactive laid off its AI team in April 2026, Dr Luke Dicken found himself out of a job. His title at the time? Head of AI — for one of the biggest videogame publishers on the planet, the company behind Rockstar Games, Gearbox, and the upcoming GTA 6. You’d assume someone in that role would be a full-throated advocate for the technology. You’d be wrong.
In a detailed interview with gaming outlet The Side Quest, Dicken — now running his own consulting firm, LuDic AI — spoke at length about his genuine reservations concerning generative AI in game development. His scepticism isn’t that of an outsider lobbing criticism from the margins. It’s informed by years sitting at the intersection of innovation and industry reality.
What Dicken’s Team Was Actually Building
The story of Dicken’s AI team begins not at Take-Two, but at Zynga back in 2019 — well before ChatGPT entered the public consciousness and generative AI became shorthand for all things artificial intelligence.
His team was an R&D innovation group based in Zynga’s San Francisco headquarters, focused on applying traditional machine learning to real game development problems. They built systems that analysed how individual players interacted with mobile games — around 40 different behavioural metrics — then adapted the experience on the fly. One project, Spell Forest, proved the thesis worked: Zynga could measurably improve retention by quietly tuning the experience to suit each player’s style.
By the time Take-Two acquired Zynga in 2022 and ChatGPT went viral, Dicken’s 25-person team suddenly found itself handing governance over generative AI tools company-wide. Of those 25 staff, only three or four were directly working on genAI. The majority carried on building traditional AI systems that were quietly effective but unglamorous.
“Generative AI is not something that I have ever been particularly passionate about. It was something I think there’s a moral obligation to see managed as best as can be.”
— Dr Luke Dicken, former Head of AI, Take-Two Interactive
Why He Doesn’t Trust Generative AI
Dicken’s concerns are multi-layered. On the ethical side, he points to the now-well-documented practice of training large language models on artists’ and writers’ work without consent. “I want to be able to look my friends in the eye and know that I am not making their life worse,” he told GamesIndustry.biz in a separate, earlier interview.
Then there’s the technical criticism, which cuts even deeper. Dicken describes LLMs as fundamentally biased towards the statistical average of their training data — always pulling towards the middle of the bell curve.
“At some level of abstraction, LLMs are a next word predictor system,” he said. “That means it is statistically always going to be biased towards the mean of what it sees in its data. If you don’t know code or are a bad coder, AI can make you a mediocre coder. But if you are a good coder, these systems can also make you a mediocre coder. It’s regression to the mean as a service.”
He also flags the fragility of these systems. A small change to training data can produce dramatically different outputs — and companies using LLM pipelines are, in his words, “outsourcing” critical tasks to a model that could fundamentally change overnight without warning.
The Poisoning-the-Well Problem
Perhaps Dicken’s most pointed concern is about what happens when the generative AI hype cycle inevitably cools. The games industry has seen AI bubbles before. Each time, an oversold wave of enthusiasm leads to disillusionment — and that disillusionment often sweeps away legitimate, genuinely useful technology along with the hokum.
Traditional AI techniques — the kind his team built at Zynga and Take-Two — have real, proven applications in game development. Adaptive difficulty systems, player profiling, procedural content generation done right. Dicken worries that if generative AI craters spectacularly, executives and developers will throw the baby out with the bathwater.
“My worry is that generative AI is poisoning the well,” he said plainly. “I don’t think there is enough sophistication and nuance to retain the traditional stuff. For LLMs, we have already stumbled into the trough of disillusionment.”
Interestingly, Dicken also notes a geographic split in player attitudes. Western audiences — particularly in North America and Europe — are markedly more hostile to generative AI in games than their counterparts in Asia. He’s cautious about drawing firm conclusions from that, noting that the loudest online voices don’t necessarily represent the average player.
What This Signals for the Industry
There’s an undeniable irony in the head of AI for a company like Take-Two — whose CEO Strauss Zelnick has himself questioned some AI applications — being the one to sound the loudest alarm bells. Dicken has compared it to being a gold dealer who won’t sell you gold: “The guy should be selling you the gold if he’s actually got the gold.”
What makes his position worth taking seriously is precisely that it doesn’t come from ignorance or protectionism. He’s not a developer worried about job security, or a critic who’s never touched the tech. He ran the AI strategy for one of gaming’s most powerful publishers for over a year. His read on where the industry is heading — that the enthusiasm is outrunning both the ethics and the evidence — carries real weight.
The question now is whether studios and publishers listening to his warnings will act, or whether the financial pressure to chase AI-driven cost savings will win out regardless.
