Take-Two's former AI chief on how to stop AI from taking over the tools that actually work
Summary
Take-Two disbanded its 25-person AI team in April 2026, a unit founded at Zynga in 2019 with proven ROI on mobile KPIs.
The team ran predictive and adaptive AI, not generative AI; only 3-4 members handled GenAI governance.
Former head of AI Dr. Luke Dicken warns the GenAI hype cycle risks discrediting all AI tooling industry-wide.
"My worry is that generative AI is poisoning the well. I don't think there is enough sophistication and nuance to retain the traditional stuff."
01
Take-Two killed a validated R&D unit, not a GenAI experiment
Dicken's team originated as a 2019 Zynga skunkworks, predating ChatGPT and the GenAI label entirely.
Its core product: a machine-learning profile system tracking ~40 behavioral metrics to adapt gameplay per player in real time.
2020's Spell Forest validated the thesis with measurable impact on core business KPIs.
At peak, the 25-person team supported multiple Zynga studios and later governed AI across all of Take-Two.
Only 3-4 members worked on GenAI governance; the rest did traditional AI development.
02
Dicken flags compounding quality, reliability, and business risks in GenAI
"LLMs are a next word predictor system... statistically always going to be biased towards the mean."
"If you are a good coder, these systems can also make you a mediocre coder. It's regression to the mean as a service."
On reliability: minor training data changes can cascade unpredictably across model behavior.
"One little change somewhere can have a cascading effect across the neural network. The fact you don't have control over that scares the shit out of me."
Dicken cites tech analyst Ed Zitron's argument that AI economics do not stack up on ethics, legal, or business grounds.
03
If GenAI disappoints, executive appetite for all AI tooling collapses with it
Dicken draws a direct parallel to prior AI hype cycles: overpromise, flood funding, underdeliver, discredit the category.
The specific publisher risk: GenAI backlash drags proven predictive and adaptive AI systems down with it.
"For LLMs, we have already stumbled into the trough of disillusionment."
GenAI hype has made studios more receptive to AI conversations broadly.
But that receptiveness is shallow and unlikely to survive a correction.
04
Right GenAI posture depends on company size, legal exposure, and values
Dicken rejects a universal policy; context determines the correct stance.
Small studios facing closure: near-unrestricted GenAI use may be rational survival, with ethical tradeoffs acknowledged.
Large publishers like Take-Two: legal risk dominates, requiring structured oversight on training data and model governance.
"The morally correct answer is no genAI. The business correct is just enough genAI, and where you draw that line is going to be values-dependent."
Events
Companies
Games
Locations
—
COMPILED BY GAMES ATLAS EDITORIAL
RELATED ARTICLES
EXPLORE MORE
Comments (0)
No comments yet. Be the first to share your thoughts!