Pim de Witte

The through-line

Across every appearance, Pim de Witte returns to one core conviction: that video, specifically the gameplay video generated by millions of gamers, is a more powerful foundation for building the next generation of AI than text, and that this makes games-native companies better positioned than language-model labs to build embodied AI agents and world models [3][4][5]. This belief has hardened into a strategic choice rather than just a research thesis: de Witte has been willing to walk away from very large offers to keep building independently toward that vision [5], and he frames the coming shift in AI's capability and speed as being on a scale far beyond prior technological revolutions [1].

On games and video as the real substrate for AI

De Witte's central research and business bet is that gameplay video, at the scale of billions of clips, is a uniquely rich dataset for training AI systems that need to understand and act in dynamic, embodied environments, positioning General Intuition's work on agents, video understanding, and world models as distinct from the text-first approach of the major language model labs [3][4]. This thesis is explicitly framed as an alternative to building AI "rooted in words," with games instead treated as the training ground for the next wave of AI [5]. General Intuition's mission is described as extremely ambitious, spanning agent research, video understanding, and world modeling as interconnected problems rather than separate tracks [4].

On choosing independence over acquisition

De Witte turned down a reported $500 million offer from OpenAI in order to instead raise $134 million and build General Intuition as an independent AI lab [5]. This decision is presented as a deliberate bet that an independent, games-focused lab could pursue a research agenda that a larger, text-centric organization would not prioritize, underscoring how central the games-as-data thesis is to his strategic thinking rather than being a side interest [5].

On the scale and speed of AI's impact

De Witte has framed the current AI shift as carrying roughly ten times the impact of the industrial revolution, arriving roughly ten times as fast, situating his own work on agents and world models within a broader conviction that this transition will be unusually large and unusually compressed in time [1].

Takeaways

• General Intuition's research agenda treats gameplay video as a primary dataset for building embodied AI agents and world models, distinct from text-centric approaches [3][4].
• De Witte turned down a reported $500 million OpenAI offer and instead raised $134 million to keep building General Intuition independently [5].
• He frames his company's mission as spanning agent research, video understanding, and world modeling together rather than as separate efforts [4].
• He has described the current AI transition as roughly ten times the impact of the industrial revolution and arriving roughly ten times as fast [1].