Hideo Kojima’s Wild Vision for AI and Zero‑Gravity Games

Hideo Kojima, best known for Metal Gear Solid and Death Stranding, has never been shy about chasing ideas that sound impossible when he first talks about them. In a recent interview with Japanese outlet Nikkei Xtrend, reported internationally by multiple gaming publications, he described two particularly audacious goals for future projects: a game designed to delight and train an AI, and another to be played in genuine weightlessness rather than simulated zero gravity on a regular screen.

These ambitions go beyond flashy marketing. They touch on deep questions about how games might evolve as artificial intelligence advances and as human activity gradually extends further into space. To understand the implications, it helps to look at where game technology stands today, how Kojima has historically treated new tech, and what it would actually take to make these ideas work.

Kojima’s Longstanding Obsession With the Future

Kojima’s comments about AI-focused and weightless games are consistent with a career spent treating video games as experiments in where technology and culture might go next. His previous projects have repeatedly anticipated or reflected future trends:

• Metal Gear Solid 2 explored information control, digital manipulation, and post-truth themes years before social media became pervasive.
• Metal Gear Solid 4 imagined a world of privatized militaries and tightly networked weapons systems controlled by data infrastructure, echoing later debates around autonomous weapons and surveillance.
• Death Stranding built an asynchronous online system where players indirectly collaborate by leaving structures and paths for others, mirroring the growing interest in social, persistent, and community-driven worlds.

His studio’s technical talks at the Game Developers Conference (GDC) have shown how far Kojima Productions is willing to push game AI and tools to realize particular design goals. For example, a 2021 AI postmortem of Death Stranding detailed how they had to reinvent navigation systems so enemies could traverse complex terrain with the same freedom as the player, using dynamic navigation meshes and custom tooling for environment costs.¹

Seen in that context, the idea of building a game that serves as a kind of curriculum for an AI, or designing a title intended to be played physically in microgravity, is not a random detour; it’s the next step in a long-standing pattern of using games as laboratories for future technologies.

A Game that “Delights” an AI: What Would That Mean?

When Kojima talks about a game that can delight an AI, he is not referring to the simple non-player character routines that already populate games. He’s talking about the possibility of an AI agent playing the game itself, learning from it, and perhaps even being the game’s primary audience.

From Scripted Behavior to Learning Systems

Traditional game AI mostly consists of hand-authored behavior trees, state machines, and pathfinding. These systems are reliable and optimized, but they do not actually learn. In the broader AI field, however, reinforcement learning and self-play have produced systems that can master complex games by repeatedly playing them and improving over time. Research examples include:

• AlphaGo and AlphaZero, which learned to play Go and chess at superhuman levels via self-play and deep reinforcement learning.²
• OpenAI’s Dota 2 bots, which trained in a high-speed simulation to coordinate in a large, complex multiplayer environment.³

These systems treat games as training grounds. Kojima’s suggestion is essentially to invert that relationship: design a commercial game around the needs of a learning AI, while still giving human players something interesting to experience.

Designing for AI as a Primary Audience

A game meant to train and entertain an AI would have to satisfy two overlapping needs:

• Rich, structured data for the AI: consistent rules, clear signals of success or failure, and a way to encode game states compactly.
• Meaningful play for humans: engaging systems, interesting narrative context, and feedback that feels emotionally satisfying rather than purely mathematical.

Practically, such a game might feature:

• Multiple abstraction layers of the same world, where humans see rich audiovisual detail and storytelling while the AI receives a simplified but precise representation of the same environment.
• Shared learning progression, where AI improvements over time genuinely change how the game behaves for human players, leading to evolving difficulty or new emergent behaviors.
• Co-training structures, where humans might guide, correct, or collaborate with the AI, turning play sessions into a kind of interactive annotation and training process.

This dovetails with current research interest in human-AI collaboration and interactive learning, where human feedback is integrated into AI training to make systems more aligned and efficient.⁴

Why Kojima Emphasizes Training, Not AI-Generated Art

In coverage of his interview, outlets have highlighted that Kojima appears less interested in AI replacing artists and more interested in AI as an underlying system: a way to build more responsive worlds, smarter enemies, or adaptive narratives. That mindset aligns with many developers who see generative models as tools for systemic complexity rather than as a shortcut to content production.

Some potential benefits of that approach include:

• More believable behavior: Enemies or allies that adapt to each player’s style, rather than relying on static patterns.
• Dynamic scenarios: Missions or events that are recombined based on player history, creating fresher experiences across playthroughs.
• Long-term evolution: Games that feel different years after launch because underlying AI models have continued to learn from aggregate player data.

Of course, this also raises ethical questions about data collection, consent, and what it means for a commercial entertainment product to double as an AI training pipeline. Those concerns are now part of broader regulatory discussions around AI in many regions, including the European Union’s AI Act, which seeks to manage risks and transparency in high-impact AI deployments.⁵

Imagining a Game Built for Weightlessness

The second idea Kojima floated is, if anything, even more physically ambitious: a game designed to be played in weightlessness. This isn’t just setting a game in space visually; it suggests a game meant for people actually experiencing microgravity, such as astronauts aboard the International Space Station (ISS) or participants in parabolic “zero-G” flights.

How Microgravity Changes Human Movement

In true microgravity, the basic assumptions of everyday motion vanish. There is no up or down, no friction with the ground, and any push or twist can send your whole body spinning. Astronaut training materials from NASA emphasize that movement becomes a matter of subtle pushes, careful orientation, and learning to think in full 3D.⁶

Designing a game around this reality would mean:

• Gestural controls tied to whole-body orientation, perhaps using tethered controllers, gloves, or hand tracking.
• Environmental interaction where players move by pushing off surfaces or gripping handholds, making the physical structure of a space a key part of gameplay.
• Motion sickness mitigation by aligning visual cues with gentle, predictable physical movements, since mismatch between visual motion and inner-ear signals can increase nausea.

Instead of simulating weightlessness on a flat screen for someone sitting or standing on Earth, a Kojima weightless game might exist as a hybrid: a normal version for terrestrial players, and a special mode or variant tailored for people in actual microgravity, perhaps as part of research, tourism, or long-duration missions.

From Space-Themed Games to Space-Played Games

Plenty of games have tackled space settings with simulated zero gravity, from realistic spaceflight simulators to sci-fi shooters featuring free-floating combat. But these are built around the constraints of conventional play spaces. A title engineered for real weightlessness would flip that assumption and treat the environment of a space station or spacecraft as a controller.

Some potential design angles might include:

• Spatial puzzles where players must maneuver themselves and in-game objects through a real physical space, aligning their bodies with virtual overlays.
• Cooperative tasks requiring multiple players to coordinate movement and orientation, echoing the teamwork needed in actual space operations.
• Physiology-aware pacing that respects the realities of working hours, fatigue, and the limited space aboard orbital platforms.

If such a project were ever realized, it could have surprising spillover effects: insights into human-computer interaction in microgravity that benefit astronaut training, space tourism experiences, or experiments on cognition during long missions.

AI Systems, Space, and Player Experience: A Quick Comparison

The two concepts Kojima described may sound separate, but they both revolve around redesigning games for nontraditional participants and places. The table below summarizes how they differ and overlap.

Why These Ideas Matter Beyond One Studio

Even if Kojima never ships a literal AI-delighting game or a title meant for orbital play sessions, simply articulating these goals helps push the medium forward. Historically, ambitious visions in gaming have often catalyzed new technology or influenced other developers, even when the first attempts were rough or commercially risky.

Several broader trends make his ideas especially timely:

• Rapid progress in off-the-shelf AI tools lowers the barrier for studios to experiment with adaptive systems and player-tailored content.
• Increasing commercialization of spaceflight, including private space stations and tourism concepts, creates more contexts where entertainment and training in microgravity will be relevant.
• Blurring lines between games and simulations mean that techniques developed for a Kojima-style experiment could easily carry over into education, professional training, or scientific visualization.

For players, the practical outcome might be more subtle but no less important: games that feel less static, that grow with them, and that explore physical experiences—like truly three-dimensional movement—that are difficult to access any other way.

Frequently Asked Questions

Is Kojima actually making a game just for AI to play?

Based on public reporting, Kojima has expressed a desire to create a game that can delight and train an AI, but there is no official announcement of such a project as a product currently in development. His comments are best understood as a statement of intent and curiosity about the future of game AI rather than a formal reveal.

How is this different from AI opponents already in games?

Most existing game AI follows predefined rules and does not learn significantly once shipped. The kind of project Kojima describes would treat the game as a continuous training environment, allowing AI agents to improve and change behaviors over time, potentially in response to how players actually interact with the game.

Could a weightless game be played on Earth?

Yes. While Kojima’s idea points toward a title designed for real microgravity, many of its mechanics could be adapted for VR or motion-controlled systems on Earth that simulate aspects of weightless movement. However, the full experience he imagines would rely on players literally floating.

What are the risks of using games to train AI?

Key concerns include responsible handling of player data, transparency about how gameplay is used in training, and safeguards against AI systems learning harmful strategies or biases. Regulatory bodies and policymakers are increasingly focused on such issues, especially for AI systems that affect safety-critical domains, but games sit in a relatively low-risk category compared to applications like healthcare or finance.⁵

Why is Kojima often associated with experimental ideas?

Throughout his career, Kojima has used games to explore emerging technologies and social issues, from cinematic storytelling on early consoles to networked social systems in Death Stranding. His willingness to publicly speculate about ambitious concepts has made him a prominent figure whenever discussions turn to the future of interactive media.