Mars Ind is a science and technology company. Our areas of study include games, developer tools, and cognitive science, including AI and SI Replicants, for game developers, engineers, and scientists. We create worlds to explore, instruments to build with, and systems that help us understand perception, learning, and the mind.
- DISCIPLINESGames • Developer Tools • Cognitive Science
- BUILT FORGame Developers • Engineers • Scientists
- PRINCIPLEAugment people. Never replace them.
01 / Manifesto
Progress should make us more human.
We believe technology earns its place by expanding what people can imagine, understand, and create. Games are worlds worth crafting. Developer tools should remove friction, not remove the creator. Intelligence research should deepen our understanding of life, not diminish its value.
AI exists to augment humans, never to replace them. Human judgment, authorship, and responsibility remain at the center of our work.
Our synthetic intelligence research is directed toward cognitive science: studying learning and neurological function, with the long-term ambition of contributing to research on neurological diseases such as Alzheimer's. This is a research purpose, not a claim of a treatment or cure.
Build meaningful worlds. Extend human capability. Pursue understanding with care.
02 / Charter
The commitments behind the work.
- Keep people in control. Design AI to support human skill and decision-making, never to make people disposable.
- Serve creators and investigators. Build original games and useful, dependable tools for game developers, engineers, and scientists.
- Research for human benefit. Direct SI toward understanding cognition, learning, and neurological function, with the aim of supporting disease research.
- Distinguish ambition from evidence. State limitations clearly, test our claims, and separate research goals from demonstrated capabilities.
- Respect agency and consent. Treat people's data, voices, creative work, and participation with care.
- Build for understanding. Favor clear interfaces, reproducible methods, and tools that strengthen the user's own knowledge.
03 / Replicants
Replicating human perception.
Mars Ind uses Replicants as the umbrella name for both AI and SI. We call them Replicants because they are designed to replicate human perception.
- Micro Replicants
- Lightweight, domain-specific systems designed to run directly on shared-hosting CPUs or phone CPUs/GPUs.
- Mini Replicants
- More general-purpose systems designed for local use on PCs.
- Replicants
- Full-size systems designed for dedicated AI boxes.
Size describes scope and target hardware; AI and SI describe how a Replicant learns. Replicants is both the umbrella name and the name used for full-size systems.
04 / Intelligence framework
Two approaches to learning.
In Mars Ind's terminology, AI and SI describe different learning approaches. General and superintelligence terms describe learning capacity, not performance on a single task.
Artificial intelligence
Our AI Replicants use external learning from large datasets. Neural-network weights are updated periodically in training sessions, rather than continuously in real time during use.
- AI Artificial Intelligence
- Neural networks capable of complex tasks, with less-than-human learning capacity.
- AGI Artificial General Intelligence
- Human-level learning capacity within the artificial intelligence family.
- ASI Artificial Superintelligence
- Greater-than-human learning capacity within the artificial intelligence family.
Synthetic intelligence
Our SI system already learns in real time, without training datasets or pretrained knowledge. SI Replicants use internal, continuous learning through experience and are designed to simulate real neurological functions.
- SI Synthetic Intelligence
- Real-time learning designed to simulate neurological functions.
- SGI Synthetic General Intelligence
- Human-level, real-time learning capacity within the synthetic intelligence family.
- SSI Synthetic Superintelligence
- Greater-than-human, real-time learning capacity within the synthetic intelligence family.
These are Mars Ind's definitions, not a universal scientific taxonomy. AI also includes methods beyond neural networks, and some AI systems use online or continual learning. The general and superintelligence labels describe learning levels, not a claim that those levels have been achieved. Simulating neurological functions does not by itself establish consciousness or medical effectiveness.
Our working method: human-led collaboration
COLLABORATION PROTOCOL
Across every division, one principle holds: the human stays in the loop. AI and SI extend judgment, craft, and imagination; they do not hide the work behind a magic button.
01 // AGENCY
Human direction remains primary.
You set intent, constraints, and standards. The system accelerates the work without taking ownership away from its operator.
Every output is a new input.
Human edits and synthetic transformations move in a deliberate loop until the result earns approval.
Tools amplify understanding.
Code or media produced without domain knowledge is still guesswork, whether it came from a person or a model. Informed operators build reliable systems.
HUMAN Direct + Edit
⟶
AI / SI Generate + Transform
⟶
HUMAN Judge + Refine
IMAGINGCanvas control + img2img iteration
VOICETimeline craft + synthetic performance
REPLICANTSHuman intent + synthetic cognition
DEVELOPER TOOLSDomain control + intelligent automation
No human degradation. No synthetic mystique. Better work through collaboration.
