AGI
→
ASI
Exploring the long-term frontier of increasingly general, autonomous and capable machine intelligence.
What happens when machine intelligence becomes increasingly general?
AGI represents a long-term research direction toward artificial systems capable of learning and solving a broad range of intellectual problems rather than being limited to a narrow task.
ASI refers to a hypothetical future state in which machine intelligence substantially exceeds human capabilities across many important domains.
KryomAI treats this frontier as a research problem spanning intelligence architectures, reasoning, autonomy, learning, infrastructure, evaluation and alignment.
A progression in
machine capability.
Specialized Intelligence
Systems optimized for defined tasks, domains and capabilities.
General Intelligence
Systems capable of transferring knowledge and solving diverse unfamiliar problems.
Artificial General Intelligence
A research frontier involving broadly capable systems operating across many forms of intellectual work.
Superintelligence
A hypothetical frontier where machine intelligence substantially exceeds human capabilities across many domains.
The problems behind
the frontier.
Progress toward increasingly capable intelligence requires advances across multiple interconnected research areas.
General Intelligence
Research into systems capable of learning, reasoning and adapting across a broad range of domains and unfamiliar problems.
Autonomous Systems
Developing architectures that can maintain objectives, plan over longer horizons and operate through continuous feedback.
Machine Reasoning
Exploring inference, abstraction, planning, verification and problem solving as foundations of increasingly capable intelligence.
Self-Improvement
Investigating how intelligent systems may evaluate their capabilities, identify limitations and improve their performance.
Scalable Intelligence
Exploring architectures, infrastructure and learning systems capable of supporting increasingly sophisticated machine intelligence.
AI Safety & Alignment
Researching reliability, controllability, evaluation and alignment as intelligence becomes increasingly capable.
Intelligence is
an interconnected system.
General intelligence is unlikely to emerge from a single capability. It requires systems where perception, memory, reasoning, learning, planning and action work together.
Capability and
responsibility together.
More capable intelligence creates both opportunities and new technical challenges. Reliability, evaluation, controllability and alignment therefore remain important research problems.
KryomAI's long-term direction considers the development of advanced intelligence together with mechanisms for monitoring, evaluation and responsible deployment.
Build increasingly capable intelligence while continuously improving our understanding of how to evaluate, control and align it.
How far can
intelligence go?
KryomAI exists to explore that question through research into increasingly capable, general and autonomous machine intelligence.