Neither an arrival date nor a path is established. Self-improvement is one proposed scenario, not a guaranteed property; compute, reliability and control remain separate questions.
Drag to rotate · or use the arrow keys
Add levels and explore how the connections become more complex. This is a visual hypothesis about a system beyond human scale.
Conceptual diagram. Its controls do not measure progress or predict dates.
About this eventSam Altman described superintelligence as possible within a few thousand days, not as a dated forecast.
The phrase frames a direction of debate rather than a calendar. The source provides no testable date, definition of ASI or achievement criteria, so the map treats it as an industry view rather than an event or release promise.
Source: Sam Altman · The Intelligence AgeAbout this eventDario Amodei argues human-level is within a few years; Demis Hassabis puts genuine AGI a few years out, not here.
About this eventOne possible scenario is a system accelerating its own development. A sustained improvement loop still needs to be demonstrated; high model scores do not establish it automatically.
About this eventThe higher the capability, the more a goal mismatch costs. Not philosophy but an engineering requirement on the reliability gate.
Primary material behind this dossier: papers, lab publications and official reports.