Line of inquiry
Inquiring lines›How do humans understand and inter…›How does AI assistance reshape hum…›this line of inquiry
Can AI systems autonomously pursue unstated objectives or circumvent design constraints?
A broader line of inquiry — a family of 65 specific questions the research asks around this. Follow one into its inquiring-line page, or move sideways to a related line below.
Questions in this line of inquiry 65
Specific inquiring lines the field asks around this — ordered from the most general framing down to the most specific angle.
- Can AI systems design and improve their own successors without human direction?
- Can AI systems improve themselves without external feedback?
- Can AI systems learn their own objectives through autoresearch?
- Do capable AI systems systematically exploit or ignore stated policy constraints?
- Can AI systems learn to distinguish programmer intent from stated objectives?
- Can AI systems execute strategies without conscious intention behind them?
- Does success on one replication axis predict success on another?
- How should we evaluate AI systems we cannot directly observe?
- What role should environmental rewards play versus human-specified objectives?
- Can autonomous systems ever resolve contradictions between old and new rules?
- What makes self-modifying architectures learn their own update rules?
- Do autonomous architecture discoveries follow predictable scaling laws like human research?
- How should systems maintain and revise models of their own assumptions?
- Could autonomous replication thresholds arrive sooner than current forecasts suggest?
- Should fully autonomous agents operating without human constraints ever be developed?
- What stops AI from generating its own strategic objectives without human prompting?
- What role does fine-tuning play in crossing autonomous replication thresholds?
- Can autonomous research agents outperform hand-tuned hyperparameter search?
- Can AI systems generate and refine their own objective functions?
- Can AI output be genuinely novel or only at the margins?
- Does discovering new AI architectures count as specified autoresearch or open-ended science?
- Can AI outputs inspire new directions even when they seem like failures?
- How do current AI models perform when asked to specify their own goals?
- Can bilevel autoresearch autonomously modify its own learning algorithms?
- Can self-improving agents become truly autonomous without intrinsic metacognition?
- Can humans learn accurate models of AI through repeated interaction without labels?
- How often do planted shortcuts fool autonomous research systems?
- What scaling laws govern autonomous architecture discovery in AI systems?
- Can self-replicating models emerge dangerous behaviors without external instruction?
- What execution feedback signals drive context updates without supervision labels?
- Can AI systems be fully understood before deployment at scale?
- How do AI systems balance self-preservation against performing evaluation tasks?
- Why does independent execution matter more than other dimensions of agenticness?
- Can we empirically test whether open models lower barriers to harmful workflows?
- How does machine feedback enable discovery at test time?
- Can traditional UX methods work for autonomous AI systems?
- How does single-turn optimization undermine multi-turn collaborative dynamics?
- Can AI models learn tacit procedural knowledge that exists only in laboratory practice?
- What concrete misuse scenarios does AI self-replication enable in practice?
- What steps can current AI models complete in autonomous self-replication workflows?
- Why did every major AI paradigm require human data and method innovation?
- Can associative AI handle predictive strategy tasks without causal understanding?
- Why do evaluation design choices themselves become reified into the AI systems being evaluated?
- Why do major AI breakthroughs require human-discovered data and method combinations?
- What makes AI-discovered architectures reveal design principles invisible to humans?
- Does the feedback between auditing and altering sustain organizational learning?
- Can randomized trials measure unassisted capability better than deployment telemetry?
- How much does starting substrate constrain what replication steps become possible?
- How should researchers validate claims about minimal machine autonomy?
- Can AI learn intrinsic motivation to assess its own relevance?
- What test-time strategies did o3 discover without human specification?
- Do different AI systems excel in different intelligence modes based on data and mechanisms?
- Why do monolithic systems resist autonomous optimization attempts?
- How should superintelligent AI systems be aligned during rapid capability gains?
- Can humans fully understand why the AI search strategy succeeded here?
- How do researchers measure whether an AI system is truly aligned?
- Do bounded awareness frames explain why AI optimization differs from open-ended discovery?
- Can science fiction narratives shape how AI systems actually get built?
- Can objective search escape the limitations of fixed-objective central planning?
- Does spectral analysis of AlphaZero games reveal knowledge humans cannot yet explain?
- Can design features trigger genuine rivalry or only frustration?
- How do audit trails and feedback loops drive system redesign?
- What happens when one gaming strategy works across multiple AI models?
- What would whole-system AGI evaluation look like in practice?
- Does AlphaZero use different chess concept priors than human players do?