AGI Timelines: When Could Advanced AI Arrive?¶

Purpose: Understand expert predictions on AGI timelines and what they mean for Australian planning Audience: Government, business and community leaders | Time: 15-20 minutes

"When will we get AGI?" is one of the most common questions about artificial general intelligence—and one of the hardest to answer. Public estimates range from the late 2020s to many decades away, but they often describe different capability thresholds and use different kinds of evidence. This page separates institutional positions, individual forecasts and research surveys, then explains what the uncertainty means for Australian planning.

What the experts say¶

Frontier AI developers: Institutional positions and executive forecasts¶

Statements from frontier developers are useful because they reveal what those organisations are preparing for. They should not be treated like independent forecasts: definitions differ, executives may be expressing personal judgement and developers have commercial and strategic incentives.

The defensible conclusion is that several leading developers are actively preparing for much more capable systems. Their statements are relevant short-horizon scenarios, not consensus forecasts.

Australian and Australia-connected perspectives¶

Australian perspectives span the same disagreement. Toby Walsh (UNSW) has used 2062 as a longer planning horizon while arguing for sustained policy preparation rather than confidence in a precise arrival date. Helen Toner, who was born in Melbourne and later served on OpenAI's board, frames the disagreement through three unresolved questions about scaling, AI-assisted research and autonomous agents; those questions are used below to explain why forecasts diverge.

Epoch AI reviews published forecasting literature and maintains empirical datasets on model capabilities. METR measures the length of software tasks that frontier models can complete autonomously; its observed trend is a capability indicator, not a direct AGI arrival forecast.

Broader researcher surveys¶

The 2023 survey of 2,778 AI researchers reported a 50% aggregate probability of unaided machines outperforming humans in every possible task by 2047. The same survey produced later estimates for full automation of occupations. These results are highly sensitive to the definition and question wording.

AGI Timeline Predictions Compared¶

Leopold Aschenbrenner, a former OpenAI researcher, argues for AGI around 2027 in his Situational Awareness scenario analysis.

Key insight: Near-term developer claims and broader researcher surveys are not directly comparable. For planning, treat the late 2020s as a high-impact stress-test scenario while building capabilities that remain useful across longer horizons.

Why do AGI timeline predictions vary so much?¶

Helen Toner (Australian AI policy researcher, former OpenAI board member, Director of Strategy at Georgetown CSET) identifies three fundamental unresolved questions that determine whether—and how soon—transformative AI will arrive:

Beyond these three questions, five additional factors drive dramatic variation:

Definition ambiguity: "AGI" ranges from benchmark performance (arguably achieved) to economically transformative (automating most jobs) to true general intelligence (human-level across all domains) to superintelligence. Frontier labs use looser definitions; academics stricter ones.

Scaling vs. breakthroughs: Frontier labs believe current approaches will scale to AGI (more compute + data + algorithms). Academic sceptics argue fundamental breakthroughs are needed for robust reasoning, understanding and transfer learning. Both could be right—scaled systems might automate jobs without solving all hard AI problems.

Information and incentives: Labs have privileged access to capabilities but incentives to appear cutting-edge. External researchers lack internal data but have less financial stake. Both see different parts: labs see rapid gains, researchers see persistent limitations.

Compute trajectory uncertainty: Short forecasts often assume continued rapid growth in training compute, investment and algorithmic efficiency. Physical limits such as chips and power, economic limits such as diminishing returns or regulatory constraints could extend timelines.

Recursive improvement wildcard: If AI systems meaningfully accelerate AI research, progress could compress dramatically. If human oversight remains necessary, progress stays incremental. This is the widest uncertainty—recursive improvement could compress decades into years.

What do AGI timelines mean for Australian planning?¶

Planning under uncertainty: We cannot know the exact timeline, but we can prepare across scenarios. A practical approach is to build capabilities that work across multiple futures rather than betting on one forecast.

Planning question: Are we prepared if advanced capabilities arrive sooner than expected? Australian organisations should allow for the time needed to build evaluation expertise, governance arrangements and continuity plans.

What leading indicators should Australia watch?¶

Rather than relying on predictions, monitor these concrete signals that timelines are shortening or lengthening:

SafeAI-Aus will track these indicators and update guidance as evidence accumulates. Subscribe to our newsletter for updates on significant developments.

Common questions¶

Back to Preparing for AGI