Follow up to “Transformative AI, existential risk, and real interest rates” (Chow, Halperin, Mazlish, 2025) – Test and Conclusion

introduces a novel approach for forecasting the timelines for transformative AI by analyzing macro-financial markets—specifically, the behavior of long-term real interest rates.

TEST: apply this framework to current situation and explain what the output tells us about transformative AI timelines.

CONCLUSION: detail in analysis following

If markets were rationally aggregating all known information about technological trajectories, and taking forecasts of AI experts seriously, we would see the major signal—an upward shift in the long-term real interest rate term structure.

The absence of this signal suggests that society-wide expectations about AI remain conservative, cautious, or skeptical in the aggregate, despite high-profile advances in generative AI.

This “outside view” (market prices) currently contradicts the most rapid “inside view” timelines coming from certain AI researchers and forecasters.

Analysis Follows: (Perplexity.ai)

1. Theoretical Mechanism

Aligned AI (very high growth): If investors expect AI-driven explosive growth (e.g., 30%+ global GDP growth per year), they treat future consumption as less valuable (abundance lowers marginal utility), pushing up long-term real rates.
Unaligned AI (existential risk): If investors think AI could cause human extinction, future consumption has close to zero expected value, again pushing up real rates.
Both outcomes result in much “flatter” time preferences: society is incentivized to consume sooner, and to save/invest less for the long term. This pushes up the price for borrowing (the real interest rate).

2. Empirical Analysis

Empirical evidence provided in the paper (across 59 countries, 35 years):

Long-term growth expectations are consistently and significantly associated with higher long-term real interest rates.
The link is robust across panel regressions, differences, and sub-samples (including the G7), with coefficients typically above one.paste.txt
Their data comes predominantly from inflation-indexed government bonds (“TIPS” in the US, equivalents in UK, Canada, Australia) and professional forecaster surveys.

Table: Key Relationships

Macro Variable	Empirical Finding
Long-term GDP growth expectations	Positively correlated with real interest rates
Long-term growth volatility	Slightly negative effect on real rates
Short-term (0–5 yr) GDP growth forecasts	Negative/ambiguous effect
Country risk (CDS spreads)	Positively correlated with real rates

3. Applying the Framework to the Current Situation (2025)

What Does the Data Say Now?

A. Long-Term Real Interest Rates (circa 2025)

As of mid-2025, 10- to 30-year inflation-linked government bond yields in the US, UK, Canada, Australia (and other developed economies) remain historically low, often below 2% real. For example, US 30-year TIPS yield hovered ~2%, UK/Canada/Australia even lower.Transformative-AI-existential-risk-and-real-interest-rates.pdf+1
No pronounced spike or structural jump in long-term real rates—absolutely no sign of the double-digit levels predicted by this framework for transformative AI.

B. Growth Expectations

Professional economic forecasts and market pricing (e.g., implied from equity markets, long-range GDP growth forecasts) do not indicate general expectations of explosive (>30% p.a.) future growth.

C. Market Implications

If investors broadly expected transformative AI (either “growth” or “doomsday”) within relevant investment horizons (10–30 years), the theory and empirical results indicate we should observe substantially higher long-term real rates than we do.
The market is not pricing in either massive upside or existential catastrophic risk from AI in the coming decades.
This finding is robust across major economies with liquid bond markets and credible monetary regimes.

4. Interpretation: Forecast for Transformative AI Timelines

Key Takeaway:

The current level of long-term real interest rates is a strong, market-driven indicator that investors and, by extension, society do not expect transformative AI (either explosive growth or existential risk) to arrive in the next 10–30 years.

If markets were rationally aggregating all known information about technological trajectories, and taking forecasts of AI experts seriously, we would see the major signal—an upward shift in the long-term real interest rate term structure.
The absence of this signal suggests that society-wide expectations about AI remain conservative, cautious, or skeptical in the aggregate, despite high-profile advances in generative AI.
This “outside view” (market prices) currently contradicts the most rapid “inside view” timelines coming from certain AI researchers and forecasters.

5. Qualifications & Caveats

Markets are not omniscient: It is possible that AI timeline risk is underestimated, the topic is too new or unclear for financial markets, or it could be slow to be incorporated into prices (especially for “all or nothing” tail risks).
Bond yields also reflect secular factors: Demographics, central bank policy, liquidity demand, etc.
Private vs Public actors: Some transformative AI development may be done by entities not represented in financial markets.
Nevertheless, this method provides a rigorous baseline and an important, falsifiable “reality check” on AI forecasts grounded not in expert speculation but in global market consensus.

Conclusion

Applying this macro-financial framework to 2025 data outputs a clear signal:
Markets do not foresee transformative AI arriving soon enough to warrant a major repricing of long-term real interest rates. Thus, the “default” market expectation is that explosive, society-changing AI—whether positive or negative—remains more than a generation away. This does not rule out rapid surprises, but it quantifies the current market consensus in a rigorous, transparent way.paste.txt+1

References:

Chow, T., Halperin, B., Mazlish, J. Z. (2025). Transformative AI, existential risk, and real interest rates. (Full text in attached PDF and paste).Transformative-AI-existential-risk-and-real-interest-rates.pdf
Empirical tables and discussion in the same source: Tables 1, 2, 5, and related discussion.paste.txt

Please indicate if you would prefer a tabular or graphical summary of expected relationships, or require an extractable summary for your research notes.

Bankwatch

PRODUCTIVITY Progress in era of Artificial Intelligence