ai-research-survey

scan.json (19814B)

---

 {
   "paper": {
     "title": "Explosive growth from AI automation: A review of the arguments",
     "authors": ["Ege Erdil", "Tamay Besiroglu"],
     "year": 2023,
     "venue": "arXiv",
     "arxiv_id": "2309.11690"
   },
   "scan_version": 2,
   "active_modules": ["survey_methodology"],
   "methodology_tags": ["theoretical", "meta-analysis"],
   "key_findings": "Standard economic growth models consistently predict explosive growth (>30% annual GWP growth) when AI can substitute for human labor, driven by increasing returns to scale, rapid digital workforce expansion, and massive transitory effects. The authors evaluate nine counterarguments (regulation, bottlenecks, alignment, slow automation, etc.) and find none decisive, estimating roughly even odds of explosive growth this century conditional on AGI. The most credible counterarguments involve regulation and non-accumulable production bottlenecks, but quantitative analysis suggests these are unlikely to fully block growth acceleration.",
   "claims": [
     {
       "claim": "Semi-endogenous growth models predict explosive growth when AI substitutes for human labor, even with ideas getting 'harder to find' (β as low as 0.68).",
       "evidence": "Section 2.1 derives the condition λ/ϕ + β > 1 for hyperbolic growth, using Bloom et al. 2020's estimate of λ/ϕ ≈ 0.32, yielding a threshold of β ≈ 0.68.",
       "supported": "moderate"
     },
     {
       "claim": "Even exogenous growth models without increasing returns predict explosive growth if AI worker costs are below ~$150,000/worker at savings rates ≥ 0.2.",
       "evidence": "Section 2.2 and Appendix D derive the condition c̄ ≤ s^(10/7) × 150,000 $/worker, using current hardware cost estimates and Carlsmith 2020's brain computation estimates.",
       "supported": "moderate"
     },
     {
       "claim": "Partial AI automation (90% of tasks) could increase output by 100x with moderate complementarity (ρ = -1).",
       "evidence": "Section 2.3 derives level effects from a CES production function, showing f^((1-ρ)/ρ) scaling factor; Figure 1 shows the results for various parameter values.",
       "supported": "moderate"
     },
     {
       "claim": "P(explosive growth this century | AGI this century) is about as likely as not (~50%).",
       "evidence": "Section 4 synthesizes all arguments. This is stated as a subjective probability estimate based on the analysis, not derived from a formal model.",
       "supported": "weak"
     },
     {
       "claim": "Regulation is unlikely to block explosive growth indefinitely, analogous to failed attempts to restrict Industrial Revolution technology diffusion.",
       "evidence": "Section 3.1 draws on the historical case of England's protectionist policies (Jeremy 1977) and arguments about declining AI training costs and national security incentives.",
       "supported": "weak"
     }
   ],
   "checklist": {
     "artifacts": {
       "code_released": {
         "applies": true,
         "answer": false,
         "justification": "No code repository or computational artifacts are provided. The paper contains mathematical models that could have been implemented and shared."
       },
       "data_released": {
         "applies": true,
         "answer": false,
         "justification": "No dataset released. The paper uses parameter estimates from cited literature but does not release compiled data."
       },
       "environment_specified": {
         "applies": false,
         "answer": false,
         "justification": "This is a theoretical/review paper with no computational experiments requiring environment specification."
       },
       "reproduction_instructions": {
         "applies": true,
         "answer": false,
         "justification": "While the mathematical derivations are in the appendices, there are no step-by-step reproduction instructions for the quantitative estimates."
       }
     },
     "statistical_methodology": {
       "confidence_intervals_or_error_bars": {
         "applies": false,
         "answer": false,
         "justification": "This is a theoretical review paper that derives analytical results from growth models rather than running experiments with empirical measurements."
       },
       "significance_tests": {
         "applies": false,
         "answer": false,
         "justification": "No empirical experiments are run; the paper derives analytical results from economic growth models."
       },
       "effect_sizes_reported": {
         "applies": false,
         "answer": false,
         "justification": "No experimental comparisons are made. The paper reports model-derived quantities (e.g., growth rate thresholds) rather than empirical effect sizes."
       },
       "sample_size_justified": {
         "applies": false,
         "answer": false,
         "justification": "Theoretical paper with no empirical sample."
       },
       "variance_reported": {
         "applies": false,
         "answer": false,
         "justification": "No experimental runs; results are analytical derivations from models."
       }
     },
     "evaluation_design": {
       "baselines_included": {
         "applies": true,
         "answer": true,
         "justification": "The paper compares multiple growth model types (semi-endogenous, exogenous, CES) and evaluates arguments against each other as a structured review."
       },
       "baselines_contemporary": {
         "applies": true,
         "answer": true,
         "justification": "References are contemporary: Davidson 2021, Bloom et al. 2020, Cotra 2020, Carlsmith 2020, Trammell and Korinek 2020."
       },
       "ablation_study": {
         "applies": false,
         "answer": false,
         "justification": "No system with components to ablate; this is a theoretical review."
       },
       "multiple_metrics": {
         "applies": false,
         "answer": false,
         "justification": "No empirical evaluation with metrics; the paper analyzes theoretical model predictions."
       },
       "human_evaluation": {
         "applies": false,
         "answer": false,
         "justification": "No system outputs to evaluate; this is a theoretical review paper."
       },
       "held_out_test_set": {
         "applies": false,
         "answer": false,
         "justification": "No empirical evaluation requiring train/test splits."
       },
       "per_category_breakdown": {
         "applies": true,
         "answer": true,
         "justification": "The paper provides detailed per-argument analysis with quantitative assessments for each of the 3 pro-arguments and 9 counter-arguments, including parameter sensitivity analysis (Figure 1, Table 3)."
       },
       "failure_cases_discussed": {
         "applies": true,
         "answer": true,
         "justification": "Each section discusses conditions under which the argument fails. E.g., Section 2.3 lists three weaknesses of the transitory effects argument; Section 2.2 discusses when computational costs are too high."
       },
       "negative_results_reported": {
         "applies": true,
         "answer": true,
         "justification": "The paper explicitly discusses scenarios where explosive growth would not occur and assigns substantial probability to that outcome. The conclusion states 'high confidence in explosive growth is unwarranted.'"
       }
     },
     "claims_and_evidence": {
       "abstract_claims_supported": {
         "applies": true,
         "answer": true,
         "justification": "The abstract claims are hedged ('could plausibly lead to') and supported by the detailed analysis in the body. The abstract also notes 'high confidence in this outcome is unwarranted,' which matches the discussion."
       },
       "causal_claims_justified": {
         "applies": true,
         "answer": true,
         "justification": "Causal claims are framed within formal economic models with explicit assumptions. The paper uses conditional language ('if AI can substitute for labor, then models predict...') and derives results analytically from stated premises."
       },
       "generalization_bounded": {
         "applies": true,
         "answer": true,
         "justification": "The paper is explicit about conditioning on 'AI capable of broadly substituting for human labor' and notes uncertainties about regulatory responses, production bottlenecks, and automation pace. Section 4 cautions against overconfidence."
       },
       "alternative_explanations_discussed": {
         "applies": true,
         "answer": true,
         "justification": "The entire structure of the paper is to present arguments and counterarguments. Section 3 presents 9 alternative explanations for why explosive growth might not occur, each with quantitative analysis."
       },
       "proxy_outcome_distinction": {
         "applies": true,
         "answer": true,
         "justification": "The paper explicitly defines 'explosive growth' as annual GWP exceeding 130% of prior maximum, distinguishes GDP measurement from consumer surplus (Section 3.6), and discusses what GDP does and does not capture."
       }
     },
     "setup_transparency": {
       "model_versions_specified": {
         "applies": false,
         "answer": false,
         "justification": "No AI models are used in the paper's methodology; it is a theoretical review."
       },
       "prompts_provided": {
         "applies": false,
         "answer": false,
         "justification": "No prompting is used; this is a theoretical review paper."
       },
       "hyperparameters_reported": {
         "applies": true,
         "answer": true,
         "justification": "Key model parameters are stated: α = 0.7, β, ρ ranges, savings rates, depreciation rates (Tables 2, 3), Bloom et al. estimates (λ/ϕ ≈ 0.32)."
       },
       "scaffolding_described": {
         "applies": false,
         "answer": false,
         "justification": "No agentic scaffolding is used."
       },
       "data_preprocessing_documented": {
         "applies": false,
         "answer": false,
         "justification": "No data collection or preprocessing; the paper uses parameter estimates from existing literature."
       }
     },
     "limitations_and_scope": {
       "limitations_section_present": {
         "applies": true,
         "answer": true,
         "justification": "Section 4 (Discussion) and Section 4.1 (Open questions) serve as a substantive limitations section, discussing multiple sources of uncertainty and unresolved questions."
       },
       "threats_to_validity_specific": {
         "applies": true,
         "answer": true,
         "justification": "Section 4.1 lists 6 specific open questions including uncertainty about brain computation estimates, robotics costs, alignment failures, and the economic value of superhuman intelligence."
       },
       "scope_boundaries_stated": {
         "applies": true,
         "answer": true,
         "justification": "The paper explicitly defines explosive growth quantitatively (Section 1), conditions conclusions on 'near-complete AI automation,' and notes that 'predicting explosive growth requires extrapolating economic models beyond their empirically validated domains.'"
       }
     },
     "data_integrity": {
       "raw_data_available": {
         "applies": false,
         "answer": false,
         "justification": "No empirical data is collected; the paper uses published parameter estimates from existing literature."
       },
       "data_collection_described": {
         "applies": true,
         "answer": true,
         "justification": "The paper cites specific sources for all parameter values used: Bloom et al. 2020 for R&D returns, Carlsmith 2020 for brain computation estimates, FRED for US labor data, etc."
       },
       "recruitment_methods_described": {
         "applies": false,
         "answer": false,
         "justification": "No participants or samples are recruited; this is a theoretical review."
       },
       "data_pipeline_documented": {
         "applies": false,
         "answer": false,
         "justification": "No data pipeline; analytical derivations from published parameter estimates."
       }
     },
     "conflicts_of_interest": {
       "funding_disclosed": {
         "applies": true,
         "answer": true,
         "justification": "The acknowledgments state: 'We are grateful to Open Philanthropy for support for this project.'"
       },
       "affiliations_disclosed": {
         "applies": true,
         "answer": true,
         "justification": "Author affiliations are listed: Ege Erdil (Epoch AI), Tamay Besiroglu (Epoch AI, MIT FutureTech)."
       },
       "funder_independent_of_outcome": {
         "applies": true,
         "answer": false,
         "justification": "Open Philanthropy has publicly advocated for AI safety and has significant interest in AI timelines and impact forecasting, which are directly related to the paper's conclusions about explosive growth probability."
       },
       "financial_interests_declared": {
         "applies": true,
         "answer": false,
         "justification": "No competing interests statement is provided. Epoch AI's mission involves AI forecasting, which could create organizational interest in particular conclusions."
       }
     },
     "contamination": {
       "training_cutoff_stated": {
         "applies": false,
         "answer": false,
         "justification": "No pre-trained model is evaluated on any benchmark."
       },
       "train_test_overlap_discussed": {
         "applies": false,
         "answer": false,
         "justification": "No pre-trained model is evaluated on any benchmark."
       },
       "benchmark_contamination_addressed": {
         "applies": false,
         "answer": false,
         "justification": "No pre-trained model is evaluated on any benchmark."
       }
     },
     "human_studies": {
       "pre_registered": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "irb_or_ethics_approval": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "demographics_reported": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "inclusion_exclusion_criteria": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "randomization_described": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "blinding_described": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       },
       "attrition_reported": {
         "applies": false,
         "answer": false,
         "justification": "No human participants."
       }
     },
     "cost_and_practicality": {
       "inference_cost_reported": {
         "applies": false,
         "answer": false,
         "justification": "Theoretical review paper; no computational method with inference costs."
       },
       "compute_budget_stated": {
         "applies": false,
         "answer": false,
         "justification": "Theoretical review paper; no computational experiments."
       }
     },
     "survey_methodology": {
       "prisma_or_structured_protocol": {
         "applies": true,
         "answer": false,
         "justification": "No systematic review protocol, PRISMA diagram, or structured search strategy is described. The paper appears to select arguments and references ad hoc."
       },
       "quality_assessment_of_sources": {
         "applies": true,
         "answer": true,
         "justification": "The paper critically evaluates the strength of each argument it reviews, using a defined likelihood scale (Appendix A, Table 4) and providing quantitative assessments of each argument's strength."
       },
       "publication_bias_discussed": {
         "applies": true,
         "answer": false,
         "justification": "No discussion of publication bias in the economic growth literature being reviewed."
       }
     }
   },
   "red_flags": [
     {
       "flag": "Subjective probability estimates without formal calibration",
       "detail": "The paper's central conclusion — roughly 50% probability of explosive growth conditional on AGI — is a subjective estimate based on the authors' synthesis of arguments, not derived from a formal probabilistic model. The likelihood scale in Appendix A provides structure but the assignments are still judgment calls."
     },
     {
       "flag": "Funder alignment with conclusions",
       "detail": "Open Philanthropy, which funded the work, has significant programmatic interest in AI timelines and existential risk, areas where the explosive growth thesis has direct implications. Epoch AI also focuses on AI forecasting. Neither conflict is disclosed."
     },
     {
       "flag": "Selective argument framing",
       "detail": "The paper rates all 9 counterarguments as 'unlikely' or 'very unlikely' to block explosive growth individually, which may understate joint probability even though the authors acknowledge correlation between arguments in Section 4."
     }
   ],
   "cited_papers": [
     {
       "title": "Artificial intelligence and economic growth",
       "authors": ["Philippe Aghion"],
       "year": 2018,
       "relevance": "Foundational model of AI automation effects on growth, including Baumol bottleneck effects."
     },
     {
       "title": "Could advanced AI drive explosive economic growth",
       "authors": ["Tom Davidson"],
       "year": 2021,
       "relevance": "Direct predecessor to this paper, estimating 30% probability of explosive growth conditional on advanced AI."
     },
     {
       "title": "Are Ideas Getting Harder to Find?",
       "authors": ["Nicholas Bloom"],
       "year": 2020,
       "doi": "10.1257/aer.20180338",
       "relevance": "Key empirical paper on diminishing returns to R&D, provides parameter estimates central to the growth models."
     },
     {
       "title": "Forecasting TAI with biological anchors",
       "authors": ["Ajeya Cotra"],
       "year": 2020,
       "relevance": "AI timeline forecasting using computational requirements of the human brain as anchors."
     },
     {
       "title": "Economic growth under transformative AI",
       "authors": ["Philip Trammell", "Anton Korinek"],
       "year": 2020,
       "relevance": "Review of economic models exploring growth scenarios from moderate acceleration to explosive growth under AI."
     },
     {
       "title": "Scaling laws for neural language models",
       "authors": ["Jared Kaplan"],
       "year": 2020,
       "arxiv_id": "2001.08361",
       "relevance": "Foundational scaling laws paper relevant to AI capability forecasting and compute requirements."
     },
     {
       "title": "Training compute-optimal large language models",
       "authors": ["Jordan Hoffmann"],
       "year": 2022,
       "arxiv_id": "2203.15556",
       "relevance": "Compute-optimal training scaling laws (Chinchilla) relevant to AI development cost projections."
     },
     {
       "title": "Survey of hallucination in natural language generation",
       "authors": ["Ziwei Ji"],
       "year": 2023,
       "relevance": "Cited in the alignment difficulties section regarding LLM hallucination as a deployment barrier."
     },
     {
       "title": "How Much Computational Power Does It Take to Match the Human Brain?",
       "authors": ["Joseph Carlsmith"],
       "year": 2020,
       "relevance": "Provides key estimates of brain computation costs used to calibrate the digital workforce growth model."
     },
     {
       "title": "Algorithmic progress in computer vision",
       "authors": ["Ege Erdil", "Tamay Besiroglu"],
       "year": 2022,
       "arxiv_id": "2212.05153",
       "relevance": "Estimates rate of algorithmic efficiency improvements, relevant to AI capability forecasting."
     }
   ]
 }