REVIEW 3 major objections 6 minor 1 cited by
Shaping AI's Impact on Billions of Lives
T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash
Pith's one-line read This paper argues that AI practitioners should deliberately steer AI toward the common good, and that a blueprint of 18 concrete milestones funded by inducement prizes and research centers can still maximize AI's benefits and limit its…
desk verdict A high-profile agenda-setting blueprint, not a research result; the 18 milestones are mostly sketches, and the paper says so itself. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The central mechanism is the distinction between elastic and inelastic demand for labor, applied to AI-driven productivity gains. When demand is elastic, a drop in price from productivity gains causes a large increase in quantity, so employment grows (programmers, pilots); when inelastic, productivity gains shed jobs (farming). The paper pairs this with the principle of human-AI augmentation rather than replacement, which both improves productivity and keeps humans in the loop as safeguards. The supporting machinery is the 18-milestone blueprint, each tied to an inducement prize of at least $1 million and optionally to a three-to-five-year research center, intended to direct AI research toward goals like a worldwide tutor, a healthcare aide, a disinformation detective agency, and an AI scientist's aide.
What would settle it
Track total employment in U.S. K-12 education and healthcare while AI aides (a teacher's aide, a healthcare aide) are deployed at scale; if productivity rises but employment falls or hours shrink in these sectors over a multi-year window, the paper's elasticity assumption is contradicted. Alternatively, estimate the price elasticity of demand for healthcare and education from historical cost changes; an elasticity below 1 would falsify the paper's job-creation logic.
Extended reading notes
Core claim
The central claim is that society still has a choice about AI's trajectory, and the AI practitioner community can and should deliberately work for the public good through a new innovation infrastructure. The core discovery is a framework: five recurring guidelines from expert interviews, applied to six domains (employment, education, healthcare, information and social networking, media and entertainment, governance and security, and science), yield 18 concrete milestones. The load-bearing economic mechanism is demand elasticity: productivity-enhancing AI increases employment when the demand for the good or service is elastic (as with programming and air travel) and decreases it when inelastic (as with agriculture). The paper asserts that education and healthcare are elastic, so AI aides that make teachers and clinicians more productive would create more jobs while also improving quality and access. It proposes to fund the milestones through $1M+ inducement prizes and short-horizon multidisciplinary research centers, with a public-private partnership to coordinate.
Load-bearing premise
The paper assumes that demand for education and healthcare is elastic, so that AI-driven productivity gains will increase rather than decrease employment in those fields; if demand is actually inelastic, its central recommendation to focus AI on these sectors could cause job losses.
Editorial extensions
If this is right
- AI tools that remove bureaucratic drudgery for teachers and nurses would be adopted first, reducing burnout and making future AI use more likely.
- Aiming AI at elastic fields like education and healthcare could turn today's labor shortages into employment growth, contrary to public fears of mass replacement.
- Inducement prizes of $1M+ per milestone could catalyze practical AI systems (tutor for every child, rapid upskilling, narrow and broad medical AI) within a few years.
- A coordinated public-private partnership modeled on the semiconductor and automobile precedents could make AI development safer and more beneficial.
- Rigorous evaluation (randomized controlled trials, natural experiments, post-deployment monitoring) would become the standard for high-stakes AI deployments, while lower-risk tools could be assessed by the marketplace.
Reading between the lines
- The elasticity argument may extend beyond education and healthcare to other shortage sectors like elder care and skilled trades, but the paper does not analyze those; testing those domains would probe the generality of the framework.
- If the elasticity assumption fails for healthcare (for instance, if cost reductions lead to consolidation rather than expanded care), the employment gains the paper predicts could reverse, and the same AI aides might accelerate job losses.
- The prize-and-center model itself could be tested against a control set: compare milestone completion rates for goals with prizes versus comparable goals without, though the paper does not propose such a test.
- The paper's claim that AI can reduce misdiagnosis could be evaluated with a prospective randomized trial comparing AI-assisted clinicians to standard care, which would settle whether the human-AI team actually improves patient outcomes.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This position paper argues that the AI community should consciously steer AI development toward the common good, rather than choosing between laissez-faire and heavy-handed regulation. It proposes five pragmatic guidelines—human-AI collaboration, targeting productivity gains in elastic-demand fields, removing drudgery from current jobs, adapting to geographic differences, and measuring outcomes—and illustrates them across six domains: employment, education, healthcare, information/news/social media, media/entertainment, governance/national security, and science. The paper then presents 18 'concrete milestones,' recommends funding them through inducement prizes and ad hoc research centers financed by philanthropy (with the Laude Institute as an example), and gives one fully specified prize example in Appendix II (Rapid Upskilling). The evidence base is qualitative: expert interviews, selected historical examples (programmers, pilots, agriculture), micro-economic productivity studies, and illustrative AI successes in science. The paper contains no new derivations, models, or datasets.
Significance. If taken as a policy and research agenda, the paper is significant because of the seniority and expertise of its authors, its attempt to synthesize expert opinion from a wide range of stakeholders, and its concrete call for inducement prizes and research centers. A notable strength is the fully specified Rapid Upskilling prize in Appendix II, which includes measurable criteria (income gain threshold, completion rates, documentation, scalability) that an independent evaluator could verify. The paper is also transparent about several limitations, notably its admission that the 18 milestones are one-paragraph sketches without detailed winning criteria. However, the central claim of providing '18 concrete milestones to guide AI research' is only partially supported, and the employment argument rests on an unquantified elasticity assumption for education and healthcare. These issues are load-bearing for the paper's recommendations, though they are fixable within the scope of a revision.
major comments (3)
- [Section III and Appendix II] The paper repeatedly claims '18 concrete milestones' (Abstract, Introduction, Conclusion), but Section III states that 'the one-paragraph sketches of the proposed 18 milestones above are without much detail on what it would take to win the corresponding inducement prize.' Only Appendix II (Rapid Upskilling) provides measurable, verifiable success criteria. Most milestones—e.g., Worldwide Tutor, Broad Medical AI, Controllable AI for Curating Information Consumption—lack a quantitative metric, a deadline, or an evaluation protocol. Since the inducement-prize mechanism requires independently checkable targets, the blueprint's advertised capacity to 'guide AI research' is operationalized for only 1 of 18 milestones. This mismatch between the claim of concreteness and the evidence is a central weakness that should be fixed by either specifying measurable criteria for more milestones or explicitly reframing the 18 as directional goals rather than concrete milestones.
- [Education and Healthcare sections] The employment logic of Guideline 2—'aim for productivity improvements in fields that would create more jobs'—depends on the assertion that education and healthcare are elastic. The paper states 'we believe that education is elastic' in the Education section and 'we believe that healthcare is also elastic' in the Healthcare section, but offers no quantitative evidence or modeling of demand elasticity for these sectors. The cited example of programmers and pilots is historical and selected, and the paper does not show that the demand for education or healthcare in the relevant markets is sufficiently elastic to translate productivity gains into employment growth. If these sectors are actually inelastic, the recommended focus on productivity could reduce employment, directly undermining the paper's central employment claim. This is a load-bearing point that needs empirical grounding or a substantially qualified recommendation.
- [Employment section] The paper builds its macroeconomic optimism on microeconomic productivity studies (Minnesota attorneys, Harvard consultants, MIT writing tasks, Microsoft programming) and then acknowledges that 'microeconomic studies do not always lead to macroeconomic results, but early indicators are promising.' The leap from task-level speedups to net employment gains is not supported by the evidence presented; no analysis is given of offsetting job displacement, income effects, or general-equilibrium dynamics. Since the paper's first guideline and its employment recommendations depend on this inference, the argument would benefit from a more cautious statement of what can be concluded from the cited studies, or from additional evidence on whether productivity gains in such occupations have historically expanded or contracted employment.
minor comments (6)
- [Footnote 37] The text 'research centers wth five-year sunset clauses' contains a typo ('wth' should be 'with').
- [Table 1] Table 1 is very difficult to read in the text version: column headers and rows are run together, and the country names are not clearly aligned with the data columns. A formatted table with separate columns for each country would improve clarity.
- [Healthcare section] The sentence 'Though the path is long between the current state of medical AI and the world we envision, the rapid pace of progress in developing the underlying technology is cause for optimism' appears twice in the Healthcare section; one occurrence should be removed.
- [Media/Entertainment section] The image caption 'Entertainment before movies. Live theater on Broadway is still healthy alongside cinema.' is disconnected from the figure it describes; if the image is not essential, the caption should be removed or the figure should be placed with its reference.
- [Conclusion] In the sentence 'We also propose 18 targets to show how to deliver on those efforts while thinking carefully about dissemination strategy and, as we shall see, funding,' the phrase 'as we shall see' is out of place in a conclusion; the funding discussion is already in the paper, so the phrase should be revised or deleted.
- [Footnote 6] The referee note in footnote 6 ('A reviewer asked if programmer productivity reduced salaries...') is an unusual appendage for a journal paper; it should be integrated into the main text if it is meant to address a substantive question, or removed.
Circularity Check
No significant circularity: the paper is a policy/vision essay whose claims rest on external literature and expert consultation, not on derived predictions or fitted inputs.
full rationale
This paper does not contain a derivation chain that could be circular. It proposes normative milestones and policy recommendations, with arguments supported by external empirical studies (Brynjolfsson, Autor, National Academies, randomized trials, etc.) and interviews with named experts. No quantity is fitted from data and then relabeled as a prediction; no milestone or guideline is defined in terms of its own conclusion. The main same-author citations are in Appendix I, where AI energy shares are attributed to Patterson et al. 2022 and 2024; those are empirical measurements of Google data-center and smartphone energy use, used as supporting evidence rather than as a premise that entails the paper's recommended blueprint. The citation to Patterson's 'How to Build a Bad Research Center' similarly supports an ancillary funding recommendation and is not load-bearing for the central claim. The paper itself admits in Section III that 'The one-paragraph sketches of the proposed 18 milestones above are without much detail on what it would take to win the corresponding inducement prize'; this is a limitation on operational specificity, not circularity. There is no self-definitional step, no fitted input called a prediction, and no uniqueness argument imported from the authors' prior work. The blueprint's utility may be questioned on concreteness or evidence grounds, but those are correctness/completeness concerns, not circularity concerns.
Assumptions & free parameters
assumptions (5)
- domain assumption AI progress will continue or speed up, and practitioners can still shape outcomes.
- domain assumption Human-AI teams outperform either alone and yield larger productivity gains than automation-focused AI.
- domain assumption Productivity gains in fields with elastic demand increase employment.
- domain assumption Education and healthcare have elastic demand.
- domain assumption Inducement prizes are effective instruments for steering research.
invented entities (1)
-
Laude Institute
Cite this review
Pith. "Pith review of Shaping AI's Impact on Billions of Lives." pith.science (2026). https://pith.science/paper/DYKEBYIT
@misc{pith2026241202730,
author = {Pith},
title = {Pith review of: Shaping AI's Impact on Billions of Lives},
year = {2026},
howpublished = {\url{https://pith.science/paper/DYKEBYIT}},
note = {Machine review of arXiv:2412.02730}
}
read the original abstract
Artificial Intelligence (AI), like any transformative technology, has the potential to be a double-edged sword, leading either toward significant advancements or detrimental outcomes for society as a whole. As is often the case when it comes to widely-used technologies in market economies (e.g., cars and semiconductor chips), commercial interest tends to be the predominant guiding factor. The AI community is at risk of becoming polarized to either take a laissez-faire attitude toward AI development, or to call for government overregulation. Between these two poles we argue for the community of AI practitioners to consciously and proactively work for the common good. This paper offers a blueprint for a new type of innovation infrastructure including 18 concrete milestones to guide AI research in that direction. Our view is that we are still in the early days of practical AI, and focused efforts by practitioners, policymakers, and other stakeholders can still maximize the upsides of AI and minimize its downsides. We talked to luminaries such as recent Nobelist John Jumper on science, President Barack Obama on governance, former UN Ambassador and former National Security Advisor Susan Rice on security, philanthropist Eric Schmidt on several topics, and science fiction novelist Neal Stephenson on entertainment. This ongoing dialogue and collaborative effort has produced a comprehensive, realistic view of what the actual impact of AI could be, from a diverse assembly of thinkers with deep understanding of this technology and these domains. From these exchanges, five recurring guidelines emerged, which form the cornerstone of a framework for beginning to harness AI in service of the public good. They not only guide our efforts in discovery but also shape our approach to deploying this transformative technology responsibly and ethically.
Forward citations
Cited by 1 Pith paper
-
Confucius3-Math: A Lightweight High-Performance Reasoning LLM for Chinese K-12 Mathematics Learning
A 14B RL-trained model reports 96.2% on an internal Chinese K-12 benchmark, about 15x lower serving cost than DeepSeek-R1, with three new training tricks.
Reference graph
Works this paper leans on
-
[5]
Brynjolfsson, Erik, Alex Pentland, Nathaniel Persily, Condoleezza Rice, Angela Aristidou. The Digitalist Papers: Artificial Intelligence and Democracy in America , Stanford Digital Policy Lab , 2024 . Choi, Jonathan, Amy Monahan, and Daniel Schwarcz Lawyering in the Age of Artificial Intelligence , 9 Nov
work page 2024
-
[8]
The Latest Labor Shortage Trends & Statistics (2024) ,
Duarte, Fabio. “ The Latest Labor Shortage Trends & Statistics (2024) ,” Exploding Topics , 10 June
work page 2024
-
[10]
Now, Later, and Lasting: 10 Priorities for AI Research, Policy, and Practice
Horvitz, Eric, Vincent Conitzer, Sheila McIlraith, and Peter Stone. “ Now, Later, and Lasting: 10 Priorities for AI Research, Policy, and Practice .” Communications of the ACM 67, no. 6 (2024): 39-40. Howarth, Josh. “ How Many People Own Smartphones? (2024-2029) .” Exploding Topics , 13 June
work page 2024
-
[11]
Rehospitalizations among patients in the Medicare fee-for-service program
Jencks, Stephen, Mark Williams, and Eric Coleman. “ Rehospitalizations among patients in the Medicare fee-for-service program .” New England Journal of Medicine 360, no. 14 (2009): 1418-1428. Just Energy. Crypto Energy Consumption and Crypto Energy Explained, February 21,
work page 2009
-
[12]
Tailoring recommendation algorithms to ideal preferences makes users better off
Khambatta, Poruz, et al . " Tailoring recommendation algorithms to ideal preferences makes users better off. " Scientific Reports 13, no. 1 (2023):
work page 2023
-
[14]
Littman, Michael, Ifeoma Ajunwa, Guy Berger, Craig Boutilier, Morgan Currie, Finale Doshi-Velez, et al. “ Gathering strength, gathering storms: The one hundred year study on artificial intelligence (AI100) 2021 study panel report .” arXiv preprint arXiv:2210.15767 (2022). Loewen, Peter, et al . “ Global Public Opinion Poll on Artificial Intelligence ,”
arXiv 2022
-
[16]
Scaling deep learning for materials discovery
Merchant, Amil, Simon Batzner, Samuel Schoenholz, Muratahan Aykol, Gowoon Cheon, and Ekin Cubuk. “ Scaling deep learning for materials discovery .” Nature 624, no. 7990 (2023): 80-85. Mitchell, Suzanne, Paula Gardiner, Ekaterina Sadikova et al . Patient Activation and 30-Day Post-Discharge Hospital Utilization . Journal of General Internal Medicine ,
work page 2023
-
[18]
Solving medicine’s data bottleneck: Nightingale Open Science
Mullainathan, Sendhil, and Ziad Obermeyer. “ Solving medicine’s data bottleneck: Nightingale Open Science .” Nature Medicine , vol. 28, 2022, pp. 897–899. National Academies of Sciences, Engineering, and Medicine
work page 2022
Show all 24 references
-
[19]
Global prediction of extreme floods in ungauged watersheds
Artificial Intelligence and the Future of Work . Washington, DC: The National Academies Press. Nearing, Grey, Deborah Cohen, Vusumuzi Dube, et al . “ Global prediction of extreme floods in ungauged watersheds .” Nature 627, no. 8004 (2024): 559-563. Ng, Andrew. Written Statement...
2024
-
[20]
Experimental evidence on the productivity effects of generative artificial intelligence
Noy, Shakked, and Whitney Zhang. “ Experimental evidence on the productivity effects of generative artificial intelligence .” Science , vol. 381, no. 6654, 2023, pp. 187-192. OpenAI. “ Understanding the source of what we see and hear online .” OpenAI , 7 May
2023
-
[21]
The economics of the music industry
Pannell, David. “ The economics of the music industry .” (2012). Patterson, David. “ How to build a bad research center .” Comm. of the ACM, vol. 57, no. 3, 2014, pp. 33-36. Patterson, David, Jeffrey Gilbert, et al . “ Energy and Emissions of Machine Learning on Smartphones vs....
2012
-
[22]
The effect of uncertainty in humidity and model parameters on the prediction of contrail energy forcing
Platt, John, Marc Shapiro, et al. “ The effect of uncertainty in humidity and model parameters on the prediction of contrail energy forcing .” Environmental Research Communications 6, no. 9 (2024): 095015. Reuel, Anka, Ben Bucknall, Stephen Casper, Tim Fist, Lisa Soder, et al ....
2024 arXiv
-
[23]
Healthcare spending as a percentage of GDP by country 2023
Vankar, Preeti. “ Healthcare spending as a percentage of GDP by country 2023 .” Statista , 9 September
2023
-
[24]
Language from police body camera footage shows racial disparities in officer respect
Voigt, Rob, Nicholas Camp, et al. " Language from police body camera footage shows racial disparities in officer respect. " Proc. National Academy of Sciences 114, no. 25 (2017): 6521-6526. Wang, Rose, Ana Ribeiro, Carly Robinson, Susanna Loeb, and Dora Demszky. “ Tutor CoPilot:...
2017 arXiv
-
[1962]
Automated identification of adults at risk for in-hospital clinical deterioration
Escobar, Gabriel, Vincent Liu, et al . “ Automated identification of adults at risk for in-hospital clinical deterioration .” New England Journal of Medicine , 383, no. 20 (2020): 1951-1960. Geraedts, Scott, Erica Brand, Thomas R. Dean, Sebastian Eastham, Carl Elkin, et al . “ ...
2020
-
[1981]
Scalable watermarking for identifying large language model outputs
Dathathri, Sumanth, Abigail See, Sumedh Ghaisas, et al. “ Scalable watermarking for identifying large language model outputs .” Nature 634, no. 8035 (2024): 818-823. Degrave, Jonas, Federico Felici, et al. “ Magnetic control of tokamak plasmas through deep reinforcement learni...
2024
-
[2012]
Embedding societal values into social media algorithms
Bernstein, Michael, Angèle Christin, et al . " Embedding societal values into social media algorithms. " Journal of Online Trust and Safety 2, no. 1 (2023). Bessen, James. “ AI and Jobs: The Role of Demand .” National Bureau of Economic Research , January
2023
-
[2013]
Foundation models for generalist medical artificial intelligence
Moor, Michael, Oishi Banerjee, et al. “ Foundation models for generalist medical artificial intelligence .” Nature , vol. 616, no. 7956, 2023, pp. 259-265. Morris, M., Jascha Sohl-Dickstein, Noah Fiedel, Tris Warkentin, et al . “ Levels of AGI for Operationalizing Progress on t...
2023
-
[2014]
Durably reducing conspiracy beliefs through dialogues with AI
Costello, Thomas, Gordon Pennycook, and David Rand . “ Durably reducing conspiracy beliefs through dialogues with AI. ” Science , vol. 385, no. 6714, 2024, p. eadq1814. Daly, Patricia. “ Agricultural employment: has the decline ended? ” Bureau of Labor Statistics ,
2024
-
[2018]
On the opportunities and risks of foundation models
Bommasani, Rishi, Drew Hudson, Ehsan Adeli, Russ Altman, et al . “ On the opportunities and risks of foundation models. ” arXiv preprint arXiv:2108.07258 (2021). Brynjolfsson, Erik. “ The Turing Trap: The Promise and Peril of Human-Like Artificial Intelligence .” Daedalus ,
2021 arXiv
-
[2022]
Getting AI Right: A 2050 Thought Experiment ,
Manyika, James. “ Getting AI Right: A 2050 Thought Experiment ,” The Digitalist Papers,
-
[2023]
Exposure to opposing views on social media can increase political polarization
Bail, Christopher, Lisa Argyle, Taylor Brown, et al ., “ Exposure to opposing views on social media can increase political polarization .” Proc. National Academy of Sciences , vol. 115, no. 37, 2018, pp. 9216-9221. Ball, John, Bryan Miller, and Erin Balogh, eds. " Improving di...
2018
-
[2024]
The labor market impacts of technological change: From unbridled enthusiasm to qualified optimism to vast uncertainty ,
Autor, David. “ The labor market impacts of technological change: From unbridled enthusiasm to qualified optimism to vast uncertainty ,” No. w30074, 2022, National Bureau of Economic Research. Autor, David, and Nathan Gardels. “ AI Could Actually Help Rebuild The Middle Class ....
2022
-
[9325]
Learning skillful medium-range global weather forecasting
Lam, Remi, Alvaro Sanchez-Gonzalez, Matthew Willson, et al . “ Learning skillful medium-range global weather forecasting .” Science 382, no. 6677 (2023): 1416-1421. Lång, Kristina, Viktoria Josefsson, Anna-Maria et al. “ Artificial intelligence-supported screen reading versus s...
2023
Reviewed August 11, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.