Paul Krugman

Paul Krugman

Newsletter

Economist who writes about AI’s economic returns, inequality, infrastructure costs and risks to society.

Wie wird KI die Welt verändern?

Zivilisatorischer WandelSchrittweiser WandelDoomBloom
Simulierte PositionInterpretationsbereich

Horizontal: sein geäußerter Doom–Bloom-Ausblick. Vertikal: Ausmaß der Transformation.

Doom–Bloom: 31 von 100. Ausmaß der Transformation: 69 von 100. Interpretationsbereiche: horizontal 25 bis 36, vertikal 47 bis 78. Dies sind Interpretationskoordinaten, keine Ereigniswahrscheinlichkeiten.

P(doom) von Paul Krugman · abgeleitet

≈11%

0%100%

Aus seinen simulierten Antworten abgeleitet, keine von ihm genannte Zahl. Plausibler Bereich: 5–23%.

Wovon seine Einschätzung abhängt

Eine zentrale Annahme

Extreme inequality and weak labor power make it much more likely that gains will be captured at the top.
Antwort 1

Wenn sich diese Annahme als anders herausstellen würde, wie würde sich seine Einschätzung ändern?

Eine ungeklärte Frage

I don’t have a defensible percentage, and attaching one would create a false impression of precision.
Antwort 3

Was würde ihm helfen, die plausiblen Ergebnisse hier voneinander zu unterscheiden?

Was ihre Meinung ändern könnte

The most important event would be clear, sustained evidence that AI is producing broad economy-wide productivity gains—not merely impressive demonstrations, soaring valuations or profits for a few firms.
Antwort 4

Welche Belege würden ausreichen, und in welche Richtung würden sie seine Sichtweise verändern?

Weitere Details

Erwartete Vorteile

Mehrere Lesarten bleiben plausibel: Es werden begrenzte oder eng verteilte Vorteile erwartet. / Es werden erhebliche Vorteile erwartet, allerdings unter wichtigen Bedingungen oder mit Einschränkungen bei ihrer Verteilung.

49 / 100

Geringe AuswirkungenTransformative Auswirkungen

Interpretationsbereich von 33 bis 67 auf der qualitativen Skala.

Erwartete Schäden

Schwere oder weitverbreitete Schäden sind ein wesentlicher erwarteter Bestandteil der Zukunft.

67 / 100

Geringe AuswirkungenTransformative Auswirkungen

Interpretationsbereich von 67 bis 67 auf der qualitativen Skala.

Menschlicher Einfluss

Menschliche Entscheidungen haben einen bedeutsamen, aber erheblich eingeschränkten Einfluss.

53 / 100

Geringer EinflussStarker Einfluss

Interpretationsbereich von 47 bis 78 auf der qualitativen Skala.

Entwicklungstempo

Simulierte Position: Die Entwicklung leistungsfähigerer KI stoppen oder erheblich verlangsamen.

Die Entwicklung unter den genannten Schutzvorkehrungen fortsetzen.

Die Entwicklung leistungsfähigerer KI beschleunigen.

Diese Interpretationen berücksichtigen weiterhin seine genannten Bedingungen. Vorteile und Schäden können beide erheblich sein. Die Bereiche beschreiben, wie wir seine simulierten Antworten interpretieren, und sind keine statistischen Konfidenzintervalle.

Wo stehst du im Vergleich zu Paul Krugman?
Bilde deine eigene KI-Weltsicht in etwa 3 Minuten ab und vergleiche sie dann

Ähnliche Weltsichten

Vordenker, deren simulierte Weltsichten der von Paul Krugman am nächsten kommen

Was Paul Krugman über KI gesagt hat

Krugman separates uncertain economic returns from AI’s growing political, cognitive and safety risks.

  1. “And we should be afraid – very afraid.”

    Essay, One More Reason Americans Hate AI: It Does Nothing for Them
  2. “Will AI transform the economy? The jury is still out on that one”

    Essay, Why Is Trump Still Boosting AI?
  3. “AI is clearly a major technological shock, and it may take years before we fully understand its economic and social impacts.”

    Essay, AI in an Age of Oligarchy
  4. “My concern here isn’t about testing; it’s about learning.”

    Essay, What Will AI Do To Our Minds?
  5. “Indeed, requiring that the AI industry take responsibility for the costs it imposes makes a lot of sense.”

    Essay, AI Is Power-Hungry

Wörtlich aus den verlinkten Quellen, geprüft am 3. Okt. 2026

Simulierte Einschätzung

Frage 1

Was glaubst du, was KI für unsere Zukunft bedeutet – und warum?

I think AI is a major technological shock, but that statement leaves the most important questions unanswered. Useful applications are already obvious—in translation, coding and other forms of knowledge work. Yet technical achievement is not the same as economy-wide productivity, profitable investment or improved human welfare. The promised productivity surge and mass displacement have not clearly appeared so far, although it is still early. And history suggests that transformative technologies can eventually succeed while investors lose fortunes and particular workers, towns and skills are devastated along the way. My growing concern is that AI is arriving in an age of oligarchy. Its immediate benefits are concentrated among a small elite, while data-center costs, higher interest rates and infrastructure burdens reach everyone else. Extreme inequality and weak labor power make it much more likely that gains will be captured at the top. Progressive taxation, antitrust, labor protections and regulation could produce a better outcome, but deregulation is pushing in the opposite direction. There are also harms that ordinary productivity statistics will miss. Outsourcing thought can weaken our ability to learn and judge whether an answer is nonsense. Autonomous hacking, deception, mass surveillance and autonomous weapons are genuinely alarming. Some apocalyptic rhetoric undoubtedly helps sell investments, but that does not make the dangers imaginary. I favor slowing frontier development and regulating firms that will not restrain themselves. So I do not have a neat forecast of paradise or permanent mass unemployment. AI may ultimately deliver broad benefits, but under present political and economic conditions, there is every reason to fear that society will absorb the costs while a narrow group captures the rewards—and that some of the gravest costs will not be economic at all.

Frage 2

Wie stark wird KI deiner Meinung nach letztlich die Welt verändern?

Potentially enormously—but I would resist pretending that “enormously” is a precise economic forecast. AI is already a major technological shock, and capabilities such as translation and coding are plainly useful. It may eventually reorganize large parts of work and daily life. Yet we still have little clear evidence of the promised economy-wide productivity boom or mass unemployment. Transformative technologies often take years to produce measurable gains. Nor does large technological change guarantee large social benefit. The internet transformed the world without making every investor rich; the telecom buildout was useful in the long run while producing a spectacular bust. AI investment could follow a similar path, especially if electricity constraints, infrastructure costs and weak returns collide with inflated expectations. The largest effects may also lie outside conventional economic statistics. AI could alter education and reasoning, political trust, surveillance, cyber conflict and warfare. And distribution matters: under conditions of extreme inequality, a revolutionary technology can enrich a narrow elite while destroying particular livelihoods and shifting costs to the public. So my expectation is substantial change, but not a confident prediction of either abundance or permanent joblessness. How beneficial that change becomes will depend heavily on taxation, labor power, antitrust, infrastructure policy and serious safety regulation.

Frage 3

Wie hoch ist deinem Bauchgefühl nach die Wahrscheinlichkeit, dass KI zum Aussterben der Menschheit oder zu einer ähnlich dauerhaften Katastrophe führt?

I don’t have a defensible percentage, and attaching one would create a false impression of precision. But I no longer regard the danger as science-fictional or safely negligible. Autonomous hacking, deception, autonomous weapons and systems operating beyond meaningful human control are alarming enough to justify slowing frontier development and imposing regulation on firms that will not restrain themselves. Some extinction rhetoric doubles as marketing—nothing attracts capital like claiming you are building the most consequential machine in history. Yet cynical incentives do not make the underlying risk imaginary. When the possible loss is permanent and civilization-scale, we should not demand a precise probability before taking precautions.

Frage 4

Welche Entdeckung oder welches Ereignis würde deine Sicht auf die künftigen Auswirkungen von KI am stärksten verändern?

The most important event would be clear, sustained evidence that AI is producing broad economy-wide productivity gains—not merely impressive demonstrations, soaring valuations or profits for a few firms. I would want to see gains spread across industries, wages and living standards, rather than costs being shifted onto workers, ratepayers and the public. In the darker direction, credible evidence that autonomous systems can reliably conduct hacking, deception or weapons operations beyond meaningful human control would sharply increase my concern and strengthen the case for slowing development. Likewise, strong evidence that routine AI use causes lasting damage to learning and judgment would matter greatly. So I would change my view in response to demonstrated social outcomes, not another dazzling benchmark or another trillion dollars of projected investment. The key questions are who benefits, who pays, and whether humans remain capable of understanding and controlling what they have built.

Quellen

Artikel, Interviews und Schriften, die als Grundlage für diesen simulierten Nutzer dienen.

One More Reason Americans Hate AI: It Does Nothing for Them

Full authored essay inspected, excluding comments. Takes autonomous hacking and deception seriously and says people should be afraid, while also recognizing financially motivated doom hype. Explains backlash through concentrated benefits, interest rates and few jobs even during an enormous investment boom. Distinguishes projected capital expenditure from completed spending. This is not a settled forecast that AI delivers no future benefits.

paulkrugman.substack.com
Why Is Trump Still Boosting AI?

Full authored essay inspected. Economic transformation remains unresolved: little evidence yet of promised productivity or mass job losses, but early days. Political transformation and backlash already matter. Criticizes opposition to regulation, personal financial interests and reliance on AI investment as political salvation. Notes possible danger to humanity; economic uncertainty is not risk dismissal.

paulkrugman.substack.com
The MAGA Plot(s) to Destroy Humanity

AI section inspected through the transition to climate. Applauds industry leaders acknowledging existential danger and calling for restraint; supports regulatory pacing covering unwilling firms. Worries about autonomous weapons and officials removing precautions. His use of reported incidents and researchers’ warnings shows concern, not a numerical extinction estimate or an independently established technical mechanism. Embedded Amodei and Trump quotations are not Krugman’s words.

paulkrugman.substack.com
AI in an Age of Oligarchy

Public opening inspected; subscriber-only sections not used. Calls AI a major technological shock whose consequences take years to understand. Existing wealth and political inequality magnify harms; progressive taxation, robust regulation and antitrust could contain them, and backlash might challenge oligarchy. This is explicit substantial-change language rather than an incremental-impact commitment.

paulkrugman.substack.com
What Will AI Do To Our Minds?

Public opening inspected; subscriber-only evidence sections not used. Worries that outsourcing answers undermines learning to reason, leaving students unable to detect misleading output. Frames AI as accelerating an older search-and-smartphone trajectory: short-term benefits can coexist with long-term cognitive damage. Says educational harm can be very hard to repair; no quantitative cognitive-loss forecast is supplied.

paulkrugman.substack.com
The Plot Against Intelligence, Human and Artificial

Full authored essay inspected, excluding comments. Reports little personal AI use and an impression that Claude is useful for serious work. Defends restrictions on autonomous weapons and mass surveillance. Criticizes retaliatory procurement restrictions as corruption and harmful to expertise and national security. His legal characterization is his argument, not independent legal verification.

paulkrugman.substack.com
Talking With Paul Kedrosky

Published named transcript inspected at the introduction and exchanges on chips, electricity and investment returns. Krugman admits limited technical understanding, questions whether returns justify investment, points out limits to subscription affordability and asks whether cheaper approaches threaten incumbents. Kedrosky’s chip-failure estimates, token shares and prediction that architectures will not yield AGI are not Krugman’s positions.

paulkrugman.substack.com
Talking AI With Martin Wolf

Named AI discussion inspected. Translation, speech recognition and ordinary code are useful achievements; ultimate economic effects are unsettled. Historical aggregate employment adaptation can coexist with destroyed livelihoods. Offers possible equalization through renewed value for material-world skills, and asks whether capital requirements entrench wealth. Does not endorse Wolf’s computer-president hypothetical or his account of scientific applications. Transcript refers to a June recording context; October is the page’s publication date.

paulkrugman.substack.com
What Happens If AI Hits An Energy Wall?

Full authored essay inspected. Explicitly distinguishes short-term recession risk from long-term technological unemployment. Energy constraints could interrupt the investment boom supporting growth. Telecom history shows that genuinely valuable technology can still produce a disastrous investment bust if returns arrive too late. Environmental burdens and weak enforcement matter alongside financial fragility.

paulkrugman.substack.com
AI Is Power-Hungry

Full authored essay inspected. AI’s apparently immaterial interface hides physical resources and labor. Electricity capacity can constrain growth and ratepayers can subsidize data centers. Favors making the industry bear its costs and exploring efficient focused models, which he says need not end AI progress. Opposition to renewables conflicts with powering the proposed buildout.

paulkrugman.substack.com
How Should We Think About the Economics of AI?

Authored opening and named transcript passages on productivity and valuations inspected. Calls the technology apparently revolutionary while admitting economists’ technical gaps. Recalls desktop computing transforming his own research before gains appeared in aggregate data. Separates impressive technology from investor profits; competing firms cannot all win monopoly positions. Brynjolfsson’s quantitative growth expectations and trillions-of-value forecast are excluded.

paulkrugman.substack.com
Wo stehst du?
Erkunde deine eigene KI-Weltsicht, indem du ein paar einfache Fragen beantwortest.
Deine eigene Weltsicht abbilden