Podcast


Central Problem

The paper confronts whether the current Artificial Intelligence hype constitutes another tech bubble—and if so, what this means for the technology’s future and society’s response. Floridi addresses the pattern of boom-and-bust cycles in technology sectors, examining whether AI exhibits the characteristic features of previous bubbles (Dot-Com, Telecom, Cryptocurrency, etc.) and warning that the bubble’s eventual burst could trigger a new AI Winter with significant consequences for research, investment, and adoption.

Main Thesis

The current AI hype cycle exhibits all the defining characteristics of a tech bubble: enormous price increases disconnected from fundamentals, new and flawed valuation paradigms, retail investor FOMO, regulatory gaps, and widespread media hype. Floridi argues that this bubble, centered on generative AI and large language models since ChatGPT’s release in November 2022, will likely follow the same trajectory as previous tech bubbles—culminating in a market correction that risks overcorrection and potential AI Winter.

The thesis draws on comparative analysis of five previous tech bubbles to extract invariant features and lessons. Despite these historical precedents, the technology industry appears to have learned nothing from past bubbles—the same patterns of speculation, inflated valuations, and unsustainable business models recur. The challenge is not to prevent the bubble’s burst but to minimize its destructive impact.

Historical Context

The paper appears in late 2024, approximately two years after ChatGPT’s public release catalyzed an unprecedented surge of AI investment, media attention, and speculative activity. This period saw AI companies achieve extraordinary valuations despite limited profitability—OpenAI, for example, was valued at over 5 billion annually.

Floridi situates AI hype within the longer history of tech bubbles: the Dot-Com Bubble (1995-2000), the Telecom Bubble (1996-2002), the Chinese Tech Bubble (2014-2015), the Cryptocurrency Bubble (2011-present), and the COVID-era Tech Stock Bubble (2020-2021). Each followed similar patterns and should have taught similar lessons—but these lessons remain unlearned.

The essay also notes AI’s own history of cycles—previous “AI Summers” followed by “AI Winters” when expectations outpaced capabilities. The current hype risks triggering another such winter, damaging legitimate AI research and beneficial applications.

Philosophical Lineage

flowchart TD
    Hegel[Hegel] --> HistoricalLessons[Historical Lessons Unlearned]
    EconomicTheory[Economic Theory] --> BubbleMorphology[Bubble Morphology]
    BubbleMorphology --> Floridi[Floridi]
    PreviousBubbles[Previous Tech Bubbles] --> Floridi
    AIHistory[AI History] --> AIWinters[AI Winters]
    AIWinters --> Floridi
    DigitalEthics[Digital Ethics] --> Floridi
    Wittgenstein[Wittgenstein] --> FamilyResemblance[Family Resemblance]
    FamilyResemblance --> BubbleComparison[Bubble Comparison]
    BubbleComparison --> Floridi

    class Hegel,HistoricalLessons,EconomicTheory,BubbleMorphology,Floridi,PreviousBubbles,AIHistory,AIWinters,DigitalEthics,Wittgenstein,FamilyResemblance,BubbleComparison internal-link;

Key Thinkers

ThinkerDatesMovementMain WorkCore Concept
Floridi1964-Digital EthicsThe Ethics of Artificial IntelligenceInformation philosophy, digital ethics
Hegel1770-1831German IdealismPhilosophy of HistoryLearning from history
Wittgenstein1889-1951Analytic PhilosophyPhilosophical InvestigationsFamily resemblance

Key Concepts

ConceptDefinitionRelated to
Tech bubbleMarket phenomenon with rapid, unsustainable growth in tech valuations driven by speculation rather than fundamentalsEconomics, Finance
AI WinterPeriod of reduced funding, interest, and research activity in AI following deflated expectationsAI History, Research Policy
FOMOFear of missing out; psychological driver of speculative investment in emerging technologiesBehavioral Economics, Bubbles
Regulatory gapSituation where regulatory frameworks are absent or lag behind market developmentsPolicy, Governance
Greater fool theorySpeculation that overvalued assets can be sold at higher prices to subsequent buyersEconomics, Speculation

Authors Comparison

ThemeDot-Com BubbleCrypto BubbleAI Bubble
Core technologyInternet/WebBlockchainGenerative AI/LLMs
New metricsEyeballs, page viewsTotal value lockedModel parameters, benchmarks
Regulatory statusEmergingLargely absentLagging (EU AI Act)
Retail participationSignificantDominantGrowing
Duration~5 years~13+ years (cycles)~2 years (ongoing)

Influences & Connections

  • Historical precedents: Floridi ← learns from ← Dot-Com, Telecom, Chinese Tech, Crypto, COVID Tech bubbles
  • Philosophical framework: Wittgenstein’s family resemblance → applied to → bubble comparison
  • Warning: Hegel’s dictum about learning from history → confirmed by → repeated bubble patterns
  • Policy implications: Bubble analysis → informs → regulatory recommendations

Summary Formulas

  • Bubble morphology: Tech bubbles share five features: disruptive technology at core, speculation outpacing reality, new valuation paradigms, retail investor FOMO, regulatory gaps and lag.
  • Nothing learned: “The only thing we learn from the history of tech bubbles is that we learn nothing from it” (paraphrasing Hegel).
  • AI-specific risks: AI hype compounds general bubble risks with AI-specific dangers including AI washing, talent war inflation, and potential new AI Winter.
  • Mitigation strategy: Focus on sustainable business models, maintain critical perspective, prioritize longer-term thinking, support appropriate regulation.

Timeline

YearEvent
1995-2000Dot-Com Bubble inflates
2000-2002Dot-Com and Telecom Bubbles burst
2015Chinese Tech Bubble correction
2017-2018First major Cryptocurrency bubble cycle
2020-2021COVID-era Tech Stock Bubble
2022ChatGPT released, AI hype accelerates
2024Floridi publishes bubble analysis

Notable Quotes

“The only thing we learn from the history of tech bubbles is that we learn nothing from it.”

“This time is not different. The AI bubble will probably follow the same pattern as the other five analysed in this article.”

“I do not hope that what I have argued in this article will make any significant difference. It is written not with hope—which would be epistemically unreasonable—but for hope—which may be morally commendable.” This annotation was normalised using a large language model and may contain inaccuracies. These texts serve as preliminary study resources rather than exhaustive references.