Posts tagged generative AI

When the Algorithm Knocks: France Faces the Future of Work

« The future is already here — it’s just not evenly distributed. »
William Gibson

A quick pulse-check

Indicator (2023-25)Latest figureSource
French jobs fully automatable5 %Sénat
Jobs partly exposed to AI, advanced economies40 %IMF
French workers who fear net job losses from AI75 %Labo
Finance teams already piloting or running AI58 %Gartner
Industrial-robot density, France (2022)180 / 10 000 workersStatista
Teachers using AI tools regularly≈ 20 %Le Monde.fr

The big picture: task take-over, not job wipe-out

France’s own Artificial-Intelligence Commission delivered a sobering — and surprisingly modest — number last spring: only one job in twenty is “directly replaceable” by current AI. The rest will merely be reshuffled, split or augmented. Global-scale anxiety persists, of course: the IMF puts 40 % of jobs in rich economies inside AI’s blast radius, meaning at least one core task could be automated.IMF Yet evidence from INSEE panel data shows AI-adopting French firms hire slightly faster than laggards, because productivity windfalls fund fresh roles in data, compliance and design. (Creative destruction is still creative.)

Sector by sector: who should sweat?

SectorTasks on the chopping blockNew (or rising) skillsCurrent tremor level
Finance & adminReconciliations, invoice coding, vanilla risk scoringData literacy, model oversight, client storytellingHigh – 58 % of teams already run AI; clerical head-counts inch down. Gartner
ManufacturingRepetitive welding, materials handlingRobot maintenance, OT-IT cybersecurityMedium-high – 6 400 new robots in 2023; density still half Germany’s. IFR International Federation of Robotics
HealthScan annotation, appointment triageInterpreting AI outputs, patient-side empathyLow – staff shortages mean augmentation, not cuts.
EducationMarking drills, drafting worksheetsDigital pedagogy, prompt-craftLow-medium – only 20 % of teachers use AI so far. Le Monde.fr
Media & creativeStock copy, basic illustrationCuration, narrative craft, IP savvyMedium – generative-AI tools flood studios; junior roles feel the squeeze.

Why the figures matter

  • Finance is already living through what McKinsey calls the “augmented-analyst” era: AI now cranks out first-pass pitch books; junior bankers edit rather than build. Clerical attrition is real, but the demand for model auditors and prompt engineers is rising even faster.
  • In factories, France’s relatively modest robot density (180 per 10 000) is a cue, not cause for comfort. If Paris wants to “ré-industrialiser” without exporting jobs to cheaper shores, cobots and predictive-maintenance AI are table stakes.
  • Hospitals fear burnout more than bots. Radiologists welcome the second pair of silicon eyes; nurses cheer paperwork-eating NLP.
  • Classrooms risk a digital divide within the staffroom: unless ministries accelerate the promised AI-literacy charter, the pupils will outrun their profs.
  • For journalists and designers, the genie is not going back in the bottle; French unions have already filed clauses limiting uncredited synthetic content.

What the state is doing — and should still do

  1. Scale training: government pledges to funnel France 2030 cash into nine AI clusters and to push CPF-funded micro-courses in data and model governance. Good — now publish an annual scoreboard of how many clerks, welders and editors actually switched careers.
  2. Audit the algorithms: the forthcoming EU AI Act will require bias-testing for HR and productivity tools. France could go further and give works councils a veto on opaque “boss-ware”.
  3. Reward augmentation, not redundancy: offer tax credits for AI deployments that raise per-worker output without shrinking payrolls.
  4. Target regional safety nets: an algorithmic risk-map (down to département-level) would flag which towns dominated by call-centres or fulfilment hubs need retraining subsidies first.

The bottom line

When the algorithm knocks, most French jobs will not be shown the door; they will be shown a new desk. The threat is less mass unemployment than mass redeployment. Whether that feels like liberation or displacement depends on politics, boardroom choices — and a national willingness to learn faster than the machines.

OpenAI vs. DeepSeek: The Business Models Race

In just a few months, generative artificial intelligence has gone from a laboratory curiosity to the next big technological frontier. Behind the excitement surrounding solutions such as ChatGPT (developed by OpenAI, backed by Microsoft) and DeepSeek (a Chinese startup aiming to rival the American giants), lie colossal challenges: profitability, sovereignty, workforce training, and even climate impact. This chronicle offers an overview of the key figures in AI, then explores three economic scenarios likely to shape the global landscape of this revolution.


1. Key Figures: What They Mean for Households and Businesses

To better grasp the stakes, the table below presents three main themes—productivity gains, employment, and energy costs—along with their real-life implications for individuals and companies.

Productivity Gains: An Inexhaustible Gold Mine?Jobs: 25% “Disrupted,” 12 Million “Created”Energy Costs and Cloud Spending: The Great Challenge
What the Figures Say– McKinsey (2023) anticipates up to $4.4 trillion per year in added value from generative AI.
– Across the EU, this could theoretically support a 2% GDP increase if widely adopted.
– Goldman Sachs (2023): up to 25% of jobs “disrupted” by 2030 (administration, customer support).
– World Economic Forum (2023): 12 million new positions focused on AI systems design and maintenance.
– Synergy Research (2023): $500 billion in cloud investments by 2026, driven by AI.
– Training a model (e.g., GPT-4) can emit hundreds of tons of CO₂ (University of Massachusetts, 2023).
Points of Comparison– $4.4 trillion is more than Germany’s annual GDP (around $4 trillion).
– For an SME, potential benefits could be seen in accounting, customer relations, and automating repetitive tasks.
– 25% is one quarter of the workforce in key sectors (accounting, telemarketing, etc.).
– 12 million new jobs is almost the active population of countries like Belgium or Greece.
– $500 billion is more than double France’s total annual budget for Education and Research.
– The carbon footprint of training a single AI model can equal thousands of long-haul flights.
ImplicationsFor households: potential drop in certain service costs (insurance, banking, legal advice), as AI can reduce operational expenses.
For businesses: need to invest in staff training and IT infrastructure to capitalize on these gains.
For households: risk of unemployment for vulnerable profiles but opportunities for younger, data-savvy professionals.
For businesses: retraining and new professions (AI consultants, “prompt engineers,” etc.).
For households: if servers increasingly consume energy, electricity bills or service fees (online hosting, for instance) could rise over time.
For businesses: sustainability becomes a major factor (cost and brand image), prompting a push to reduce data center consumption.

2. From Promising Numbers to Economic Models: The Necessary Transition

The data above highlights the vast potential of generative AI while revealing significant disparities. Much like Amazon in its early days, neither OpenAI nor DeepSeek has found the holy grail of profitability yet, despite massive investments. One relies on platform effects (OpenAI is strongly tied to Microsoft Azure), while the other leans on cost-efficiency and government support (DeepSeek and the Chinese market).

In an ecosystem where data centers consume hundreds of billions of dollars and enterprise adoption may be slower than predicted, the central question becomes: how can generative AI be monetized effectively? The table showcases the magnitude of possible gains, but also the financial, human, and environmental costs. Recent tech history (Google, Facebook, etc.) reminds us that business models often emerge empirically, shaped by trial and error as well as partnerships.

In this vibrant context, three economic scenarios stand out—each offering a distinct path to turn innovation into sustainable revenue while addressing concerns of sovereignty, competition, and environmental accountability.


3. Three Economic Scenarios for the OpenAI-DeepSeek Rivalry

Scenario A: The “Google-Style” Advertising Model

  • Principle: Provide free or freemium versions, leverage user attention, and monetize via targeted advertising (or user data sales).
  • OpenAI might thus strengthen ChatGPT’s integration with search engines (e.g., Bing) or social media.
  • DeepSeek, backed by Beijing, could favor a model less reliant on advertising, possibly state-subsidized to ensure data sovereignty and security.

Scenario B: The “Premium Licensing” and B2B Model

  • Principle: Reserve advanced versions (GPT-5, GPT-6, etc.) for clients able to pay high subscription fees, such as banks, insurers, and large industrial groups.
  • OpenAI would cover infrastructure costs by charging for exclusive access to its most powerful models.
  • DeepSeek, for its part, might offer turnkey solutions to strategic national sectors (banks, hospitals, government agencies), capitalizing on an ecosystem less open to American providers.

Scenario C: Technological Breakthrough and Cost Reduction

  • Principle: Simultaneously, advancements in AI hardware (dedicated chips) or training optimization (quantization, model distillation) could drastically reduce energy consumption.
  • Democratization Effect: Much like personal computers in the 1990s, declining unit costs would open generative AI to a wider range of players (SMEs, emerging countries), lowering entry barriers and pushing OpenAI and DeepSeek to stand out via product innovation rather than sheer infrastructure capabilities.

A Revolution to Be Invented

Between promises of productivity and risks of polarization (impacting employment and resources), generative AI finds itself at a turning point for the global economy. OpenAI and DeepSeek are its most publicized faces, yet they only represent the tip of a vast movement affecting every sector and raising fundamental questions: Who will finance the transition? How will the benefits be distributed? What rules will govern the geostrategic and environmental facets of AI adoption?

The numbers speak to a major economic opportunity, but the history of tech pioneers (Amazon, Google, Microsoft…) shows that profitability rarely follows a simple or rapid path. The scenarios outlined here illustrate the diversity of possible approaches—each with its pros and cons. As AI becomes embedded in our daily lives, this Sino-American rivalry underscores the need to craft, sometimes from scratch, a sustainable model that is both profitable and socially responsible.

Main References

  1. McKinsey (2023).
    The Economic Potential of Generative AI.
    https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier
  2. Goldman Sachs (2023).
    Generative AI Could Raise Global GDP by 7%.
    https://www.goldmansachs.com/insights/pages/generative-ai-could-raise-global-gdp-by-7-percent.html
  3. Synergy Research (2023).
    Cloud Market Reports.
    https://www.synergyresearchgroup.com/
  4. World Economic Forum (2023).
    Future of Jobs Report.
    https://www.weforum.org/reports
  5. University of Massachusetts (2023).
    Estimating CO₂ Emissions of LLMs (Energy and Policy Considerations for Deep Learning).
    https://arxiv.org/abs/1906.02243
  6. OECD (2024).
    SMEs and the Adoption of AI.
    https://www.oecd.org/going-digital/smes-and-the-adoption-of-ai/
  7. Reuters (2025).
    DeepSeek and Its Disruptive AI Model.
    https://www.reuters.com/technology/artificial-intelligence/what-is-deepseek-why-is-it-disrupting-ai-sector-2025-01-27/