AI supremacy: The artificial intelligence battle between China, USA and Europe - The Global AI Race, European Autonomy & Human Futures
Turning Point & Core Question: “What Makes Us Human?”
- Narrator frames current era as an historic hinge-moment; in ext5years AI will permeate "everything we do."
- Raises existential questions:
- Definition of consciousness & conscience.
- Unique human traits to “still be proud of.”
- Proposal to redefine ourselves from Homo sapiens (thinking) ➔ Homo sentiens (feeling/loving/meaning-seeking).
High-Stakes Global AI Race
- Competing for: careers, multi-billion $, geopolitical power, cultural influence.
- Principal blocs: United States, China, European Union.
- Additional players: Big Tech behemoths (Microsoft, Google, Meta, etc.) & VC-backed start-ups.
- Warning: “No second chances” for those who fall behind.
- Projected macro impact: AI expected to add 15\,\text{trn}toworldeconomyby2030.
European Efforts & Personalities
1. Jonas Andrulis & Aleph Alpha (Germany)
- Background: Former Apple senior AI researcher; childhood tinkerer (amateur radio, early coding).
- Founded Aleph Alpha before phrase “Generative AI” was popular.
- Headcount: 60 (multiple German sites; HQ Heidelberg).
- Funding rounds:
- Early seed: 28\,\text{M€}.
- Nov 2023 mega-round: \approx0.5\,\text{B€} led by SAP, Bosch, Schwarz/Lidl.
- Tech focus: Large multimodal foundation models for industry & public sector, not consumer chat.
- Distinguishing aims:
- European data sovereignty (on-prem HW, German data centres built with HPE).
- Explainability layer that cites both positive & negative evidence.
- Challenges:
- Relentless inbound from DAX corporates yet conservative sales cycles.
- Sudden competitive shock: release of GPT-4 (14 Mar 2023) cutting cost & raising bar.
- Media scrutiny (Die Zeit article on toxic outputs) ⇒ trade-off between safety vs capability.
- Lobbying role: Voices EU startup concerns during AI Act negotiations; calls for “fair race.”
2. Thomas Wolf & Hugging Face (France/US/NL)
- Co-founder & Chief Science Officer; company at \approx200 staff.
- Product: Open platform + hub to share / benchmark ML models & datasets (GitHub for AI).
- Core values labelled “very European” ➔ transparency, privacy, accountability.
- Funding: 235\,\text{M US$} from Google, Amazon, Nvidia, AMD, others.
- Hosted Meta’s LLaMA 2 release (open-weight GPT rival).
- Research ambitions: empirically map “value pluralism” across models (Anglo-American vs French vs German answer styles).
- Ethical tension: open-sourcing lowers barrier but eases misuse (e.g., weapon molecule generator case study—VX nerve agent found in <4 h).
3. EU Policy Landscape
- First omnibus law: EU AI Act (finalized Dec 2023) → risk-tiered regulation, bans on social scoring & mass biometric surveillance.
- Politicians (Robert Habeck, etc.) tout strategic autonomy; link to industrial grants & public-sector procurement to nurture European LLMs.
Chinese AI Ecosystem & Han Xiao (Jina AI)
- Han’s CV: Olympiad math schooling, Tencent alum, Zalando alumnus; father CS professor.
- Company structure: HQ Berlin, R&D Shenzhen/Beijing; workforce highly international.
- Product suite: Multimodal search & embedding platform (“AI for devs”); forthcoming OpenAI-competitor PaaS.
- Observations:
- At World AI Conference Shanghai, 30 LLMs released in single day—mix of BAT giants & banks.
- Prediction: eventual top models will be Chinese due to agile imitation + scale.
- Chinese state strategy (2017 “Next Gen AI Plan”): explicit goal of global AI supremacy; AI = lever for military & surveillance power.
- Infrastructure reality: >500 “city brain” command centres; nationwide CCTV densest worldwide; citizens accept due to safety narrative.
U.S. Tech Titans & Power Dynamics
- Microsoft–OpenAI alliance: 10\,\text{B US$} equity + deep Azure/Office/Teams integration ("wave of Microsoft money").
- Microsoft Research Asia (1990s Bill Gates bet) became incubator for Chinese AI leadership (SenseTime, Megvii alumni).
- Senate hearing (May 2023): Sam Altman acknowledges risk of LLM-enabled persuasion; analogy to Photoshop but “far stronger.”
- Biden White House: AI briefing up to 3× weekly; calls for trans-Atlantic coordination.
Safety, Pause Campaign & Philosophical Debates
- Max Tegmark (MIT, Future of Life Institute) history:
- 2014: “AI safety” taboo; 2023 open letter calling 6-month pause on models > GPT-4.
- Signatories: Elon Musk, Steve Wozniak, Yoshua Bengio, etc.
- Framing: risk of extinction comparable to pandemics & nukes.
- Critiques: Could double as marketing—spotlight capabilities, shift blame to regulators.
- Academic work: MIT/Google paper—LLMs can predict public-opinion polls → election manipulation scenario.
- Tegmark lab modelling “conflict dynamics” mathematically: naive public believes AI helpful ⇒ under-invests in resistance.
Open Source vs Closed Models – Pros & Cons
- Pros: democratized innovation, transparency, faster science (“Lego analogy”).
- Cons: safety guardrails removable, dual-use chemical/bioweapon design, small elite still shapes future.
- Ongoing question: What obligations on open-weight publishers?
Economic, Labor & Creative Impacts
- Knowledge-work in Europe deemed “at risk”: potential for new empire or collapse of white-collar pillar.
- Arts/Music Example – Dr. Enongo Kasango (“Sammus”):
- 2022: fully AI-generated rapper; 2023: cloned voices of Drake/The Weeknd.
- Concern: capitalism incentivizes replacement & “average-ness;” hope lies in weird/anomalous creativity.
- Chinese business view: AI cheaper, faster, inexhaustible; firms already substituting humans.
Infrastructure & Hardware Bottlenecks
- Training frontier LLMs demands \mathcal{O}(10^{3}-10^{4}) high-end GPUs (scarce).
- Smaller firms often forced into hyperscaler clouds ⇒ loss of sovereignty; Aleph Alpha builds own German datacentre with HPE.
Investment & Business Development Tactics
- Visibility = capital: panels, media, pilot projects (e.g., Heidelberg city chatbot) act as testimonials.
- Need for “money + non-monetary help” (distribution, enterprise channels) illustrated by Microsoft–OpenAI vs resource-strained EU start-ups.
Regulation, Governance & Democracy
- EU-US must “lead with like-minded friends,” per summit in Luleå, Sweden (31 May 2023).
- U.S. system intentionally slow—balance between efficiency & civil-rights; risk of industry capture.
- Call for mass public education so citizens know limits (analogy: learning Photoshop/GPS).
Ethical & Practical Implications Summarized
- Manipulation & disinformation in upcoming 2024 global elections.
- Potential loss of human agency if automation becomes ubiquitous & reliable 99\% of time.
- Need for “antidotes” (regulation, plural values, transparency, safety research).
Connections to Earlier Lectures / Principles
- Echoes classical philosophical worry about tech outrunning governance (20th-century warnings).
- Intersects with industrial-era concept of comparative advantage shifting from labor to data/compute.
- Relates to past course modules on data privacy (CCTV example) & cloud-sovereignty (hyperscaler lock-in).
Numerical & Statistical References (LaTeX)
- 10\,\text{B US$} Microsoft investment in OpenAI.
- 0.5\,\text{B€} Aleph Alpha Series B (one of EU’s largest).
- 235\,\text{M US$} Series C for Hugging Face.
- 28\,\text{M€} initial funding Aleph Alpha vs Microsoft wave.
- 15\,\text{trn}globalGDPupliftforecastby2030 from AI.
- >500 Chinese “city brain” surveillance centres.
- 30$$ Chinese LLMs announced in a single day (Shanghai, 2023).
- Senate claim: LLMs predict opinion polls within few % error.
Forward-Looking Recommendations & Study Prompts
- Monitor intersection of compute supply chains (GPUs, energy) with AI sovereignty.
- Compare AI Act provisions to U.S./Chinese regulatory stances.
- Debate merits of 6-month global pause: feasible or utopian?
- Explore role of public-sector procurement as demand-side stimulant for domestic AI.
- Reflect on personal skill development: coding + emotional intelligence to remain uniquely valuable.