AI supremacy The artificial intelligence battle between China USA & Europe / Global AI Race & European Start-up Strategies – Comprehensive Study Notes

Speakers repeatedly frame 2023-2024 as the historical pivot when “everything changes.”
Core philosophical prompt: “What does it mean to be human when machines equal or surpass us?”
- Emphasis on consciousness, empathy, curiosity, purpose, love → proposed new label “Homo sentiens.”
Prediction: within $5$ years AI will permeate “everything we do.”

Stakes: careers, economic growth, political power, military dominance.
Main blocs in competition:
- United States: Big-Tech-driven (Microsoft–OpenAI, Google, Anthropic, etc.).
- People’s Republic of China: state-aligned tech giants (Tencent, Alibaba, Baidu) & Communist Party goals.
- European Union: seeks strategic autonomy & value-aligned technology.
Prize: projected boost of $15\text{\ trillion}$ to global GDP by $2030$ .

Jonas Andrulis | Aleph Alpha (Germany)
- Former Apple AI researcher; founded company in Heidelberg.
- Goal: sovereign European LLMs for industry/public-sector (not consumer chatbots).
- Funding trajectory: € $28$ M → ~€ $500$ M round (SAP, Bosch, Schwarz-Lidl, VC, govt).
- Strategic pillars: own research, in-house data centre, HPE hardware partnership, EU lobbying.
Thomas Wolf | Hugging Face (France/US/NL)
- 200 employees; open-source model hub & collaboration platform.
- Values: pluralism, transparency, data accountability, “European-ish” cautiousness.
- Raised $235\text{ million}$ (Google, Amazon, Nvidia, AMD). Valuation $\approx 4.5\text{ billion}$ .
- Hosts Meta’s LLaMA-2; studies cultural bias (ChatGPT vs BloomChat vs regional models).
Han Xiao | Jina AI (Berlin–Shenzhen–Beijing)
- Ex-Tencent & Zalando engineer; mission: multimodal embedding & developer tools (Prompt Perfect, embeddings-as-a-service).
- Aims for top-10 benchmark positions to attract funding; balances Western & Chinese markets.

Microsoft invested $10\text{ billion}$ into OpenAI ➔ deep integration across MS products & Azure.
GPT-4 launch (14 Mar 2023): performance shock + drastic cost decline; threatens smaller competitors’ business cases.
Sam Altman U.S. Senate testimony: warns of persuasion/disinformation risks; calls for regulation.
CEOs (OpenAI, Google, Microsoft) sign statement equating AI-risk to pandemic & nuclear war.

2017 national strategy: explicit goal to dominate global AI.
Microsoft Research Asia (1990s) produced talent for SenseTime, Megvii, etc.; “incubator” of PRC AI elite.
Massive urban “City Brain” systems: > $500$ smart-city hubs, ubiquitous CCTV + facial recognition.
AI exported to authoritarian partners; narrative: safety, crime-free cities, anti-terror.
CCP concern: LLM “brains” must reflect Chinese ideology; fear dependence on U.S. hardware.

Training frontier models needs $\mathcal{O}(10^4)$ high-end GPUs ➔ global shortage.
Small firms trade equity/independence for access to compute; Aleph Alpha resists via local HPE cluster.

AI kiosk greeter, citizen Q&A (traffic on B37, child-benefit instructions).
Feedback loops: error logs sent to AA for retraining; priority = accuracy + explainability.
Goal: accelerate bureaucratic workflows, address labour shortages (“solidarity”).

Lego metaphor: open models are building blocks; users can inspect “under the hood.”
Security dilemma:
- Open access ↑ innovation & auditability.
- Also ↑ misuse (terrorist molecule design, disinformation, deepfakes).
Hugging Face internally debates gating/opt-in licensing to curb malfeasance.

Artists/Musicians (e.g., rapper Sammus) see voice cloning & AI-generated music erode copyright & income.
Chinese firms already replacing human creative labour for efficiency.
Advice: cultivate “anomalous” creativity beyond average patterns; teach children coding + unique human skills.

March 2023 Future of Life Institute letter (Max Tegmark) → pause training beyond GPT-4 for $\ge 6$ months; signed by Elon Musk, Steve Wozniak, Yoshua Bengio, etc.
Motivations: existential risk, manipulation of elections (LLMs predict public opinion).
Critics: letter also serves as marketing—highlights tech’s power & creates regulatory moat.

EU AI Act (world’s first broad AI law)
- Bans social scoring & mass surveillance; imposes risk-tier obligations.
- Germany–France–Italy tried to soften foundation-model clauses; Parliament prevailed.
- Dragos Tudorache: regulation = geopolitical power via standards.
U.S. struggles: constitutional inertia + Big-Tech lobbying; White House convenes weekly AI meetings.
Call for transatlantic alignment; inclusion of other democracies.

Attention economy: conference panels, press (“German answer to ChatGPT”), benchmarking leaderboards.
Die Zeit exposé on bias prompts ➔ Andrulis argues for user-side responsibility vs over-filtering.
Continuous lobbying tours (Brussels, Berlin, Sweden ministerial meetings, Las Vegas HPE Showcase).

Top-10 placement on public leaderboards critical for venture funding (Han Xiao conversation).
“Cambrian explosion” of startups; > $30$ Chinese LLMs unveiled in one day at WAIC Shanghai.

Andrulis: childhood ham-radio tinkerer; sleeps on office couch answering emails.
Wolf: balances research with family sailing holiday; values pluralistic ethics.
Han Xiao: math Olympiad schooling, perpetual Saturday classes; insists on lifelong learning for staff.
Tegmark: MIT professor turned safety advocate; quantitative modelling of social conflict over AI.

Will Europe scale funding from millions → billions fast enough?
Can open-source remain competitive while mitigating security externalities?
How will democracies preserve electoral integrity amid interactive disinformation?
Will creative labour evolve or be commoditised by “average-seeking” algorithms?
Long-term hope: human strengths in emotion & meaning (sentience) complement machine cognition.

Funding: $10\text{ billion}$ (Microsoft → OpenAI), $28\text{ million}$ → ≈€ $500\text{ million}$ (Aleph Alpha), $235\text{ million}$ (Hugging Face), $4.5\text{ billion}$ valuation.
Economic impact: $15\text{ trillion}$ by $2030$ .
Hardware scale: $10^4$ + GPUs; $>500$ smart-city “brains.”
Future of Life pause request: $\ge6$ -month moratorium.

Understand dual-use nature of generative models; weigh openness vs control.
Track hardware supply chains; compute availability shapes geopolitics.
Monitor regulatory developments (EU AI Act) as potential career niche.
Cultivate unique human skills (creativity, empathy, critical thinking) alongside technical literacy.