Alvin Wang Graylin

persona · active · confidence: medium · reviewed 2026-09-21 · tier: semiannual

Who he is

Affiliation (role check run 2026-09-21): No longer at HTC. Senior Fellow for Technology, Asia Society Policy Institute — Center for China Analysis; Digital Fellow, Stanford Institute for Human-Centered AI (Digital Economy Lab); Lecturer, MIT; Chairman, Virtual World Society (nonprofit for safe XR/AI adoption); Advisory Council, Braver Angels; ECE Advisory Board, University of Washington. His own bios add "Professor of AI/Tech Policy, University of Washington" — the UW page found shows only the advisory-board seat, so that title is unverified. Writes the Abundanist Substack (~2025–). Ex-HTC: China President 2016–2023, then Global VP of Corporate Development; still "Global VP, HTC" in Feb and Jun 2024 bios, no HTC role by Apr 2025 — exact departure date not found in any public source. Co-author with Louis Rosenberg of Our Next Reality: How the AI-powered Metaverse Will Reshape the World (Hachette, 2024-03; foreword by Neal Stephenson).
One-line position (v1, kept): An optimist who insists on being a realist — AI and immersive tech can deliver genuine abundance, but only if the governance work (aligned incentives, an AI Bill of Rights, US–China cooperation rather than race dynamics) is done deliberately rather than left to markets.
One-line position (2026 revision): "A race is not a race if the competitors are running toward different destinations" (Quincy Institute, 2026-09-18). The US is over-spending on an AGI sprint it cannot win outright, under-spending on diffusion and shock absorbers, and heading into a debt-plus-displacement squeeze; the fix is cooperation abroad and an automation levy, a "GI Bill for the AI age" and universal basic infrastructure at home. The metaverse has left his vocabulary.

Discipline & technical bet

Engineer by training (BSEE UW 1993, VR and neural-net focus; MIT SM EECS + SM Sloan 2000), 30+ years an operator — Intel, IBM, Trend Micro, four venture-backed startups on both sides of the Pacific, then HTC's XR business for nearly a decade — re-based since 2025 as a policy strategist. His technical wagers now, all public and checkable: (1) algorithms beat scale — ~2x error reduction per 10x compute vs "5 or 10x" from algorithmic change (CFR, 2026-07-08); rhymes with Sutskever's "age of scaling is over." (2) Open-weight models converge on closed ones — Chinese open models from <2% of token traffic (late 2024) to a majority among top models (2026); US lead down from 12–14 months to 2–3 (his figures, Quincy 2026-09-18; not independently verified). (3) Danger does not scale with model size — a sub-100M-parameter drug-discovery model produced 40,000 candidate chemical agents in six hours; the threat is "a 30-billion-parameter uncensored model, a competent harness, and an unknown creator," so compute-threshold rules aim at the wrong thing. (4) China's edge is the system, not the model — energy (his figures: 3.9 TW vs 1.34 TW US capacity, 2025), manufacturing, robot volumes (Unitree ~30–40K quadrupeds, 5–10K humanoids, 2025), 300M-patient health data (WAIC report, 2026-08-13). The XR bet survives only as the Virtual World Society's "Bridge to One" initiative and an AWE USA 2026 slot.

Key claims (Says)

Notable predictions — with falsifiable checks

Revealed behavior (Does)

Feels

Fears the "young and impetuous AI system," not the superintelligent one (2024) — and, by 2026, a US fiscal-political breakdown more than misalignment. Wants it to be true that a bridge person can still exist between Washington and Beijing; the "trillionaires vs war vs shared good" triad (TED, 2025-11) is about his own relevance as much as the world's.

Hears

Chinese lab and policy contacts in person (WAIC), Asia Society's China analysts (Triolo, Loke), Stanford Digital Economy Lab (Brynjolfsson's GDP-B and Canaries work, cited verbatim), the Quincy Institute restraint school, TED. Absent: the frontier-lab safety community and the DC China-hawk consensus he argues with.

Sees

The Chinese AI ecosystem from the floor — chip generations, robot volumes, hospital data pools, which labs have a business model — a vantage no other bench member currently has in person. And, from the HTC decade, what a hardware platform bet looks like when the market does not arrive on schedule.

Incentive map

Re-derived from scratch. The v1 conflict is gone; a different one replaced it. v1 discounted him as "an HTC executive arguing the AI-powered metaverse reshapes society." He no longer works for HTC and no longer argues that. What he sells now is the "wrong race" thesis to a US policy market where it is a differentiated, contrarian brand — fellowships, speaking slots and the Asia Society seat all price on his being the cross-Pacific bridge. He cannot say: that the CCP's AI regulation is primarily a control regime; that the US export-control lead is worth keeping; or that engagement has failed — each would collapse the brand and the access behind it. His China figures (3x electricity, 2–3 month gap, 34x cheaper models) are the strongest available for his argument and were not independently checked here. Unknown: which 100+ startups he has backed and how China-weighted they are — the incentive-map gap. Rule-25 discount: moderate on US–China framing and China-side numbers; low on his fiscal and displacement arithmetic, which cites third-party data (CBO, Brookings, Canaries) he could drop at no cost.

Theories aligned with

What he's reacting against

Where he overlaps / splits (with Rich)

Track record

Empirical vs normative

v1 read him as heavily normative — scenario work plus what institutions should build — and advised reading him for "what should we build," not "what happens next." Half-obsolete: his 2025–26 writing carries a real empirical load (energy capacity, capex, token-price deflation, entry-level payrolls, China's chip generations and robot volumes) on top of an unchanged normative core (UBII, automation levy, CERN/IAEA for AI, GI Bill, Marshall Plan). Read him for both — but check the China-side numbers, which are chosen to serve the thesis. Conflict note (CONVENTIONS rule 25): moderate; see incentive map.

Weak spots / open questions

Rich's take

(your synthesis here)

Delta log

2026-09-21 — v1→v2 migration + validation (bulk run)

Sources

migrated v1→v2 2026-09-21 · backup: _archives/alvin-wang-graylin.html.bak-20260921 · schema v1 added 2026-08-21 from a podcast guest list · AI & Society domain