Who's Missing from Trump's AI Superintelligence Lunch? Three Power Gaps Shaping the Industry

The author, a software testing veteran, examines a widely circulated photo of a White House East Room lunch on September 29 — dubbed the "superintelligence lunch" — where Trump, Vice President Vance, Speaker Johnson, and cabinet members sat across from AI industry leaders including Huang Renxun (Nvidia), Elon Musk, Mark Zuckerberg, Jeff Bezos, Sundar Pichai, Satya Nadella, and the founders of OpenAI and Anthropic.

Landscape One: Promises vs. Verification

Official statements claim companies committed to safety assessments, independent audits, layered controls, and intrusion prevention. Trump called these "morally binding commitments"; Speaker Johnson described them as voluntary principles. When pressed, Anthropic's Dario Amodei admitted the rules are "still under discussion." The author notes this phrase is familiar to engineers: "Unverifiable promises equal no promises." In software engineering, a claim of safety requires test reports; data compliance requires audit logs. An agreement without acceptance criteria or review mechanisms is merely a draft requirements document, not a deployed system. The risk is not necessarily bad faith, but that "we'll self-regulate" becomes the only constraint, quietly shifting risk onto unaware users.

Landscape Two: Who Profits, Who Pays

Trump praised data centers as "very positive" and urged communities not to oppose them, promising local benefits. The author, drawing on experience reviewing local informatization projects, points out infrastructure costs extend far beyond construction: electricity from the grid, new power plants, upgraded lines, land, water, tax breaks — all public resources. A typical mismatch emerges: gains concentrate among a few shareholders and large users, while costs spread across all ratepayers. "You may never have touched a large model, but your electricity bill may already have paid for it." Trump's argument that forcing firms overseas would lose the industry is understood, but keeping factories domestic doesn't eliminate costs — it just relocates them onto ordinary people served by that grid. The author warns against "local optimum" thinking: the table's occupants feast while the bill-payers remain unaware.

Landscape Three: Rule-Makers and Rule-Takers

Gathering the entire industry apex to discuss safety standards is necessary, but standards also function as barriers. "Whoever meets the standard first gains first-mover advantage; whoever can afford the assessment cost stays in the room." Small firms are locked out, concentrating voice further in giants, and regulation risks becoming a competitive tool designed to favor incumbents. Historical parallels: post-1998 financial crisis regulation was shaped by large financial institutions; early internet standards were set by the strongest vendors. The author concludes: "'How safe is safe enough' sounds technical, but at its core is a power issue. Rules defined at whose table define interests from whose perspective. If your representative isn't at that table, your concerns only get 'incidental consideration.'"

On the Name "Superintelligence"

Trump prefers "superintelligence" over "artificial intelligence" — a savvy framing that rallies more ambition than a neutral term. But researchers must clarify: "The name is a posture, not a solution." Whatever the label, data still leaks, jobs still get displaced, electricity prices still rise. Users gain no extra safety from a rebrand.

In the End

The photo is strikingly quiet: no arguments, no drama, just a lunch, a roomful of people, a few anodyne promises. That quietness is worth pondering — AI's direction, standards, and cost allocation are being settled in an unremarkable afternoon. "The landscape is often not announced; it is defaulted into being." The author, drawing on decades of software engineering, is not overly worried about AI "running away" — that's an engineering problem solvable through verification, testing, iteration. The deeper concern: "When all this is decided at a small table, who speaks for ordinary people?" That question applies not just to this lunch but to the author's own role. "Seeing clearly isn't for commenting from on high; it's for knowing where to apply pressure — can academia build an independent AI evaluation system instead of ceding safety standards entirely to others? Can we educators embed engineering ethics and algorithmic power into required curricula so the next generation of engineers enters the field with power awareness? The missing chair at that table must be brought over by someone. If no one sits in it yet, this lunch is only the story's beginning."