Vivienne Ming

Vivienne Ming's core thesis, repeated across her SuperNova appearances, is disarmingly simple: technology has no moral content of its own. "AI is just a tool. It does good because you do good" ¹. This is not a throwaway line but the organizing idea behind everything else she says. She is openly hostile to the framing of "AI for good" as a category, and even more dismissive of companies that treat AI itself as the product. "There's a lot of AI companies out there whose product is AI. Again, that doesn't make any sense. AI does something — do something with it" ¹. For Ming, the tool metaphor cuts against both utopian and dystopian hype: a hammer builds a house or breaks a skull, and the algorithm is no different. The moral weight sits entirely with the person wielding it.

Her definition of what AI actually is in practice is unusually concrete and unglamorous. "Modern AI, it's any brief expert human judgment made cheaper, faster and better than a human can do it" ¹. This framing lets her make a specific, almost mundane prediction: expensive repetitive expert labor, the kind performed by analysts, lawyers, and doctors, will be displaced not by some singular superintelligence but by small, targeted software projects that automate one narrow judgment call at a time ¹. It's a deflationary view of AI that strips away the mysticism and replaces it with an economic argument, one she extends directly into the labor market. In her promo for the same conference she previewed the sharper edge of this idea: AI will transform the creative economy and generate jobs nobody has imagined yet, but "the problem is most of us won't be qualified for those jobs" . The optimism about new categories of work is paired, characteristically, with a blunt warning about the gap between the jobs created and the people available to fill them.

Ming's stories function as proof of concept for the tool argument, and she uses them to make a more specific point about what AI can and cannot substitute for. The example of autistic children wearing expression-recognition glasses is her strongest case: the technology didn't just hand them a prosthetic for reading faces, it taught them to read expressions themselves, and because the emotions showed up embedded in real social situations rather than isolated on flashcards, they also absorbed empathy as a byproduct ¹. This is the shape of her argument throughout, that a well-designed tool aimed at a real, specific deficit can produce learning and capability that outlast the tool itself. She contrasts this sharply with the CIA-adjacent competence-rating AI that "didn't learn competence, it learned that a 'competent face' in America is simply an old white guy" and consequently failed to predict how well the CEOs it rated actually ran their companies ¹. The two stories together form her working test for whether an AI application is legitimate: does it target a real, well-understood problem, or does it launder a bad question through a black box and produce a confident, wrong answer.

This is also where she draws her firmest boundary. "AI cannot solve problems you don't know how to solve," and she is scornful of anyone who thinks you can throw data at an algorithm and expect a perfect solution to materialize, calling that belief a recipe for making the world worse ¹. Her diabetes-hacking and refugee-reunification examples serve the opposite lesson, that if you already understand a problem well enough, AI can be the lever that makes solving it economically feasible at scale. "If you know how to fix something that's seriously wrong with this world, you can build a tool that can actually change people's lives" ¹. She frames many of the world's hardest problems as persisting not because they're unsolvable but because there simply aren't enough experts to go around, and AI's real contribution is changing the economics of expertise so that scarce judgment becomes abundant ¹.

There's a recurring skepticism in her talk about where "innovation" naturally drifts without deliberate direction. "You get the end result of all innovation anywhere on the internet: it animates cats," she says, a wry way of pointing out that clever technology left to market incentives alone tends toward the trivial ¹. This is the flip side of her opening claim: tools don't aim themselves, so if you want AI to matter, someone has to point it at something that matters, and do so on purpose, now, rather than waiting for permission or resources. "Don't wait, don't wait till you're rich... go out and plant a tree" ¹. She ties this directly to her research interest in purpose as a predictor of good life outcomes, arguing it can be measured through evidence of sacrifice for something larger than oneself, "when old men plant trees" ¹.

Taken together, her position across these sources is consistent rather than evolving: she arrived at SuperNova already committed to a demystified, tool-based view of AI, used it to preview a labor-market warning about the creative economy , and then spent the keynote itself building out the argument with concrete cases, both cautionary and inspiring ¹. The practical takeaway she leaves audiences with is a kind of test they can apply to any AI project: does it do something specific for a real, understood problem, does it teach or expand human capability rather than just replace it, and is someone treating it as a means to a chosen good rather than as an end or a product in itself. Her sharpest one-liners, that AI is a hammer, that a company selling AI as its product makes no sense, that innovation without direction just animates cats, all function as compact refusals of hype, aimed at forcing the listener back onto the actual human purpose behind the technology.