Timnit Gebru warns AI science is corporate-run

safeguard independent – Computer scientist Timnit Gebru says American science—especially in AI—has become corporate-driven and sloppy, pointing to practices she calls basic but widely ignored. She argues incentives must move away from corporations and the military toward scientific c

In the AI world, Timnit Gebru says the danger doesn’t start with bad ideas. It starts with the everyday shortcuts companies choose to take.

Gebru. a computer scientist and engineer who founded the Distributed AI Research Institute. describes a field in which she’s seen basic safeguards routinely disregarded. “In my field. ” she says. “I would describe it as corporate-driven and sloppy.” She points to practices such as not testing on your training data. failing to provide transparency. and not keeping appropriate documentation—mistakes she argues can make results not just flawed. but misleading.

The concern runs deeper for Gebru because of what she believes powers the work. In her view, the incentives guiding American science are misaligned with what science is meant to serve. “Number one, the incentive needs to be decoupled from corporations or the military,” she says. Funding, she argues, should come from elsewhere—guided primarily by scientific curiosity, but also by people’s well-being. “That’s the number-one thing. ” she adds. before stopping short of trying to dress it up as anything less fundamental: “That’s where I’ll start and stop.”.

Her most forceful answers aren’t abstract. She’s quick to name where she’s finding reasons to keep pushing anyway: small organizations and younger people building tools despite limited resources.

Gebru says she’s constantly inspired by a New Zealand–based company called Te Hiku Media. Using a small budget. she says. it builds tools for language revitalization “and a whole bunch of other things.” She and her team have tried to adopt some of the company’s practices. hoping other people will follow suit. For her, the key lesson is speed—how ideas can spread when small groups talk to each other. “I’m inspired by how fast movements can spread like this. ” she says. “with small organizations talking to and inspiring one another.”.

One detail stands out: Te Hiku Media builds its own cluster—a setup she describes as “kind of like a small data center”—instead of relying on Google and Amazon and other cloud services. That choice, in Gebru’s telling, is proof that constraints can force innovation rather than kill it.

She describes how quickly discouragement can set in when people fixate on the amount of money and resources they think science demands. “You can get very discouraged thinking about the amount of money and resources it takes to do science,” she says. But when you see what small organizations can do with few resources. she argues. it becomes possible to pursue a specific curiosity even with limited means.

For early-career scientists, Gebru’s advice is equally direct. “Question everything,” she says, and question the source of the information you receive. She warns against believing “any kind of hype that you hear from companies.” The only safe approach. she argues. is persistence: follow the rabbit holes. ask questions. and don’t let the first story you hear become the last one you accept.

Her definition of being a scientist is almost stubborn in its simplicity: “Being a scientist is really asking why and how. ” she says. “Is this really true?” She emphasizes patience—waiting long enough to keep asking questions—rather than settling for answers from the media. She also urges scientists to question incentive structures and where information is coming from. and to keep going down rabbit holes during experiments. even when the process becomes hard. “You have to struggle,” she says.

Gebru’s comments come from a life spent building in environments where questions and community matter. She is the founder and executive director of the Distributed AI Research Institute. an independent organization conducting community-rooted research into technological tools and systems. She is also co-founder of Black in AI. a nonprofit focused on increasing the presence. inclusion. visibility. and health of Black people in the field of artificial intelligence. In addition. she serves as treasurer for AddisCoder. a nonprofit dedicated to teaching algorithms and computer programming to Ethiopian high school students.

In the past few years, she says, the field itself has changed—fast, and in a way that troubles her. She says that AI. and the wider computer science world. has moved from the margins into the center of public attention. For her, that wasn’t the problem at first. She describes being drawn to computer science by the thrill of asking questions about efficiency and resource use—what’s the fastest way to do something. how to do it with fewer resources. how little memory can be used. She talks about a shift from brute-force approaches to “the clever way. ” where algorithms come in. and says that puzzle-solving energy was what made the work exciting.

But in her view, corporate-driven priorities have pushed the field toward something less compelling. If researchers pretend they have “all the data in the world. ” “all the resources in the world. ” and every chip they could ever get. she says. the project becomes less about discovery and more about proving what can be done under unrealistic assumptions. “That’s not exciting to me,” she says, arguing that “you cannot innovate that way.”.

After laying out what she believes is broken—scientific incentives tied to corporations and the military, and AI practice that she describes as insufficiently tested—Gebru leaves readers with a message that lands like a call to protect the integrity of science before it becomes too easy to misuse.

The edited interview includes a request for support for Scientific American. framed as a moment that requires readers to “stand up and show why science matters.” It notes that Scientific American has served as an advocate for science and industry for 180 years. and that the publication says this is a critical time in its two-century history. The message also describes the kinds of content a subscription supports—news coverage of decisions that threaten labs across the U.S. podcasts. infographics. newsletters. videos. games. and writing and reporting.

Gebru’s argument is not that science should be romantic or untouched by power. It’s that the incentives now steering high-profile work have begun to corrode the basic discipline of testing. documentation. and transparency—and that the field will only regain its edge when it starts rewarding curiosity and well-being again.

She finds encouragement where she looks for it: in small organizations building their own infrastructure, in younger researchers willing to keep asking why, and in anyone brave enough to keep following the rabbit holes long after the hype moves on.

Timnit Gebru Distributed AI Research Institute Black in AI AddisCoder AI research scientific incentives science journalism transparency documentation training data testing language revitalization Te Hiku Media independent science