Building powerful AI tools is only part of the equation; making them responsibly so they work as intended is the next frontier, and one that Sarah Bird is tackling.
Bird, who helped create ONNX — a kind of common language for AI programs so different tools can “speak” to each other — is now chief product officer of responsible AI at Microsoft. One of her team’s missions is to ensure that AI agents are not just trusted, but worthy of that trust.
“You need to give agents access to a lot of important things and the ability to take action, but you want them to stay focused and stay on task with what you told them to do and not allow them to do anything else,” Bird says. “Designing systems to enable us to do that is really, really fun and challenging.”
Bird’s work is shaped by a long-term view of innovation. She returned to Microsoft in 2019 after a stint at Facebook, now Meta — drawn back, she says, by Microsoft’s early and sustained investment in responsible AI, even when it was still considered an academic concern and customers weren’t paying much attention.
“Once the whole world woke up to why this was so important, we had been working on it the whole time, so we were really well positioned to start moving quickly when our customers needed us,” she says. “Microsoft’s leadership just absolutely gets that we really need to invest in all of the other dimensions of AI — security and safety and privacy.
“If we don’t get it right now,” Bird says, “we don’t get the huge, amazing, long-term future.”
She also credits the collaborative “One Microsoft” culture with making it possible to bring together deep layers of experts in fields as different as privacy, threat intelligence, identity, fairness and even biological systems to help solve problems that no single discipline could tackle alone.