Bypassing Systems with Gary Bradski [Idea Machines #9]

12/30/18 • 58 min

In this episode I talk to Gary Bradski about the creation of OpenCV, Willow Garage, and how to get around institutional roadblocks.

Gary is perhaps best known as the creator of OpenCV - an open source tool that has touched almost every application that involves computer vision - from cat-identifying AI, to strawberry-picking robots, to augmented reality. Gary has been part of Intel Research, Stanford (where he worked on Stanley, the self driving car that won the first DARPA grand challenge), Magic Leap, and started his own Startups. On top of that Gary was early at Willow Garage - a private research lab that produced two huge innovations in robotics: The open source robot operating system and the pr2 robot. Gary has a track record of seeing potential in technologies long before they appear on the hype radar - everything from neural networks to computer vision to self-driving cars.

Key Takeaways

Aligning incentives inside of organizations is both essential and hard for innovation. Organizations are incentivized to focus on current product lines instead of Schumpeterian long shots. Gary basically had to do incentive gymnastics to get OpenCV to exist.
In research organization there's an inherent tension between pressure to produce and exploration. I love Gary's idea of a slowly decreasing salary.
Ambitious projects are still totally dependent on a champion. At the end of the day, it means that every ambitious project has a single point of failure. I wonder if there's a way to change that.

Notes

Gary on Twitter

The Embedded Vision Alliance

Video of Stanley winning the DARPA Grand Challenge

A short history of Willow Garage

In this episode I talk to Gary Bradski about the creation of OpenCV, Willow Garage, and how to get around institutional roadblocks.

Key Takeaways

Aligning incentives inside of organizations is both essential and hard for innovation. Organizations are incentivized to focus on current product lines instead of Schumpeterian long shots. Gary basically had to do incentive gymnastics to get OpenCV to exist.
In research organization there's an inherent tension between pressure to produce and exploration. I love Gary's idea of a slowly decreasing salary.
Ambitious projects are still totally dependent on a champion. At the end of the day, it means that every ambitious project has a single point of failure. I wonder if there's a way to change that.

Notes

Gary on Twitter

The Embedded Vision Alliance

Video of Stanley winning the DARPA Grand Challenge

A short history of Willow Garage

Previous Episode

Accelerating Biotech with Jun Axup [Idea Machines #5]

Link to this Episode in Overcast

In this episode I talk to Jun Axup about accelerating biotechnology, how to transition people and technology from academia to startups, the intersection of silicon valley and biology, and biology research in general.

Jun is a partner at IndieBio - a startup accelerator specializing in quickly taking biotechnology from academic research to products. She has both started companies and did a PhD focused on using antibodies to fight cancer. This experience gives her a deep understanding of the constraints in both the world of academia and equity-funded startups and what it takes to jump the gap between the two.

Key takeaways:

Biology is reaching a cusp where we can truly start to use it to do things outside the realm of traditional medicine and therapeutics. These new products fit more cleanly into the silicon valley startup ecosystem.
The gap between research and products in people's hands is not just a technical gap, but a people one as well. Indiebio is built to address both - guiding both the research and the researchER out of the lab.
While the capital overhead has come down, biology-based innovation still require different support systems than your standard computer-based innovations.

Links

Jun's Homepage

IndieBio

Flight from Science

Langer Lab Case Study (Paywalled)

No transcript this week - trying a different production flow. If you feel strongly, please let us know at [email protected].

Next Episode

Hacking Politics with Craig Montouri [Idea Machines #8]

In this episode I talk to Craig Montouri about nonprofits and politics. Specifically their constraints and possibilities for enabling innovations.

Craig is the executive director at Global EIR - a nonprofit focused on connecting non-U.S. founders with universities so that they can get the visas they need to build their companies in America. Craig's perspective is fascinating because contrary to the common wisdom that innovation happens by doing an end run around politics, he focuses on enabling innovations through the political system. It's eye opening conversation about two worlds I knew little about so I hope you enjoy this conversation with Craig Montouri.

Key Takeaways:

There is a lot of valuable human capital and knowledge left on the table both by the US immigration system and the university tech transfer system.
Nonprofits need to find product-market just as much as for-profit companies making products. And just like the world of products, there's often a big difference between what people say their problems are and what their problems actually are.
Political innovation is different than other domains for several reasons - it both has shorter and longer timelines than other domains and in contrast to the world of startups, politics needs to focus on downside mitigation instead of maximizing upside.

Resources

Global EIR

Craig on Twitter(@craig_montouri)

NPR piece on Global EIR