Helping (Robotic) Hands

Startmate applications close on September 30 and at Blackbird we’re here all the time, so drop us a line.

Driverless cars, drones that deliver food and machines that work the warehouse are all robots. Robots are exciting because they let the technology industry solve the real world’s biggest problems not just the ones stuck inside a screen.

Robotics in 2016 is intriguing not only because the smartphone wars have decimated the price of electrical components but large advances are being made in machine vision (how can robots see the world around them) and deep learning (after figuring out the world, what decision and action should they take).

At Blackbird and Startmate we’ve invested in Zoox (autonomous transportation), Baraja (Lidar), Flirtey (Drone Delivery) and Propeller (3D drone mapping). Here are some more ideas that excite me, so do get in touch (@nikiscevak) if you’re thinking through them:

1. Mapping data networks: Many problems can be better solved through data network collaboration. For instance in a completely different industry, Sift Science aggregates online fraud data and helps merchants make decisions on whether to accept payments. Can you aggregate, label and structure real world map data for driverless cars and indeed any other autonomous robots to help them make better decisions? It’s very fashionable to say that Google and Facebook have all the data but unless it is correctly labeled and classified then it’s useless data. See a glimpse of what might be possible with what Comma.ai did recently with releasing 7 hours of highway footage and times that by a billion.
2. Better Eyes: Great advances are being made in Lidar technology (the sensor you see on top of the Google Koala car) to dramatically bring down the cost of allowing robots to see with precision. Using better software algorithms can you use cheap smartphone cameras to bring the fidelity of 3d vision dramatically up? Can you take cheap sensors like Xbox Connect to another level down the cost curve? Can the sensors be really low power so little drones can use them too?
3. *Simulated worlds:** Testing the real world is hard. Google has driven just 1.5 million miles with its self-driving cars when ideally you’d like billions. Can you build a simulated testing environment that allows you to take real world snippets and recombine them into monte carlo simulations and then verify that those virtual tests are correctly calibrated to the real world? The network data approach here also allows greater possibilities. Improbable in London is a good generic example.
4. Poop-scooping shitty unit economic on-demand startups: Can you take sharing economy businesses that have bad unit economics like cleaning, massages, food delivery and haircuts and replace the humans with robots? Examples: Picking apples, making coffee and delivering pizzas.

Fire up!