Crowdsourced talent, artificial intelligence, and drone racing — AlphaPilot has a unique combination of ways to promote innovation.
Lockheed Martin, HeroX, and the Drone Racing League have partnered on the AlphaPilot AI Drone Racing Innovation Challenge, which the organizers claimed is the first large-scale open innovation challenge of its kind. The third race in the World Challenge will continue tomorrow in Baltimore, kicking off the Artificial Intelligence Robotic Racing Circuit.
HeroX is a spinoff of XPRIZE co-founded by Peter Diamandis, Christian Cotichini, and Emily Fowler. It offers cash to individual problem solvers in a “Crowdsourcing 2.0” model.
Six years ago, serial entrepreneur Cotichini had sold his fourth company to Dell when he met Diamandis. “We started collaborating on the idea of democratizing the XPRIZE model, which passed the ‘shower test,'” he told The Robot Report. “We wanted to make the power of the crowd available to everyone and to make a global labor market for the knowledge economy.”
“We’re really proud of what we’ve accomplished with minimal capital,” he said. “We’re ready to scale and are looking to be the Airbnb of human ingenuity.”
He and Kyla Jeffrey, leader of the HeroX customer success team, answered the following questions:
Competition for technical talent is particularly tight right now — how do challenges such as AlphaPilot address that need?
Cotichini: Crowdsourcing is a medium, a labor model. There are many different use cases. Lockheed Martin found it was a good approach for them, robotics, and AI.
With more than $2 million in prizes, this is the Olympics of crowdsourcing, acting as a spotlight on the development of autonomous drones.
This is also a platform for learning and developing skills to contribute to the global talent pool and Lockheed Martin’s network of innovators. It will change the lives of hundreds of people.
What makes HeroX and AlphaPilot different from other challenges?
Cotichini: HeroX is the only general-purpose crowdsourcing platform that has hit enough scale to do large volumes of projects like this. By “general purpose,” we mean that any company can send out for any problem that’s ethical, legal, and safe.
Horizontal positioning is important to us — it creates a true capacity for organizations to try a “boil the ocean” strategy that incubators tell you not to do.
Competitors are intrinsically motivated by the vision of the challenge rather than just the monetary prize and are going for audacious breakthroughs. HeroX has hundreds of teams, and the power of crowdsourcing talent is that they’re hiring the project, not the project hiring them.
AlphaPilot is at the finalist phase with nine teams, but 2,600 solvers signed up. Lockheed Martin gets use cases for solutions.
How does HeroX handle so many potential solvers?
Jeffrey: It’s how the projects are set up. In this case, Lockheed Martin set up three sequential tasks for a quantitative assessment by a panel of judges aggregated from experts in the field.
We set criteria so that submissions would be easy to judge. In AlpaPilot, Phase 2 was a vision test, and teams had to identify gates with machine vision algorithms. The third task was building a model to fly in simulation at MIT’s AI lab.
Cotichini: This follows a lot of best practices we’d apply to crowdsourcing. We have separate qualitative and quantitative or subjective and objective criteria, and we work with clients to remove biases. Rather than screen for skill sets, we look for results.
With our criteria, we have an automated scorecard. Using an API [application programming interface], the simulation generated scores and a leaderboard for Phases 2 and 3 of AlphaPilot.
We’re really looking to provide gamification as part of our toolkit. With a World of Warcraft model, solvers can see how to iterate and how they’re doing.
What hardware is AlphaPilot using? How much customization do teams do?
Cotichini: Lockheed Martin provided the teams that move on in the drone races with NVIDIA Xavier GPUs as well as drones meeting the specifications of the Drone Racing League.
On the software and AI side, what’s the baseline? Is code reviewed, or just drone performance?
Cotichini: During the qualifying tests this spring there was a review of the teams’ frameworks to see if their code was elegant and scalable, but the three tasks heavily weighted toward quantitative results.
Lockheed Martin wanted to make sure the software met the criteria for the challenge and didn’t pose any intellectual property risks.
For the World Challenge, the Drone Racing League has the perfect test bed for testing AI pilots. It’s egalitarian and transparent.
Are the milestones visible to the public?
Is mentoring or other support available to the teams?
Jeffrey: HeroX and Lockheed Martin offered webinars early on in which participants could ask questions live.
There are nine finalist teams left competing in the race, with about 70 people total. Outside of looking for mentors, they’re also looking for people to work with. Some runner-up teams have merged with the top teams.
At HeroX, we’re pleased to see collaboration and bringing in outside parties.
How have the organizers collaborated on AlphaPilot?
Jeffrey: The team at MIT played a critical role for simulation and testing. Everyone is committed to the end goal of improving drone autonomy.
Cotichini: I had a background in B2B startups, and when I first started with crowdsourcing, I had anxiety at first. I’ve learned a lot in the past five years. When you give people a chance to participate, don’t micromanage it — there’s a fundamental human desire to compete. It’s not about winning; it’s about inclusion, which is powerful.
I’m also amazed to learn that we’re wired to collaborate; we’re social animals. We see that desire a lot in these crowdsourcing projects, bringing talent to our customers. That’s the egalitarian nature of HeroX.
How important is machine learning to autonomous drone operations, and what level are you looking for? Is this applicable to all potential drone applications?
Cotichini: There are two prize levels. One is for the AI that beats all other AIs. Just finishing a race would be amazing. The fastest team will get $1 million.
The other prize is a stretch goal — if any AI beats professional human racers, its team can win an additional $250,000. We don’t really expect this to happen this year.
Drones must be completely autonomous. One of the reasons that Lockheed Martin went to the crowd is that AI is very much focused on cloud computing and unlimited electrical power. The question was how to start with the idea of an AI pilot running on milliwatts and letting innovators come up with solutions that it’s not going to get from experts.
Will you circle back with promising advances by runners-up?
Cotichini: We’re looking at HeroX as a new labor model, and we’ll see how hard it is to win this competition. You can’t have innovation without massive amounts of failure.
But in a knowledge economy, a lot of things are more routine. Jobs are still scored, but with more of a reputational scoring, like in the gig economy.
We’re on a power curve, where on the high end is the breakthrough $1 million crowdsourcing, then a long tail of gigs. We want to give HeroX solvers the opportunity to participate and engage in the overall ecosystem.
The aspiration of the platform is solvers as heroes. HeroX lets them follow a “hero’s journey” through a career path, expanding from similar knowledge work through smaller prizes.
Beyond the monetary prize, does Lockheed Martin gain access to the advances of AlphaPilot’s winning teams, or can they develop it further and commercialize it themselves?
Cotichini: The intellectual property stays with the team. Lockheed Martin will ultimately work with the winner and can negotiate to acquire the technology.
It’s a great way to find talent to contract with or to hire. $1 million is not a lot for AI or drone development. It’s a good bang for your buck to get talent and build a positive brand.
What’s the proportion of large to small prizes at HeroX?
Cotichini: There are three major brackets. Every challenge is personalized and unique, but the high end is more than $200,000. This is different from XPRIZE, where the high-end prizes are in the millions.
At the midrange are more technical solutions and larger ideation challenges, such as how to inspect oil storage tanks. We did a challenge for a venture capital firm that works with limited partners to find specific solutions to invest in, doing a deal flow with crowdsourcing. It had a problem one of its investors wanted to solve, and they used HeroX to find someone with a solution.
On the low end is ideation, picking logos, writing content — smaller tasks. Our platform is like a tool kit to build multiphase challenges with reusable templates. We’re working with consulting companies and other crowd vendors so they can provide expertise and use our network for publishing challenges for crowdsourcing and innovation.
The Robot Report is launching the Healthcare Robotics Engineering Forum, which will be on Dec. 9-10 in Santa Clara, Calif. The conference and expo will focus on improving the design, development, and manufacture of next-generation healthcare robots. Learn more about the Healthcare Robotics Engineering Forum, and registration is now open.
How does crowdsourcing talent at HeroX change the model for innovation?
Cotichini: We look at the demand side versus supply side, which is different from every other crowdsourcing model. Others mostly recruit a crowd with specific expertise and then sell it to companies. We don’t do that — 80% of 2,600 solvers we recruited for AlphaPilot were off our platform, not in the HeroX community. There are 900 million highly educated potential solvers accessible on the Internet — that’s an overwhelming talent pool to target. More than half of college graduates get jobs outside their field of expertise.
Jeffrey: I have a degree in marine biology. Like a high school diploma, getting a college degree is a solid choice. You choose the field you’re interested in, and it’s factory-installed, like an app on an iPhone. You can always add more apps or skills.
With specialization, degrees are not seen as that important. If you want to learn engineering, there are resources online that weren’t available 20 years ago. Anyone can become an expert in something.
Crowdsourcing gives the opportunity to prove your talent and genius. We’re birthing new careers from these projects that people can put on their resumes — “I was a finalist in the Lockheed Martin AlphaPilot challenge.”
Millennials get that they’re in charge of their own education, learning on demand. We can never stop learning, especially in technical professions.
Part of our mission at HeroX is creating job liquidity, moving to an optimum we’ve already created with the Internet and the knowledge economy. We want to help make that global market. Once employers have to compete for talent rather than talent competing for jobs, we’ll start to see balance in the income disparity, which is a current existential threat to economies and politics.
Like social marketing, crowd enablement will eventually be a standard tool because of competitive forces — that’s the beauty of the free market. We’re also adding to academia and corporate labs what’s possible for innovation.