Jimi Crawford: Another use case that we’ve been working on is simply counting cars. It seems simple. It’s not actually all that simple because the satellite images have relatively low resolution—about a meter or so per pixel. There’s not very many pixels. We’ve been able to train to actually count cars accurately in parking lots, roads, and bridges. One of the things we can do is track how many people are going to shop in all the different retailers in the US so we can give you an idea at the end of the Christmas season as to how many people were shopping versus the year before.
We can look across the US and see major effects of cold weather in the northeast. We can also get an idea, and we’ve been tracking this the last couple of months, whether the US economy is slowing down or speeding up. There are other applications as well. For instance, we’ve heard from a lot of insurance companies we’ve talked to that car density is one of the biggest and most important determinants of car accident rates.
We’re starting work to assess trends in car densities in major cities around the world as an input to major insurance companies and how they set their rates around the world. This simple problem of counting cars turns out to be applicable to quite a number of different industries when you can do it at scale across countries and cities around the world.
Sramana Mitra: Where are you gaining traction in terms of your customer base? Who is buying this kind of solution?
Jimi Crawford: The three areas where we’re doing the most right now is investors and asset managers. There’s a lot of asset management companies out there that are managing billions of dollars in assets. They’re interested in all kinds of things. If they were investing in Walmart, they want to know what’s going on with the car counts. If they’re investing in construction companies doing business in China, they want to know how fast the city is growing.
If they’re investing in industries that are sensitive to international trade, they want to know how many ships are in all the ports and how many containers they are offloading. The second big vertical for us is government. All of these questions about food security, international development, how many cars are crossing the borders, and port activity are interesting to quite a few different government agencies. The third area is insurance. For crop insurance, measuring corn yield and soybean yield.
Sramana Mitra: If you were to look at your industry in general, talk about some open problems that could be solved by a different entrepreneur working with a different framework, but applying some of the AI principles that perhaps you are familiar with. What are some possibilities?
Jimi Crawford: I think there are several interesting possibilities. Deep learning, as a trend, is one of the most significant trends in artificial intelligence right now. That’s just because in 2015, for the first time ever, computers got better in object recognition than humans. You give a computer a picture and say, “Is this a tree?”. Computers can now do that object classification more accurately than humans, which is amazing. This is something that’s been a goal in the field for decades.
That has all kinds of implications. It has implications for self-driving cars to figure out what it is they’re seeing—whether it’s a tree or a person that will most likely move. It also has applications for processing imagery out of cellphones and images that people are uploading to social media sites. Imagery that comes out of security cameras or imagery that comes out of other fixed cameras. I think there will be a whole plethora of startups generated on taking deep learning to process imagery instantaneously.
I think we’ve just scratched the surface of what can be done with that. Similarly for speech understanding, there have been vast improvements over the last few years in the ability to take human speech and figure out what the words are. I think we’re only beginning to see the possibilities of how that could be applied in various business opportunities.
Sramana Mitra: Very good. Thank you for your time.
This segment is part 2 in the series : Thought Leaders in Artificial Intelligence: Jimi Crawford, CEO of Orbital Insights