DAPPCON 2019: Decentralized Earth Observation - Julien Bouteloup (Flyingcarpet)

Transcript

so flying carpet we've been building flying carpet over the past two years and so flying Kaba is a prove observation is design-wise there for earth observations so we start building using drones that was like a year and a half ago we actually want to Papua New Guinea and we had a drone flying for analyzing the prediction of coconuts and coffee and I was like running in on a smart contract and then we completely shift to using satellites so basically I'm gonna go through all the different products that we've been building in

flying carpet and and then focus on something that we call the prediction market for physical assets so we have like partnership across the biggest satellite providers so what we do is basically we build machine learning models that fetch information from satellites and then we can use it as an Oracle so we can actually calculate the life prediction of coffee in Brazil or we can do like analysis of a rooftop and then what you can do you can actually plug this into any physical any prediction market for physical community we build them this is actually everything is live so you can test it but for example we build the first model

request so people could it was like a year and a half ago so you could actually pass the bounty to get someone using a drone for example in middle of London to go and then select something as a task and people could actually go and then make money so that was our first prove our concept using drones for Earth Observation and then we also build the first Mechanical Turk for annotating images for machine learning training this is also life we can test it it's kind of like AWS machine learning and then we focused on creating machine learning on different different assets so for example you can calculate the

production of coffee in Brazil and we did partnership with one the biggest Kofi trader so then you can actually sell so you you have the smart contract you have the the machine learning that basically get hash into a smart contract and then the machine learning we extract information using satellite images in in Brazil and then we use it for making bet on the coffee market coffee market is one the biggest market for a community we we've been we sent someone from so v1 flying carpet we want Microsoft and National Geographic AI progress which is actually the biggest competition in machine learning and we actually we focus on

climate change so we did like cackle biomass global warming so now we have someone in Antarctica is actually analyzing the the melting speed of the melting ice over there and then I will explain how it's connected to blockchain all the stuff so we also release the proof of observation sticking mechanism where you can actually run this is on Twitter you can go and you can see live if I know Internet let's see for Internet but actually gives a good speed okay so this is something we did last

last year you we use this to fetch information and double-check information that we were getting from satellite using Google Maps or searching information and you could plug this into an Oracle that was giving you output for physical specific physical asset in this case is actually cut tackle so you can make like 100 hundred died under $20 just by running this sticker this yes so then we we partner with a company

that so uh for this presentation I will focus on carbon carbon offset carbon credits carbon credits is actually one the biggest market we've and coconuts and cacao for a community trading so in carbon offset what you can do you can actually have like shin link for example you can capture data from space knowing the concentration of carbon knowing the concentration of biomass in a specific forest and you can basically tokenize the specific absorption of carbons in a per square meter in the forest and then what you can do you can actually build a prediction market on top of it you can ask people can ask question on the market saying I predict that it because the concentration of carbon in specie forest will be higher than last year and you can also sell this credit to

specific company so we have backed by the European Commission well basically the idea is like they buy carbon credits from flying carpet using the this company called wafer is shown in a coaching I will show you this is actual online it's called echo tree so you can actually buy trees that are tokenized like tokenize like mother absorption of carbon is tokenized and so what we did we got sponsored by nosies for running a prediction market on the top of carbon credits on the top of like any physical asset you can actually make pretty prediction so for example you can say I believe that the speed of the melting ice will be faster in Antarctica in

Greenland then last year you can actually make a bet and what we do to resolve this bet we basically trigger using channel link which figure and a vehicle that is basically as a machine learning algorithm that is running on Isaac cloud computing system or they could be running an Isaac - now it's like running just on s3 but basically then it gives the output and can resolved the market I will show you demo so yeah that's what we have we have like different machine learning models like anyone can actually push any machine learning models inside the system is like kind of like an MRI but we don't actually run competition like anyone is welcome to push the model as long as the model feet specific data

set you can resolve a specific that I set but we are like extraction of data of Shane and Shane we have the nosiest mercury smart contract for the prediction market so I'm gonna show you so if you go for this one if you go this is actually local oz because we it's faster but if you go to p.m. so you need some link token and also like some dye we we have like a set for that it's on test met some we could buy and then yeah so then you have the market I'm gonna go back to localize for fast it's gonna be faster so you see for

example this one they are not the same questions but someone actually create a market and this data is being fetched from Shane link so it's actually real data and it's backed by European Commission because the data is using eco tree eco tree is a company that has a piece of forest in in Europe and they have plenty of trees and they know exactly what is the concentration of the forest of the of the the carbon offset of the carbon so then actually you can actually fetch its information using chain link which is an Oracle and then you can buy a few different market for example like wool more than 30 metric tonne of co2 being absorb in the specific flows this is actually in French but this is a GPS location so you can actually and then you can trade the

market so if for example you say this is using nosies LSM are smart phones right you have all information if you want to run your own prediction market you want to fetch this and you can actually create any market for physical community using channel and using a noisy sunshine link you can actually go to get up flying capitulation market and we explained how to run your own flying carpet prediction market or you can actually make a bet everything is described so you need together we also send you some some tokens linking dye to make to make a few bets so for example what happen in this case if I if I think that will not happen so I select any way I can

I can select how much I'm playing Tibet and they will tell you the percentage you receive if you win to that then you can approve then you've got a meta mask this is actually as this one is on being kadai but then approves on the market that was created nosies and so what happened is like on the date of this one the by January 20 2020 what will happen is like the smart controller would trigger an Oracle that this article as a link to the machine learning model that

can fetch information about the specific forest and also the certificate that was issued by the European Commission so you can actually correlate the data between those two different systems and so this is a proof ok so then you close and normally if there was like no bug here you could see it would be the percentage of succeeding or failing and then you can go to your positions there's no all positions generally gives them an equation first that making sense or questions so then you can you can also like for example what we do like we

fetch more than just one market so you can actually go to you will fetch from a specific company like still using a co tree and you can actually go in and create a market and you can specified all the information for the specific market [Music] yeah I think it's a I just wanted to make this demo like more focused in the prediction market side of proof observations and you can actually render the the stuff online if you go to the github or flying carpet you guys have any questions or questions okay I think it's time

[Applause]