Exclusive Interview With Founders of Ankr Network

Just recently we came across new blockchain project called Ankr Network which is a decentralized artificial intelligence computing platform. Ankr is distributed Cloud Computing on Trusted Hardware. Cloud Computing is projected to be a trillion dollar market, yet it is monopolized by some of the largest tech conglomerates in the world.

Ankr strives to build a resource efficient blockchain framework that truly enables Distributed Cloud Computing (DCC) and provides user-friendly infrastucture for business applications.

We interviewed Ankr founders to learn more about them and the project.

Tell us a little bit about yourself. How did you get started in blockchain?

Chandler: Back in 2014, Ryan and I were college roommates at UC Berkeley. I was studying EECS, and got interested in Bitcoin, which led me to join what later became Blockchain at Berkeley. I tried to sink any extra money I had lying around into Bitcoin.

I started doing more of my own research, talking to friends at different schools who were also interested in this field, and becoming more actively involved with Blockchain at Berkeley. I had a close friend who was a PhD at the time, and she got me started on this idea of trusted hardware. In August of last year, I mentioned to Ryan I wanted to build something around the combination of this technology but I don't think Ryan thought too much of it at the time.

Stanley: I was Chandler's super manager while he was an intern at Amazon AWS. My background is in cloud computing and I was a lead engineer with the first round of EC2 instances. In my time at Amazon i also contributed to Alexa Rankings. While Chandler was here, we had previously talked about his concerns over the high margin and high barrier to entry to run an EC2 instance. Chandler and I had also talked about his interest in trusted hardware and blockchain, so when Chandler came to me later to discuss this project, it followed naturally that we would strike up a new working relationship.

Ryan: For me, I was a Business and Statistics major at Berkeley. Chandler was the third person at Berkeley I met, and he was the one who convinced me to buy Bitcoin with the money I had been saving for a car. I didn't really have time to learn to drive at the time anyway. Together we bought around 22 Bitcoin at around 500 dollars apiece.

When Chandler started working at Amazon, we talked a lot about how AWS was one of the only ones to provide the kind of service they did, so they had the kind of rapid exponential growth even though it was originally a small margin business. I was working at an investment bank, and I was fascinated by this rapid growth. In college, I remember I was running highly computationally intensive code for lots of statistics homework assignments, and sometimes it would take hours. Running it on Amazon EC2 just wasn't an option because it was too expensive.

What was your inspiration behind Ankr Network? How did it all start?

Chandler: After I talked to Ryan about wanting to dive deep into blockchain and explore how I could make use of trusted hardware, I started coding, building and prototyping. That started in late summer of last year, and by sometime in the winter I went to Ryan very excited about the SGX paper. I had been busy building out Oracle networks and I knew I wanted to solve this problem of being able to provide cheaper decentralized cloud computing. I'd researched some of the other projects in the space, but I didn't want to follow the same route specifically because it had a really high barrier to entry for non-technical people.

At the same time, I was talking to Professor David Anderson of the Berkeley Open Infrastructure for Network Computing (BOINC) project. So I went hunting for a more elegant solution, but I knew I had two crucial problems to solve: (1) it's not private, and that's less appealing to businesses, and (2) they were having trouble because BOINC didn't have an incentive scheme to encourage volunteers. That's when things clicked and I decided that trusted hardware would be our differentiating point, and the blockchain consensus mechanism layer would help us solve what had been the incentive problem in BOINC.

Who are your competitors and what do you consider your main competitive advantage?

Chandler: dApps are getting increasingly popular, and they're placing a growing computational burden on the blockchain ecosystem. As they become more computationally expensive, the ecosystem needs a decentralized marketplace for cloud resources.

Our competitors in the blockchain space are services like Golem, SONM, DFinity, IExec, Perlin and Hypernet. Most of those services are Ethereum based, and they all provide cloud computing in some form, usually through smart contracts on a blockchain.

The Ethereum platform is tied to the current Proof of Work-Proof of Stake hybrid consensus mechanism, and platforms like Golem and SONM use virtualized computing containers to manage, and have the smart contracts as the execution step. Participants collect on the computation fee provided as a reward for someone filling a computation request. And IExec uses an existing grid computing platform, which limits the number of sources of income. Dfinity and Hypernet are examples of projects where the computation is limited by the smart contract it's executed in.

Ankr is focused on providing a computing platform with innovations at the consensus level. Proof of Useful Work makes use of the pointless mathematical computation wasted in Bitcoin, and the SGX ensure that computation executes exactly like it's supposed to. Where our competitors are worried about TPS (transactions per second), Ankr is going to provide users with several sources of income- mining, transactions and cloud computing- and change some very basic things about the protocol that just make common sense.

Ankr isn't limited by a smart contract because the miners make up the "cloud computing." Furthermore, because Ankr provides their miners with a computation fee as well as a mining reward, they're able to accept smaller computation fees, and provide lower computation costs for the same amount of work done. By moving the whole margin lower, we maximize computing efficiency, and at the same time we provide more profit.

What is the biggest problem currently in the market?

Chandler: The market right now is high margin and currently exists as an oligopoly. Ankr is trying to find a way to make certain things cheaper. A large amount of people are just using single-function computing resources (AWS lambda).There are still going to be functions that AWS can provide and Ankr cannot, but out of the say hundred functionalities Amazon provides there are some that are more valuable than others. Those, and the ones that make sense in a decentralized setting, are the ones we want to focus on for Ankr.

The idea of a decentralized computation platform can be difficult to grasp. When you evaluate the space, you find that most projects trying to accomplish similar goals are all trying this in a manner focused on non-centralization. But when that's your viewpoint, you run into problems like the fact that running all the computation in a large smart contract with lots of nodes is terribly expensive. It's really very different from having lots of chips running in parallel. If you think about the cloud computing problem from AWS' perspective, their platform runs like there are tens of millions of CPUs doing the work. So we took that model, and learned from BOINC. We were able to solve the redundancy, incentivization and confidentiality problems there.

To expand on the redundancy problem: Imagine you have one job, and you break it into five pieces, and send a fifth of the original problem to five nodes each. If one of the five nodes takes forever to execute, that means your overall problem takes forever to finish, which is problematic. BOINC solves that by sending that one job to five nodes instead. If they all come up with the same answer, it's most likely the right answer. The problem is less that the answer might not be right, but this is how they ensure that it's the right answer.

Another common example is that of evaluating 2,000 pictures to find aliens. It's a computationally expensive task, and the node could say it looked at all 2,000 pictures and with high probability they will conclude that they didn't find an alien. But we brought in SGX to act as a proctor that could guarantee that the CPU had executed. It doesn't need to guarantee correctness, it just needs to guarantee execution because correctness is built into the machine.

Ankr has been regarded as one of the top projects at the moment. For other blockchain entrepreneurs, how did you guys achieve such positive notoriety?

Chandler: Last January, there was a time when you could have a whitepaper with anything on it and you could still raise money, but that time passed really quickly. Now, it's really important to be able to prove yourself in this competitive space. We have the benefit of good technological development. We found problems and real places where decentralization could benefit people. Just the problem statement: "AWS is too centralized", would have been difficult to run with. Suppose there is a centralized business where there is very low margin. Decentralizing when there is almost no pie to be divided up among many people becomes very difficult. But we stayed true to what we thought the problem was, and we were lucky enough to be able to showcase that to investors, and our community is able to band together behind this goal. I genuinely think we're headed in the right direction.

Can you break down Proof of Useful Work for laymen?

Chandler: Proof of Useful Work has all the flavors of Proof of Work's "In Math We Trust." Suppose Bitcoin is the teacher, grading students answers. All the nodes/students are doing their homework (which in Proof of Work is just pointless brute force computation to find the nonce). Suppose there are 1,000 students, and the first one to solve the answer gets the reward. In Proof of Work, that means they build the block, collect the mining reward, etc. All the other 999 nodes have now wasted their time and energy to work on this question, and now they get nothing. Now sometimes this is necessary, as in the case when the professor isn't a centralized agent, but the protocol itself. But it's clear this is a very wasteful way to achieve decentralization.

Consider instead Proof of Useful Work where the protocol is once again the professor. But this time, the students are no longer doing the same hard math questions. This time, they're doing useful computation for other projects/organizations. These organizations are all paying those students to do that computation. Maybe one project pays 20% of the students, and another project is only paying for 10 students, but everyone who's working is getting paid. All the students are doing different homework tasks at the same time, and getting paid by potentially different parties. This is where trusted hardware comes in; we need something to act as a proctor/professor who essentially has a pair of omniscent eyes and monitors every single student very closely.

You can understand the way that works by the following metaphor: If a student writes 10 lines within a period of time, he or she has the opportunity to participate in the Poisson process. This Poisson process has a lot of relatively complex parameters, but most importantly, it has the number of instructions to the CPU and the age of the CPU (we don't want people to use computers that are way too old!) Then through this process, we randomly choose one person to get the block reward. But everyone is also still getting paid by the enterprises at the same time. It models the real world, where people get paid regularly but can also run into serendipitous rewards. And at the same time the CPUs can utilize their idle computational resources, and that's the basis of the potential for our Universal Basic Income long-term dream.

To try one more analogy: In photosynthesis, light and water generate sugar, and the byproduct is oxygen. In our case, we'd like to make the analogy that Proof of Useful Work as a consensus protocol is at its core trying to generate a new block. It's still a ledger, and the main purpose of the consensus protocol is to generate the block. But the side product, which is very important, is the useful work-- the oxygen.

Other: Anything else you'd like to add/include? Get creative/controversial.

Chandler: A lot of people have FUD- which stands for fear, uncertainty and doubt- about the fact that we're young. But we'd like to face this challenge by these naysayers, head-on. In this space, you have plenty of young entrepreneurs. The industry is tremendously young, and if it's new to us, it's new to everyone. We're one of the youngest in the space, and we started right out of college-- we won't deny that-- but we're set on learning everything that comes up, and faster. We want to address this question of youth in two perspectives.

First of all, the people who are really building the technology right now, are four key engineers: Stanley, Xiuyan, Songliu, Giacomo. They're leading development, and they're all incredibly qualified to be doing that. Stanley worked at AWS for years, Xiuyan and Songliu have close to 20 years or experience in security at Palo Alto Networks and Gigamon respectively. Giacomo is a PhD and his thesis paper was actually on the sharing economy. The people who are leading development, know what they're doing, and it's all sound technology.

And secondly, in terms of business and leadership, we're young but we have the advantage of courage and diligence. We're more energetic, and physically we're at the peak of our lives. We're pulling all-nighters and working hard, and we love talking, meeting and connecting with people. We have good connections, and I personally think one of the main reasons we're successful right now-- the key reason these four older engineers are willing to listen to Ryan and I, is because we're good at finding workarounds, and solving unique problems. Being young is one of our key advantages. We have nothing to lose, and all we want is to chase our dreams. Ryan and I had good track records, but we decided to quit excellent jobs and chase this dream because we confidently believed-- and we had people who believed-- that we were perfectly capable of following through.

Brief about founders' history

Chandler and Ryan were college roommates. Chandler was EECS and Ryan was Business and Finance/Statistics. Everything really started back in 2014. Chandler was the third person Ryan met in Berkeley. Back then, Chandler was interested in Bitcoin, and highly recommended to Ryan that they invest any extra capital. Ryan was initially going to buy a car with the money, but he didn't have time to learn how to drive. So they all bought around 22 Bitcoin around 500 dollars apiece, and they held onto it for the long time it took to eventually blow up in value.

Chandler joined Blockchain at Berkeley, did some research outside, had friends in different schools who were also in this field, discussed with a lot of them extensively, and he had a very close friend, PhD, who started him on the journey in trusted hardware. In August of last year, he mentioned to Ryan that he wanted to do something about blockchain. He started playing around with code and prototyping, but it took until around winter that Chandler realized the potential of the whole SGX paper.

He discovered the idea of the Oracle service, coded some parts of the idea himself, discussed with various people about the potential. He had a supermanager, Stanley, while he was interning at Amazon. Stanley was one of the early Amazon Cloud engineers, wrote some of the earliest code and led the Alexa Rankings team. Stanley had a good working relationship with Chandler, and a long impressive history in AWS, specifically EC2. Chandler took his original concerns of the high margin and high barrier to entry it takes to run EC2 to Stanley, and Ryan discussed this with Chandler as well.

For the first few years, there was exponential growth from a very small margin business, and now we also have Google Cloud but from a finance/banking investment banking perspective. Amazon became valuable as a company specifically because of AWS. As a statistics major, Ryan was also running highly computationally intensive code, and in R sometimes it would take hours. It was too expensive to use Amazon EC2 to run it just for homework, and so Ryan had both a finance and usage perspective. Chandler had the blockchain and trusted hardware perspective.

Well aware of other projects, but they didn't follow the same route, specifically because it has too high of a barrier to entry for non-technical people, but they were looking for a more elegant solution. Trusted hardware became their differentiating point. Chandler was in connection with David Anderson of the BOINC project, but very quickly we can see why businesses haven't adopted this. Trusted hardware along with the blockchain token helps solve the incentive problem in BOINC...

The market and Ankr

The market right now is high margin and currently exists as an oligopoly. Ankr is trying to find a way to make certain things cheaper. There are still going to be functions that AWS can provide and Ankr cannot. A large amount of people are just using single-function computating resources (AWS lambda). Let's say that right now Amazon provides 100 functionalities because they have to. But there are functionalities that are more valuable than others, but there are certain of those that we'd like to focus on that make sense in a decentralized setting.

The decentralized computation platform is too complex, and there are too many limitations. If you think about it, most projects approach this problem in a way that's not centralized. They're running all the computation in a large smart contract, with lots of nodes. That's a lot different from having lots of chips running in parallel. AWS runs like there are tens of millions of CPUs doing the work, so we took that model, and learned from BOINC. It's just that BOINC doesn't have the ability to solve the redundancy, incentive and confidentiality problem.

To expand on the redundancy problem: if you have a job and you distribute it to five nodes, probably one of the five nodes hasn't executed it yet, which becomes problematic. So BOINC instead sends that one job to five nodes, and if they all come up with the same answer, most likely that is the right answer. The problem is less that the answer might not be right, but they want to be sure that the task is executed. Another nice example would be the finding aliens in 2,000 pictures example. The node could say they looked at all 2,000 pictures and with high probability they could conclude that they hadn't found an alien. But the SGX is better because it can tell you as a proctor that the CPU has executed. It's not guaranteeing that it's correct because computers are machines that execute presumably on the same code, but it does guarantee execution…