Raven Protocol AMA with Sherman Lee in the Binance AMA Series
Dear Raven Protocol Community!
Sorry to all those who couldn’t make the AMA on November 1st, 2019. I know it was very short notice between the announcement and the actual AMA hosted by Richie (our favorite Binance Angel). This was a very special AMA. We shared with everyone our latest plans on working with real customers with real use cases. It was supposed to last for only an hour. Sherman was glued to his laptop for over 2 and a half hours and many cups of coffee to power through. At one point, 200–300 questions came simultaneously. This may have been the most successful AMA ever hosted on t.me/BinanceDEXchange :) Thank you so much to the Binance Chain community for having us.
Here is the full transcript. You don’t want to miss a single word.
INTRODUCTION
Q. Hello Mr. Sherman! How are you doing today?
A. Hey Richie, I am doing well! Thank you so much for having me here today 🙏
Q. So, before we start the show. Could you kindly tell us a bit about you and Raven Protocol?
A. Yes of course. I will be glad to introduce myself and Raven.
Hey Binance DEX community! I am Sherman from Raven Protocol. We are working on decentralized and distributed AI training. Our value proposition solves a very specific problem — the speed of AI training. It typically takes about 2–3 weeks to train a 1M image dataset on say AWS. Raven is building a distribution framework from the ground up that takes this down to 2–3 hours.
The team has been working in AI and Machine Learning for years. Personally, I’ve been doing ML since 2008. Rahul has built a ML platform with enterprise customers. Kailash wrote the book on Building Next-gen Generative Models https://www.amazon.in/ Generative-Adversarial-Networks-Projects-next-generation- ebook/dp/B07F2MY1QH. Shannon worked on content generation using AI.
Q. Okay. Thank you Mr. Sherman. Hope AI won’t reach to the level where they could start hosting AMA’s 😂
A. It will be a long time before AI can reach the level of sophistication as you Richie. But I’m sure it’s close enough in terms of the tech today help you augment your work to make it more efficient!
SEGMENT ONE: [QUESTIONS FROM THE AMA HOST]
Q. Raven is working on a deep technical problem. Can you please explain more about how it works?
A. Raven is creating a network of compute nodes that utilize idle compute power for the purposes of AI training where speed is the key. A native token is the key to bootstrapping a nascent network.
We want to incentivize and reward people all over the world to contribute their compute power to our network. Additionally, we will reward token holders for running masternodes which will be responsible for orchestrating the training of various deep neural networks.
Our consensus mechanism is something we call Proof-of-Calculation. Proof-of-Calculation will be the primary guideline for the regulation and distribution of incentives to the compute nodes in the network. Following are the two prime deciders for the incentive distribution:
Speed: Depending upon how fast a node can perform gradient calculations (in a neural network) and return it back to the Gradient Collector.
Redundancy: The 3 fastest redundant calculation will only qualify for receiving the incentive. This will make sure that the gradients that are getting returned are genuine and of the highest quality.
Q. How do you plan to compete with big players like AWS? Now that’s a challenge I would say, but no doubts that it can’t be achieved.
A. Going up against big players is not as scary as it seems. They have their own set ways of doing things and they can be disrupted. Amazon AWS, Google Cloud, Golem, etc., offer general cloud compute. We focus specifically on the speed of AI training. Thus, building our own distribution framework allows us to bring down the AI training time down.
Conventional distribution methods have an inherent latency in their network. Large chunks of data need to be passed between machines. Our unique approach to distribution solves latency by chunking the data into really small pieces (bytes), maintaining its identity, and then distributing it across the host of devices with a call to action: gradient calculations.
To dive in a bit, we take both Data and Model Parallelization approaches to form a different model of distribution.
A major difference between a static and dynamic computation graph is that in the former, the model optimization is preset, and the data substitutes the placeholder tensors. Whereas, in the latter the nodes in a network are executed without a need for any placeholder tensors. The benefit of a dynamic graph is its concurrency, and it is robust enough to handle the contributor addition or deletion, making the whole Raven training sustainable.
Raven is thus capable of eliminating the latency and scalability issues, with both the approaches. Hence, distributing the training of any deeper neural network and their larger datasets, by getting rid of the added dependency on the Model replication. Data is also sharded in smaller snippets.
In fact, the Model is intact at the Master Node, and the heavy lifting is distributed in the tiniest snippets of data subsets over the network of contributors. The resultant gradients, after the calculations that happen at the node/contributor ends, are sent back to the Master Node.
This creates a ton of difference, as it is easier for calculations to pass through from machine to machine, rather than creating multiple replicas of a complicated Model.
Q. What are the latest plans for Raven?
A. In our latest update we announced that the development of functions such as Back-Propagation (Back-prop) and several activation functions that are used to train vanilla Neural Network models have been completed. That was an important milestone as it allows us to engage deeper with alpha customers.
Making sure our tech works with real customers who train thousands of models a day will be no easy feat. We are at the very early stages of growing. These are real companies that will be using Raven Protocol as a core component to their technology stack.
AI is being applied to every single vertical and it’s experiencing explosive market growth. Revenues are increasing from around $9.5B U.S. dollars in 2018 to an expected $118.6B by 2025. The crazy thing is that AI is in its infancy. Raven is here building an important foundational layer for the fast growing AI industry. We have a long runway as we have support from funds who think about the long-term opportunities. Our company is at the same growth stage as post-Series A companies in the traditional equity startup world. We have 10 engineers and we are looking to hire additional AI/ML researchers.
Raven has a competitive advantage and a unique approach to distribution. We want to maintain this for some time. We can’t quite announce it yet, but we are partnering with companies in China, India, and Africa. These are three large, growing, and important markets for AI + Blockchain. Our focus is on the real use cases and real utility of our tech. Enterprise companies will purchase RAVEN to access compute power on our network. As a foundational layer to AI training, Raven sits in the perfect position to grow with these markets :)
Q. Do you think IDOs will still exist in the future and WHEN EXCHANGE?
A. HAHA if I had 1 BNB for every time someone asks us that question…
It’s fascinating how the blockchain industry moves so fast. We had ICOs, private sales, IEOs, and then IDOs. As stated when we launched the first IDO (Initial DEX Offering), it was an experiment and a learning process to push the innovations of blockchain fundraising forward. These mechanisms help important technologies come to life. I believe IDOs will exist in some form. Fundamentally it is a decentralized launchpad. It’s a way for projects to get in front of a large community on launch. Excited to see these mechanisms evolve on the Binance Chain ecosystem. Raven is honored to have played a part in shaping that future.
Regarding exchange listings… HAHA of course you know that we’re unable to comment publicly on that. It’s up to the exchange to decide when to list a project and not the other way around. We do recognize the importance of liquidity that centralized exchanges provide and in addition
to that, we are exploring many ways to provide that liquidity.
Hummingbot for decentralized market making is launching on Binance DEX next year. Thorchain’s BEP Swap (Uniswap for Binance Chain) is launching soon. It would allow you to stake RAVEN in liquidity pools and earn commissions on every trade. The Binance Chain alliance is also working on a solution to this to help all Binance Chain projects provide liquidity.
But you know Binance Dex IS an exchange. And arguably it will be the best exchange in the world. We’re still very early in that process, but hey Binance CEX has been around for over 2 years and look at the progress that’s made. Binance DEX only launched in April and has made phenomenal progress both in terms of adoption and product.
There’s a ton of activity in the Binance Chain ecosystem. Exciting to see so many people building tools to help everyone grow
Q. Can you share any interesting companies, use cases, and applications of AI/ML?
A. Deep learning is an expensive and time-consuming process. Image recognition, natural language processing, speech recognition, computer vision, etc.requires extracting millions of parameters to identify structure and patterns. It is computationally intensive. That makes it quite hard for companies today innovate even for the best use cases and applications.
Obviously meeting with AI companies has led me to meet some of the best founders in the world in this space. A few companies that I’ve gotten to know and really excited about are…
1) Fasal is an AI-powered IoT-SaaS platform for horticulture, which captures real-time data on growing conditions from on-farm sensors and delivers farm-specific, crop-specific actionable advice to farmers. Their field sensor array can be installed by farmers in India in less than 15 minutes and measures micro-climate, soil, and crop conditions. Fasal leverages machine learning to transform this field sensor data into farm- level predictions, anticipating various risks while helping horticulture farmers to reduce input costs by optimizing crop protection, irrigation, and crop nutrition. They just recently raised $1.6M seed round and we’re excited to watch their company and crops grow: https:// agfundernews.com/breaking-indias-fasal-raises-1–6m-seed-funding-to- build-out-precision-ag-across-southeast-asia.html
2) Deeptrace is the antivirus for deepfakes. If you’re not familiar with deepfakes, it is a technique for human image synthesis based on artificial intelligence. It is used to combine and superimpose existing images and videos onto source images or videos using a machine learning technique known as generative adversarial network. The company, Deeptrace, is the first-to-market deepfakes detection solution designed to guarantee the integrity of visual media. They have been all over the news including Fast Company: https://www.fastcompany.com/ 90414479/how-deepfakes-evolved-so-rapidly-in-just-a-few-years. This team being so early in the market with a large opportunity, I wouldn’t be surprised if they raised another round soon.
3) Seoul Robotics is a dynamic provider of perception software for autonomous systems. They announced SENSR over the summer, a 3D perception software platform compatible with major 3D LiDAR sensors. Seoul Robotics is now the only company to commercially provide a high- performance 3D perception software solution that is compatible with multiple brands of LiDAR manufacturers. This means they are making it cheaper to develop all autonomous vehicles and systems.
With so many founders working on challenging problems all over the world, I am further convinced that decentralization is the way forward. Talent has proven to be universal time and time again. Opportunity typically hasn’t been, but the blockchain industry is showing it can be universal too.
SEGMENT TWO [QUESTIONS FOR THE COMMUNITY]
Q. What is the projected revenue for the AI industry in 2025? How much of that market do you think Raven Protocol will capture?
A. $118.6B is the correct answer. There is no incorrect answer to the 2nd part. But remember how early we are in this space :)
Q. What is the key value proposition that Raven is solving for the AI industry?
A. We perform AI training where speed is the key. Easy one right?
Q. What are the three large, important, and growing markets Raven will enter?
A. China, India, and Africa
Q. What role with Raven play in the future of AI?
A. Pro tip: there is not incorrect answer for this. Use your imagination for how big this could be!
Q. What market cap do you think RAVEN will grow to in 5 years given that you now know our technical depth, expertise, go-to-market strategy, vision, and opportunity ahead?
A. Answering this would be price prediction which we will not do. However, I will say we currently have a low market cap and an immensely huge opportunity ahead.
SEGMENT THREE [ QUESTIONS FROM THE COMMUNITY]
Q. How did you succeed to increasing the speed of image recognition of image recognition from 2 weeks to few hour of 1M images ?
A. We’re developing our own framework written in Python and c++. Because we have no dependency on the architecture of the machine, nodes in the network can be as simple as a Javascript client. E.g. a website could install a javascript snippet and execute gradient calculations on page load.
As mentioned before, conventional distribution methods have an inherent latency in their network. Large chunks of data need to be passed between machines. Our unique approach to distribution solves latency by chunking the data into really small pieces (bytes), maintaining its identity, and then distributing it across the host of devices with a call to action: gradient calculations.
Q. How will you grow your current user base with Raven? Any plans to expand Raven globally?
A. We are currently working with a handful of alpha customers. We do have plans to expand globally and in particular China, India, and Africa. They’re very exciting markets and we get to be early in them.
Q. What do you think of the current crypto market? What impact will the changes in the crypto market have on the price of RAVEN token? Do you have any plans to avoid the price slumped like last year? How did RAVEN Survive in during long bear market? what your team strategy to gain more adoption ? Can you share us on recent achievement your team are most proud of?
A. We raised our seed round in September 2018 during a tough bear market. Our supporters who believed were in it for the technology. In our latest update we announced that the development of functions such as Back-Propagation (Back-prop) and several activation functions that are used to train vanilla Neural Network models have been completed. This allows us to engage deeply with alpha customers
Q. Hello , Here im a Raven fans since IDO and i believe that RAVEN will grow day by day. Here my question is : Right now RAVEN has small marketcap and only list on DEX which is only has less liquidity. What are you gonna do as a CEO of RAVEN rather than list RAVEN on big CEX? Maybe like more big trading event or buyback event to attach more investors? and is there any plan to has big partnership? Sorry for my question because im a long term investor and i hope i invest in the right project! thankyou!
A I believe i answered this earlier, but happy to talk it through again. We’re unable to talk about exchanges publicly. It’s up to the exchange to decide when to list a project and not the other way around. We do recognize the importance of liquidity that centralized exchanges provide and in addition to that, we are exploring many ways to provide liquidity. Hummingbot for decentralized market making is launching on Binance DEX next year. Thorchain’s BEP Swap (Uniswap for Binance Chain) is launching soon. It would allow you to stake RAVEN in liquidity pools and earn commissions on every trade.
Q. What will AI be doing in 2030? What percentage of it will be utilizing Raven’s AI infrastructure? If AI starts replacing traditional jobs, what is your opinion of a UBI / Universal Basic Income?
A. I like where your head is at! In 10 years, AI will continue to augment the work of humans. Hard to say what percentage of it will be utilizing Raven. Hopefully 100. UBI is great IMO. It will clear up headspace
Q. What is the uniqueness of the Raven token? Why should investors (including me) invest in Raven?
A. It’s a utility token. You will be alongside enterprise companies purchasing RAVEN to utilize on the network. As more and more customers onboard, the supply could become more limited. You’ll start seeing more and more utility of the RAVEN token as well through marketplaces for purchasing AI services.
Q: What do you predict Raven Protocol’s largest challenges will be in the future? How do you plan to address them?
A. The largest challenge is hiring people who know how to build a framework from the ground up. Many developers and engineers know how to use Tensorflow and train models with AWS. Only a handful in the world know how to combine research with implementation of a framework.
Q. In a short time RAVEN project attracted great interest. What do you attribute this to? Investors always support good projects. He pays attention to promising projects. Do you agree?
A. The team has been in this space for a while and is working on an important problem with a very large market opportunity. Building this at scale will have a huge impact on the world if you think about the downstream effects.
Q. Does Raven have any competitor in the crypto space? If yes, what are you doing as a company to deliver a better product and to be more preferred than your competitors?
A. Amazon AWS, Google Cloud, Golem, etc., offer general cloud compute. We focus specifically on the speed of AI training. Thus, building our own distribution framework allows us to bring down the AI training time down.
Q. What are the incentives for someone holden Raven tokens. How are token holders rewarded? How does the Dynamic Graph Computation help Raven?
A. We will have masternode staking and as mentioned earlier we are exploring liquidity pool staking were you can earn for holding RAVEN.
With Dynamic Graph Computation the benefit of a dynamic graph is its concurrency, and it is robust enough to handle the contributor addition or deletion, making the whole Raven training sustainable.
Q. What ́s your outlook on the future of cryptocurrencies? What can we do to keep increasing adoption? How can RAVEN help to achieve it ?
A. There’s no doubt the crypto and blockchain are here to stay. Adoption happens when you have product-market fit. That’s why we work with real customers who have a real need for our tech. It’d be silly to build something without having a user for it.
Q. Why did you build Raven protocol, what is the motivation for you to start this project?
Rahul and I were working on our respective AI startups. User growth was coming fast and we were paying 5–10K usd a month for AWS and the training times were slow. We wanted to have cheap/fast compute power. Rahul started building a prototype and things started actually working. No crazy story really. It was born out of our own needs
Q. What are the core strengths of the RAVEN ? Where do you see RAVEN in 5–10 years? Bitcoin use the cryptographic keys but i want to know what does you use?How can I be sure that my holdings are safe and accurate?
A. The speed of AI training. We will be the core infrastructure layer for AI/ML training for many enterprise businesses. We use Binance Chain so you be sure it’s safe.
Q. Could you talk about the security of the Raven Protocol?
What makes the Raven token unique from others?
A. In addition to data being encrypted, data is broken up into small bytes and distributed across devices. It will be impossible to piece that data back together.
Q. Do you have any plans to attract non-crypto investors to Raven Protocol and how? What are the measures to increase awareness of Raven Protocol in non-crypto space?
A. This is an amazing question! Our customers don’t need to care if we use blockchain technology or not. They just want their problems solved. We do have many traditional equity supporters who contributed to our token launch.
Q. What has so far been the most fulfilling experience ever since starting Raven Protocol?
A. The most fulfilling experience is defining the future of the AI industry. We are in its infancy right now. It’s just incredibly fun to be on the journey
Q. Between the goal of building an increasingly strong platform, improving user experience and the goal of attracting more satellites project into the Raven Protocol ecosystem. Which goal is Raven more focused at the moment? And if so, how can the Raven Protocol convince them to use Raven Protocol technology instead of their old technology?
A. We are focused on working with alpha customers right now. This is critical to real adoption. Not much convincing is needed for people to understand why they need our tech. The tough part is like any enterprise sales. Lots of meetings, demos, trials, etc.
Q. What were the major difficulties that you faced while starting this project ? And what was your motivation to continue?
A. The hardest part was during the bear market. When almost nobody believed in token networks anymore, we raised a seed round and then subsequently had to convince people who no longer believed in crypto to believe in Raven :)
Q. When looking back at your time working for RAVEN , what are you most proud of?
A. Building a new piece of technology that most people deemed impossible. When we set out to solve the AI training problem, we had no idea if our approach would actually work. So we talked to machine learning engineers, data scientists, and AI researchers at corporations who have spent billions of dollars on AI. The brightest minds in AI at Google and Facebook did not think our approach was feasible. Even if we were able to build out the compute nodes in our network, they were skeptical that the distribution of computations would actually make anything cheaper or faster. We really questioned ourselves at this point. These are people at the top of their game with all the resources in the world to do AI training. But there was a key difference that made us not listen to them. It was because they had all the resources in the world. AI engineers at Google and Facebook do not have a unique insight into the problem like us.
Q. Why did Raven choose IDO instead of IEO? Is this benefit for the Raven Protocol? How to become Contributors? What benefits can Contributors get?
A. It really came from the community when we announced we were launching on Binance Chain. They wanted to make sure we got in front of as many people as possible.
Q. Last week, the Crypto market became active from the news of the Chinese government. How would you rate Raven’s opportunity here as this is one of your target markets?
A. China is very big in AI and very big in blockchain. The two are a perfect fit for each other
Q. How does the migration to Binance Chain help in your project?
A. Binance chain is great! It’s fast and has 1s finality. Additionally, the support from the Binance Chain team was top notch. We’re very glad to have launched on it
Q. Where do I want to buy RAVEN -F66 token?
A. You can trade Raven on binance Dex https://explorer.binance.org/asset/RAVEN-F66
Q. Stablecoin is the word that I heard everyday, so do you have any plans to release a wallet for stablecoin?
A. We’ve discussed a stablecoin for AI-services. Would you feel more comfortable using stable coins instead of RAVEN?
End of AMA!!
Host: Mr. Sherman, thank you very much for taking the time to be with us today! If I may speak also in the name of the chat, it was awesome to have you here 👏👏👏
Sherman: Thanks so much to the Binance DEX community and of course Richie who is the best AMA host. I’ll leave you all with one last thought. The blockchain industry is in its infancy. We’ve come a long way and there is still a huge mountain to climb. The next 5, 10, 20 years will be incredibly crazy. We are laying the foundational layers for a whole new industry right now. Can you imagine the kind of value creation this unlocks? It will be full of twists and turns, but the journey will be fun. Let’s do this together. Come say hi on our telegram https://t.me/ravenprotocol!
Raven Protocol: Q2 2019 Tech and Community Update:
https://medium.com/ravenprotocol/q2-2019-tech-and-community-update-4f836a9a1e97
Raven Protocol: Q3 2019 Tech project development Update:
https://medium.com/ravenprotocol/tldr-raven-stayed-heads-down-building-in-q3-2019-ae5f242dc15d
Q3 Tech Update Summary:
First alpha customer adopting the RAVEN token and our protocol
Built a repository for development with engineers from various parts of the globe. Taking measures to safeguard our technology and distribution approaches.
Development of functions such as Back-Propagation (Back-prop) and several activation functions that are used to train vanilla Neural Network models have been completed.
Raven Protocol Project Review:
https://cryptocalibur.com/raven-review/
Raven Protocol White Paper:
https://drive.google.com/file/d/1FAaVKkg_CjxMj-n1yHZc6ufcVDtOU1Ct/view?usp=sharing
OFFICIAL CHANNELS:
Official Email Address: founders@ravenprotocol.com
Official Website Link: http://www.RavenProtocol.com
Official Announcement Channel: https://t.me/raven_announcements
Official Telegram Group: https://t.me/ravenprotocol
Official Twitter: https://twitter.com/raven_protocol
Official Medium: https://medium.com/ravenprotocol/
