“Neuromation is a distributed synthetic data platform for deep learning applications. Unblocking game-changing computing capacity for wide AI adoption”
A couple weeks ago, like most people, I had zero clue what this meant. Since I started researching this project my eyes have been opened, and my mind blown. Three key definitions to help understand the project.
Synthetic Data: Any production doata applicable to a given situation, not obtained by direct measurement.
AI neural network: An interconnected group of nodes, similar to the large network of neurons in a brain.
Machine learning: AI that provides computers with the ability to learn without being explicitly programmed. Solving giant math problems and self adapting when exposed to new data. The process is similar to data mining.
Neuromation is where data scientists and companies can get all their AI data analysis needs met faster and cheaper than anywhere in the world. This is possible because they reward GPU miners for using their computational power to mine this synthetic data. After the synthetic data is generated, scientists can specify how they want to train their machines to analyze the data in record time on a separate group of nodes.
Here’s how Neuromation describes themselves
“Neuromation is a technology platform that creates synthetic learning environments for deep learning of neural network. These simulations are then used for training better algorithms.
We are building the platform of distributed computing for creating artificial worlds where AI algorithms are trained on simulated sensory inputs. These synthetic worlds also have a virtually infinite supply of perfectly labeled training data. AI plays, as with video games, to learn specific tasks in the real world.
Our technology is crucial to making Deep Learning-based systems useful in the real world as they are taken up by industry. With Neuromation, the future has arrived where computers teach computers to perceive.
Our strategy is not to develop our platform in isolation, but to work with partners in select industries in order to try to bring our vision organically to life. We are developing “Neuromation Labs” that would develop synthetic data and train deep learning models on live applications. Each “lab” like Retail Automation Lab, Industrial Automation Lab, Pharma/Medicine/Biotech Lab will be a study on a specific problem in a partnership with a category leader. As our platform is fleshed out we will be moving parts of generation and training there, allowing us to organically test parts of our vision in real-life scenarios. The Labs will seed the Neuromation Platform market with initial data generator and data sets. We will also encourage our Labs partners to transact farther services through the platform, thus building the initial market for services.”
The Token
In order to do transaction on the platform you must use Neurotokens. To make it easy, Neuromation will provide a portal with a one click token buying process. The platform offers 5 services; Data generation, data labeling, data purchase, model training, and model purchase. The price customers pay for each service will depend on how much it costs for the nodes. The platform will determine the resources required for each task, then select the most efficient node pool (minimizing the cost to the costumer, while still making it extremely profitable for the nodes/miners. This brings me to the juice of this project.
Mining Knowledge instead of ETH: The reason why synthetic dataset generations haven’t seen widespread adoption yet is because of the huge deficiency of computing power. Neuromation plans to change this by offering cryptocurrency GPU miners 3-5 times what they make per hour mining crypto, to mine Neuro(synthetic data/machine learning) instead How this works is in addition to their existing mining software miners load up a neuromation computation node. This node is special because it allows miners to mine their normal cryptocurrency until Neuromation has a task available, at which time the miners will switch over to mining neuro until the task is complete.
Since I believe this to be the heart of the project, I emailed the neuromation team and asked them to break down further how their mining/node system works. Here’s what they said.
“to mine neurotokens, you will need to load our node. If your system is eligible (depending on processing capacity, bandwidth) the node will activate and wait for tasks. The nodes take on available tasks if they win an internal auction. That auction is run in tiers. Tiers are numbered from 0 to 4. Nodes in tier 0 get a higher priority (the auction will first try to distribute the task among lower tier nodes) then nodes in tier 1 and so on. When our token sale is running you only tier 4 nodes will be available.
We have a program where miners can invest various amounts into out pre-sale and secure a certain number of lower tier nodes. It is:
– 3,000 Ether for tier 0 (500 node keys)
– 1,000 Ether for tier 1 (100 node keys)
– 600 Ether for tier 2 (100 node keys)
– 200 Ether for tier 3 (50 node keys) “
Why Blockchain? Today if a data scientist wants access to a large amount of computer power he can go to Amazon and pay for it. They will charge him 12 times what it costs for GPU miners to generate it. This is a giant opportunity for Neuromation to slide in.
Distribution of Token
60 million of the 100 million total supply will be sold during the ICO.
base price = .001 or .30 cents per token + bonuses
The presale is taking place right now thru November 28 (25% bonus). you must sign up for the whitelist https://ico.neuromation.io/en/
The regular ICO starts November 28th and runs 4 weeks:
week 1 (15% bonus) week 2 (10%) week 3 (5%)
Important to note: Over the next 3 years, neuromation will burn 50% of the total supply!
2018 =30% 2019= 20% 2020=10% (hmm, this is actually 60%, i think they made an error)
The Team
This is a seriously talented team They are out there putting in a lot of work too. Winning competitions, attending conferences, and forming partnerships (hacken.io, TaaS).
CEO- Max Prasolov has been a big time player since 2001. Here’s a short informative interview he recently did. https://cryptosrus.com/interview-neuromation-ceo-maxim-prasolov/. I’m actually just going to copy and paste his introduction to the rest of his team. Note, I did look fairly deep into each one of these guys, all legit, max just did such a good job summing them up I will use his words.
“I’m a serial entrepreneur and TOP manager with 15 years of experience in different sectors from natural resources mining to industrial multimedia development. I was a part of the team who made an IPO of the iron ore company, blue chip on the London Stock Exchange. But all my life I’ve been in love with graphic novels and animation. I have written and produced several animation movies. While making them I have found out that AI can be trained by showing it the cartoons. This metaphor is very close to what we are doing in Neuromation.
I am glad to work with very smart and talented people involved in our company. Our mentor and advisor Andrew Rabinovich, who is a creator of key deep learning algorithm of Google image. My partner, investor and founder of Neuromation Constantine Goltcev, who believes in me from the beginning and he brings powerful engineers to our team. A deep learning scientist and mathematician Sergey Nikolenko, who is checking our crazy ideas with the proof of scientific method.
Fedor Savchenko, experienced CGI expert who creates synthetic data generator. Kyryl Truskovskyi, a talented researcher and engineer, who implements our hypotheses in deep learning applications. Denis Popov, former CTO of Viewdle, who is helping us to hire strongest software developers all over the world. All of these people are my partners and shareholders at Neuromation”
Advisors
They have 2 absolute beasts on their advisory board.
Andrew Rabinovich – has studied machine learning for over 15 years. has many patents and peer reviewed publication. He is the Director of Deep Learning at Google, Magic Leap!!
David Orban (Singularity University) apparently this guy has been a pretty big deal for the last 20 years. He’s definitely good at hyping Neuromation.
Verdict
Because it currently costs so much to “knowledge mine”, this giant futuristic AI market has yet to take off. Neuromation has the answer. They incorporate the mathematical power of blockchain miners to make synthetic data mining and machine training cheaper and faster than ever before. Its now possible that in the near future, thousands of deep learning projects and b2b clients will use neuromation to develop things we haven’t even heard about yet.
The team already has many labs in the works, and has recently boasted about their new Retail Lab, that is already providing “image recognition services to major retail brands”
When just reading the nerdy worded explanations of what synthetic data machine learning is all about, it could be hard to fully grasp what changes this can have on our entire system. I recommend watching youtube videos demonstrating the power of these self learning machines.
Risks
- It will take millions of micro-transactions for this platform to run once fully live and running. Like every project in the space right now, scalability is still a giant issue. I emailed the team asking how they plan to deal with this, their reply, “We are looking at HashGraph currently. EOS is also an option. For version 1 of the platform we will run an auction internally on our system and not on the blockchain, so transaction volume is not an issue until the end of 2018. For now, we will likely use HashGraph. “
- Data is the new oil, and there will be businesses aggressively competing in all sectors. Luckily for Neuromation, and their extra cheap/fast computational power, they are far ahead of the pack when it comes to knowledge mining.
- If they cant get the miners to adapt their nodes, then the project dies. I don’t think it will be hard to get them to switch though if neuros pays way better than anything else
- With the timing of this ICO coming out during this fork, it may be possible they don’t sell enough tokens to reach their final goal (which requires 60k eth) world wide adoption.
- These machines are already learning things from this synthetic data that no human has ever even thought of before. This is freaky.
Growth Potential
- Truly unlimited
- Neuromation plans to open an Enterprise Automation Lab where the synthetic data approach will help implement solutions in manufacturing, supply-chain, financial services and agricultural industries, to name just a few
- According to Gartner, by 2020 85% of customer interactions in retail will be managed by artificial intelligence.
- Here’s a quote from neuromation that sums up their potential better than I can.
“What is especially fascinating is that in only about a month of working on the inventory recognition problem we have achieved 95+% accuracy, a result others have spent years of effort and millions of dollars on. Significantly, the model performs well on real-life data without seeing anything but synthetic datasets during training. This breakthrough proves the viability and efficiency of our approach. With Neuromation Platform, the gateway to easy AI training at scale will be finally opened.
The potential demand for image recognition alone in the retail industry is enormous
It amounts to more than 40 billion images per year, according to ECR research of 72 of the biggest retailers and suppliers. Going further, we plan to create datasets that mimic human interaction with the shelf. We will be able to track customer flow and intent, creating a full simulator of the retail store, with a multitude of possible applications”
Disposition
I have to admit, after watchin numerous machine learning neural network demos on YouTube, I feel like a teenager who just touched his first real life boob. My mind is filled with wonder and amazement. That being said, you should do your own research. I’m clearly smitten, and could be ranking her too high.
This project is unique, revolutionary, and cheaper/faster than every competitor. The science they are accelerating with these large data mines could change the world in a lot of ways. If they grow their community, and get those miners to switch over their GPU power, the possibilities are endless At only .30 plus bonus and a lower supply of only 100 million, I feel very comfortable recommending this ICO. I wouldn’t sell it anytime soon either. This is a newborn baby giraffe, with the sexy genes she got from her mother! 8 out 10
Investment Details
- Symbol: NEURO
- Presale : now until nov 28th join the whitelist here https://ico.neuromation.io/en/
- Opening sale: Nov 28
- End Date : January 1, 2018
- Supply : 100 million .001 eth per + bonus