The blockchain space is full of projects without a clear use case and clear value, but, as we’ll see, there are also projects which are quite the opposite. Nebula-AI (NBAI) is creating a decentralized AI computing platform which will make AI DAPPS (which they call DAI Apps) a reality. The first product and demo that they released is Quant-AI: A cutting-edge trading price prediction tool. In this article, we’ll dive into the specifics of this project and take a look at Quant AI and the broader concept of DAI Apps.
Thanks to the use of GPUs and parallel computing for algorithms like machine learning (ML) AI has finally become widely viable. This created a huge industry which infiltrated into our daily lives. Every time you are talking to your virtual personal assistant (Alexa, Siri or Google assistant) or Facebook suggests you which friends to tag in your photos, AI is working behind the scenes.
AI is more than one of the many buzzwords we hear in the tech world. AI is often proposed as some sort of panacea for all sort of problems, which it is not. Still, its potential is so great that the estimated value for the AI market worldwide in 2018 is over $7 billion, in 2020 it is expected to grow to $17 B and in 2025 it is estimated to hit $90 B.
This technology is already overwhelmingly used in many industries. In medicine it is used to predict heart attacks or diagnose mental health problems from speech patterns (and AI actually seems to manage both those things better than human experts), in finance it is used for quantitative financial analysis and price prediction (an example of this is Quant AI itself) and the list could go on. AI is not only Alexa and Google Assistant, it is going to be a really important part of our modern lives way sooner than most expect.
Current issues of the AI industry
There are some big issues with how AI services work nowadays which Nebula-AI is trying to solve. Before explaining how those are being addressed let’s take a look at what are the problems that the AI space is currently struggling with.
Centralization and control
Since the infrastructure, which provides the service is completely centralized, it is also completely dependent upon the hosting company. Cloud providers such as Google or Amazon could decide to cease offering their services at any given moment. That means there is no guarantee that the service that you are using will exist in the future. This could jeopardize your business processes.
Since AI development is linked to cloud services the data is completely centralized. Giving too much of your data to one single company can lead to great problems when this data is exploited. This topic has been brought into mainstream discussion lately after Cambridge Analytica has used the data it obtained from Facebook to influence the outcome of the US elections. This incident gives great insight into how broad the consequences of Big Data misuse can be.
Lack of talent and a high barrier to entry
This technology is demonstrating its ever-growing potential and its development is actually rapid, but it could be exponential. Two major issues are slowing down the development of this space: the lack of AI talent and the high cost of AI development.
The lack of talent in this space means that the top talent is reportedly paid up to 1.9 million dollars. No startup can afford to pay that much. This means that only big, established and well-funded companies can afford to recruit good AI researchers which limits not only the speed of development but also the democratization of this technology. And that, given the great potential of AI is a serious issue in and of itself.
AI deployment is very expensive, in part because of the need for great amounts of computing power needed. The maintenance of large data centers is pretty costly. A high cost creates a high barrier to entry for startups, researchers, and developers that are trying to get started in this space.
How Nebula AI addresses the problems of the AI industry
Democratizing AI computing and storage
NBAI lets people pay for distributed computing power on shared machines with their tokens. This solves a number of issues. The most apparent among those is the efficiency since in this system no idle resources are being paid for. That’s in part obtained by sharing the resources with other users. But that’s not the only advantage of decentralization.
Decentralized encrypted data storage
With NBAI no central organization has access to all your data. That is obtained by the combined usage of IPFS, multiple private keys for access control and data verification which makes sure the data cannot be tampered with. This way you can be sure that your data won’t be misused or even accessed by who shouldn’t have access to it.
AI engineer training center in Montreal
NBAI has established an AI training center in Montreal where people with a background in mathematics or programming can learn fundamentals about artificial neural networks and how they’re being used for machine learning, as applied to speech and object recognition, image segmentation, modeling language etc.
A blockchain for AI computing
Most of us already do know that blockchain mining consumes lots of electricity and computing power. Actually, Bitcoin mining consumes more electricity than Ireland and it has been predicted that this blockchain mining will account for 0.5% of worldwide electricity use by the end of 2018. What many people don’t know is that this computing power is used only for the purpose of securing the blockchain.
That is a really huge amount of computing power that could be put to better use. That being said, the inefficiency of Proof-of-Work based blockchains is beyond the scope of this article. What I wish to point out is that such blockchains create a distributed computing network which is already really efficient at what it is doing. Sure, there are ways to operate a blockchain with way less computing power using other consensus algorithms. Still, the amount of computing power obtained by those networks is quite remarkable.
That’s the case because miners themselves profit more if their nodes are efficient. What NBAI envisions is that such miners could actually perform AI calculations instead simple hash calculations in order to obtain tokens.
Another big advantage of this approach is that many GPU mining machines can be converted to NBAI AI computing machines. This can potentially accelerate the network growth because many miners hold potential future nodes of the network. Still, since building decentralized networks takes time, NBAI has deals in place with large-scale third-party data centers in Quebec. This will ensure the system will be operational immediately when launched.
NBAI system architecture
Helix (Until 2018 Q3)
The first Ethereum independent blockchain (and phase) of the NBAI project is called Helix. Proof-of-Work (Ethash algorithm) will be used to secure this blockchain. Using an independent chain has many advantages. For the most part, with its own chain, the project will suffer from fewer traffic delays and some fundamental underlying proprieties of the blockchain can be tailored specifically for the very peculiar needs of such a system.
That means that the Gas (transaction) cost will be different in order to motivate miners to get profit through AI calculations instead of traditional Proof-of-Work mining. The way that it will work for nodes is that they can obtain tasks from the task pool by smart contracts. Then, after they performed the task, they submit the result and receive the token rewards. Another fundamental propriety which will be altered is the mining difficulty. This will be done in order to increase the speed of generating blocks and adjust the token production.
Still, PoW is really inefficient and has a limited scalability. Also, nobody ever solved Proof-of-Work’s 51% attack vulnerability. Those are probably the main reasons why this consensus algorithm is only used in the first phase of the project.
Orion (2019 Q1 onwards)
A fundamental property of distributed computing is that the closer the distance between nodes, the lower the cost of communication and the higher the computational efficiency. This propriety is the basis for the new consensus algorithm which is being developed by NBAI: Proof-of-Group (PoG). In PoG consensus systems and token incentives are used to ensure both efficiency and security.
In the PoG system, there are two different kinds of nodes: work nodes and ledgers. The role of a work node is to compute artificial intelligence tasks. The ledger, in addition to normal calculation, can also be responsible for allocating subtasks to all work nodes in the area (called a virtual working group). The task results are then written onto IPFS. After that is done the completed contract is verified by the ledger.
Self-organized efficient network architecture
In order to ensure an efficient communication between nodes of the virtual working groups an algorithm which creates a self-organized network topography which is efficient has been developed. Here is how it works.
When a new work node joins the system, it searches for nearby nodes. If it finds nodes with fast response time it joins them and becomes one of the worker nodes of the virtual working group. If the node doesn’t find any nodes with an acceptable response time it elects itself as a ledger. Nearby nodes will have a faster response time and will work together under a single ledger that coordinates them. Such a system ensures the highest possible network efficiency and generates an efficient network design.
No 51% attack vulnerability
The working groups compose a ledger network which uses the Byzantine consensus system for the joint ledger. This solution ensures complete safety against 51% attacks (which PoW doesn’t) and higher efficiency. Such attacks recently had some pretty bad effects on some cryptocurrencies. The most notable example is Bitcoin Gold (BCG).
Nebula AI use cases
There are a lot of potential applications for a more efficient and privacy-conscious AI computing service. Some possible applications are biomedical imaging, protein structure calculation (a generating task), computational marketing and various kinds of analysis of big data like social media feeds.
What’s more, this project offers more than just potential to speculate about. The team has developed some functional demos which run on their testnet that can actually show us some examples of what the system is capable of. Those are the aforementioned Quant AI and the Sentiment Analysis DAI Apps.
Demos offered by Nebula AI
One demo that the NBAI team has developed to demonstrate the potential of the system that they are creating is called Quant AI. Quant AI is a DAI App on the NBAI testnet which predicts the Ethereum trading price. This tool analyzes time series and trains deep learning models based on AI algorithm to forecast real-time trends and implement automatic trading of cryptocurrencies.
Advanced AI trading price prediction is something that has been accessible only to few people so far and now thanks to this initiative will become widely available and will be developed further. That’s the potential of blockchain democratization applied to AI.
The sentiment analysis DAI App is a natural language processing DAI App developed by NBAI. This tool helps users classify the polarity of a given text and extract the attitude of the writer. It is currently used as a price prediction model for trading, evaluation of consumer inclination, online conversations positioning and content inclinations.
Social media are gold mines of public opinion on any given topics. Such data can be monetized in various ways amongst which predicting market movements is the most obvious one. An example of how this tool can be used is given by this sentiment analysis of the subreddits of the main cryptocurrencies (the example given has actually been obtained using a less accurate algorithm which isn’t based on an artificial neural network).
This is one of the projects that look promising from the get-go but only get better as you continue getting to know them more. AI is a huge industry which is currently booming and is still so new that has a lot of space for improvement and suffers from various issues. NBAI sure won’t solve all the problems that this space is facing right now but it is actually addressing some of them.
From a speculative point of view, their token is just as promising as their project in general. Computing power is constantly getting cheaper (both from a hardware cost and electricity consumption standpoint) so the amount of work that you will be able to get done with the same number of tokens is going to grow over time. At the same time, also the number of users willing to pay for such a service is probably going to grow drastically. That is also confirmed by the AI service market value estimates.
More AI applications are going to be developed over time and more AI applications become possible as the amount of processing power available increases. It is hard to imagine how the NBAI token could become worthless. Sure, there is always the possibility that the whole system fails for some reason but the presence of actually functional product demos on the testnet makes it appear way less likely for this project than for most other ICOs. The ICO concluded on 20th April with $5,874,054 gathered.
No investment is 100% safe
Still, this system is under development and a lot of things can go wrong (Cryptocurrency and ICOs are always really speculative and risky investments. The NBAI whitepaper actually explains it pretty well at the very beginning in the “Risk statement” section.
One thing worth pointing out is that the very presence of such a detailed risk statement is a big plus since it demonstrates responsibility on the team’s part. You should never trust a project that claims your investment will be 100% safe and guarantees returns. That being said, the risk statement still says the truth and you should only invest what you can afford to lose.
Another thing to keep in mind is that there are some other projects that bring AI to the blockchain. Sure, every one of them has its own strengths and weaknesses but there definitely is some overlap. So, to put it shortly there are two main things that can go wrong with this project:
- The system could fail or an unsurmountable issue could stop the development.
- Too much of the market could be lost to the competitors or could not be “taken” from the centralized counterparts.