Decentralized Artificial Intelligence and Autonomous Bots (Auto-Bots) in Distributed Ledger/Blockchain Networks
Towards Decentralized and Localized Operations using Software Agents (bots) in Distributed Ledgers/ Blockchain
Today, most of the market and bot development are heavily centred around consumer and enterprise messaging bots. Bots are defined as software programs that execute repetitive tasks. The boom of messaging bots just came about recently with the introduction and adoption of machine learning/artificial intelligence (AI) algorithms that automate bot tasks. These new computer algorithms is what makes these “little software engines” so popular these days. One example is the Facebook AI messaging bot.
At a first glance, it seems that these are just some dumb software coding, sitting in our computers or website, and interacting with users through some text messaging or web tools. But its more than that. This is just the beginning of a more complex technology (re) evolution that is about to get started in our society with the convergence of decentralized AI and Blockchain. The text below describes some of these key evolution aspects and its challenges.
Bots Gets a Real Life Presence
Before we will ever see some real “hardware” or “physical” robots, circulating in our daily lives, executing some meaningful tasks in our houses and work, we will first experience the multi-task “software bots” in action. This bot category will be functioning in our home and work environment, inside our computers, executing and speeding up our daily repetitive tasks.
The next move is to incorporate these “software-agents” inside machines to handle queries, provide information or execute some small tasks. It will first appear in our appliances, media and communications tools, including our vehicles, and later will evolve towards more specialized enterprise applications. Bots, with Natural Language Processing (NLP), such as the google assistant, are currently being incorporated into our mobile phones and home devices. Today these bot assistants can interact with homeowners and home “things”, enabled by Google AI centralized cloud-based servers. The way these home bots work, with some NLP technology inside, is using our Wi-Fi networks to access the massive amount of Google central data and processing power available at their distributed central cloud servers to process all the big data using Google’s AI algorithms and produce the answers we need.
The Expert Bots: Specialized Enterprise AI Bots in Action
The next bot technology wave to be introduced will be the expert enterprise bot systems, based on applied AI, accessing some specialized knowledge data base. These are industrial or enterprise-grade bots, designed to execute specific industrial tasks, such as monitoring an industrial valve, checking temperature and making decisions to shut-down some element or system in case of potential failure. Some configuration of these bots will reside inside expensive industrial machines that work and act locally in an edge-computing environment.
These edge bots will be AI-driven, also called AI-engines, and capable of running smart contracts to communicate and log events in the system ledgers of Blockchain networks. The industrial grade-bots will work as “little brains” inside expensive million dollar machines to automate field operation or industrial processes.
AI Engines/Bots are Fuelled by Big Data
Large amount of data, or big data, is the real fuel that underpins AI. AI Algorithms are applied to interrogate, correlate and process this large amount of dataset and come up with predictions and forecasts with a high level of accuracy. This is an iterative process or control feedback mechanism required, particularly in some specific Deep Learning algorithms, using neural nets, where new data is absorbed and processed as the predictions are proved, and the results are fed back into the system to further improve and refine the neural net arrangements.
In most cases, results allows machine learning (ML) algorithms being tweaked to become more and more accurate. This tweaking process is known as the “AI-maintenance mode” that needs to occur frequently if data input is changed over time due to algorithm accuracy decrease if AI engines are not under regular maintenance by expert teams of data scientists and ML engineers. So in short, data drives algorithm processing that provides predictions, which are validated and fed back to improve the algorithm, which in turn improves predictions, and so one. It’s a complete self-reinforcing feedback loop mechanism
AI is Driven by Web-Centric Dotcom Centralized Databases
The first AI bot evolution relies on centralized data servers to provide massive amount of cloud computing data that fuels the AI algorithms for better and accurate predictions. For example, Google and Amazon implement AI and control this technology in part due to the massive amount of data, captured by their distributed centralized data servers, using the different web tools they offer to end users as a service (search engines, advertisement, email, video tags, etc.).
In a similar analogy, today’s AI bots are centralized and relies solely on this cloud database solution to process their AI data. Start-ups developing machine learning/AI algorithms currently pay millions of dollars of cloud-based costs just to test and develop their own AI engines.
AI Goes Decentralized with Blockchain Distributed Ledgers
The next challenge is how to make AI systems decentralized, with low data processing costs, and still access lots of data to function properly? This is where the Blockchain provides an answer. Moving AI systems on a distributed platform via the Blockchain allows these AI engines to connect to a network that has distributed processing power and off-chain distributed storage capability to run these algorithms in distributed ledgers, at the fraction of the cost of centralized cloud infrastructure, enabled by next generation decentralized computer data technology, the Blockchain/DLT.
In a distributed ledger configuration, decentralized bots can have access to data everywhere. They can go and connect in a peer-to-peer mode to decentralized servers, decentralized exchanges or any decentralized data feeder. These bots can even autonomously connect between themselves, in a P2P mode, to exchange and share their own data. A P2P data sharing rewarding mechanism, enabled by crypto-assets, will be important to establish the basis for this data value exchange economy.
Here there is no central authority ordering or organizing this data sharing interaction, only mechanisms of consensus that establish some clear rules and protocols to make this a transparent and trusted environment. That’s the principle of Blockchain.
Hence, the natural progression of AI bots will be a combination of bots accessing data from hybrid systems — so some data may come from a centralized data server and others from a distributed ledger network infrastructure. This combination will drive AI to the next level — to the holy grail of decentralized AI.
Challenges of Blockchain Decentralized Infrastructure
However, the Internet-based Blockchain as we known today cannot yet completely support this level of infrastructure decentralization as described. The reason is that there are not enough technology platform available yet to support this transition. This will take time as more innovative and disruptive decentralized edge, and cloud computing Blockchain/DLT solutions, including off-chain storage, are introduced to the market.
In the next three years it is expected a high level of decentralization of Internet infrastructure platforms (computing and storage), with IT web-centric dotcom experiencing some level of businesses disruption. A hybrid centralized server, combined with some degree of decentralized approach, will be common place during this transition.
Yet we are still in the early stages of this technology. There are still other issues facing decentralized AI solutions. The next big challenges and evolution for AI is how does these decentralized AI engines will handle real-time data processing, vast swathes of it, with high degree of scalability in P2P architectures? So this is also related to the Blockchain processing speed. It is well documented that transaction speeds on the Blockchain are sticky. This also poses problems with scalability. This issue is currently being addressed by the Blockchain developer’s community and new protocols are in place to improve upon this. The other issue related to this question is to have AI engines working with static, historic data from either centralized or decentralized databases versus real-time data coming from the same infrastructure to feed their data algorithms and neural nets, and a combination of these two alternatives
Edge-Based AI Engines
Another step towards the bot technology evolution is to add more computing, intelligence and processing/storage power to the edge/fog, in last mile network segment. In this configuration, AI engines will rely on a multitude of Internet of Things (IoT) sensors and actuators, collecting and processing vast amount of operating field data. Streaming IoT data will therefore, fuel the “localized” edge computing AI engines that will run local process and make decisions on the spot. This requires another level of real-time edge computation, data gathering and storage that will drive AI processing to the edges of the network. Events will be triggered and written in smart contracts that will log and register anomalies or any time-series events, that needs to be considered by mission critical operation. This will complete the cycle of cloud-edge intelligent and networked machines, running micro-transactions over ledger-based smart contracts.
AI Crypto-Trading Bots
Among all these bots categories, the AI crypto-trading bots, particularly the ones that are completely decentralized and operate on distributed ledgers, are the most sophisticated ones. These are the bots that will constantly look to interact with other bots in peer-to-peer mode, access centralized and decentralized data feeders to fuel their algorithms and make intelligent predictions and eventually decisions, based on data correlation of a multitude of data sources. A good example of use cases is in the daily trading world where the crypto markets never sleep. This level of data complexity in terms of origin and sheer volume is forming the foundation of the newest decentralised autonomous software engine or bot where speed and real time data processing is essential to cope with tremendous crypto-current market volatility.
The Distributed Ledger Network Infrastructure
Decentralized AI bot engines need new decentralized computing power and storage that will form the basis of the new decentralized Internet network infrastructure that is not existent today. This next generation Internet will be made of these computing network elements, based on peer-to-peer (P2P) configurations, that will boost the development and adoption of AI decentralized distributed applications (DApps).
As the Blockchain network infrastructure evolves to improve transaction speed, increase performance and improve security, more web-centric dotcom businesses will migrate to this distributed ledger to reduce IT costs, improve cybersecurity and create new distributed services and applications.
The end goal of all this with AI engines is to move AI processing (the algorithms and neural nets) and their “fuel” (the data) to a completely decentralized infrastructure with no central data ownership. This will be the beginning of the creation of a “super global Blockchain-enabled computer” with a large distributed brain made of decentralized processing power and storage capability, extended all the way to the edges, with IoT machines working in synch and in real-time with their peers to automate processes and operations. We call this Interchain.
The Blockchain Engineering Council — BEC is working on next gen Interchain engineering technologies, which is the evolution towards a decentralized Internet technology and the creation of unique applications on distributed ledgers.
