Home
aiCubicle is all about Artificial Intelligence & associated technologies, human-machine singularity and the direction it will take at a philosophical level.
It is about understanding where this field is today, and maybe getting hands-on to understand the technology. It is about understanding the massive shift that is underway in all walks of life, musing over the next generation use cases, and what it means to us as we develop a powerful "species" of AI driven machines.
Tomorrow's AI driven ecosystem is not complete without critical emerging support technologies like the Internet of Things (IoT), Blockchain, Cloud Computing and Quantum Computing.
The set of algorithms develop and update internal constructs that learn from information over time and they need not be fixed or finite. These internal constructs can improve over time so that the output is as per expectations or as close to that observed in the real world, with increasing accuracy. The development and improvement can happen without human intervention - like humans when they learn from experience or by "just being there".
The internal constructs of the algorithm need not be derived scientifically but may evolve and fall in place through learning from vast data. The algorithms automatically determine the internal constructs like equations, variables, coefficients or other complex structures and logic-flows to process information as they go through the data or the experience. Thus, there is an element of constant learning or a feedback mechanism so that output accuracy either improves or continues to remain relevant. Hence, "intelligence” as opposed to being dumb and incapable of improving once coded by a statistician or a scientist. Once the algorithm has stabilized (to produce reasonably good outputs), feeding it with inputs can then produce outputs that can be of value to us.
Artificial intelligence is set of algorithms that perform tasks normally requiring human intelligence (and in several cases do better).
IoT enables the communication ability among machines so that they can collaborate.
The role of Blockchain in this ecosystem is to establish a system of global truth - so that the knowledge and experience developed by the AI is immutable. Also, it ensures that the ecosystem never forgets how it has come about to its current state so that it can avoid pitfalls in the future or revert to a better "self" if required.
Blockchain enables permanent memory that is incorruptible. It never forgets!
As it stands today, Cloud is comprised of massive data storage and computing capacity distributed across multiple locations (called "data farms"). The data of end-users are stored and processed at these farms - though data is replicated to introduced redundancies, thus avoiding service failure to end-user if a farm was to fail. The end-user also is not able to precisely point out which data farm or computer is processing or storing his data - thus the location of physical infrastructure is anonymous to a user. Given that the data is redundantly spread across several computers and/or locations, Cloud algorithms ensure that the end-user is served from the best possible location to provide optimal speed and access times to the users. Theoretically, data on Cloud cannot be destroyed and is always available.
Cloud computing gives AI machines the ability to live for ever. It is always online, because it is in the Cloud.
It is about understanding where this field is today, and maybe getting hands-on to understand the technology. It is about understanding the massive shift that is underway in all walks of life, musing over the next generation use cases, and what it means to us as we develop a powerful "species" of AI driven machines.
Tomorrow's AI driven ecosystem is not complete without critical emerging support technologies like the Internet of Things (IoT), Blockchain, Cloud Computing and Quantum Computing.
Artificial Intelligence
AI is a set of algorithms that provide the best output estimate when provided with some inputs - which would usually require human like intelligence i.e. the relationship cannot be expressed as simple mathematical equations. The relationships are multidimensional and multiple layers "deep" in dependencies which usually manifests in humans as knowledge through experience (which roughly maps to the "depth" of reasoning) or gut feel (which is again based on experience which is "deeply" ingrained but not easily explained).The set of algorithms develop and update internal constructs that learn from information over time and they need not be fixed or finite. These internal constructs can improve over time so that the output is as per expectations or as close to that observed in the real world, with increasing accuracy. The development and improvement can happen without human intervention - like humans when they learn from experience or by "just being there".
The internal constructs of the algorithm need not be derived scientifically but may evolve and fall in place through learning from vast data. The algorithms automatically determine the internal constructs like equations, variables, coefficients or other complex structures and logic-flows to process information as they go through the data or the experience. Thus, there is an element of constant learning or a feedback mechanism so that output accuracy either improves or continues to remain relevant. Hence, "intelligence” as opposed to being dumb and incapable of improving once coded by a statistician or a scientist. Once the algorithm has stabilized (to produce reasonably good outputs), feeding it with inputs can then produce outputs that can be of value to us.
Internet of Things (IoT)
Today we have AI based solutions that are confined to a carefully curated environment - this is in part due to limitation of algorithms and the computing power to process data if obtained from a larger environment. Thus AI solutions are limited to certain tasks or category of tasks. A "thriving" AI would be able to get data from various sources, talk to various other machines and then take a decision, receive help of provide one to other machines. This is something that humans are good at - we can communicate with others and thus collaborate to create solutions. Machines can do this too, if they communicate with each other. While humans are a homogeneous group as a species, machines come in all shape, sizes (and even be virtual like a software running in the Cloud). If machines can communicate, they can collaborate to achieve a solution (good or bad for humans - it depends).IoT enables the communication ability among machines so that they can collaborate.
Blockchain
Blockchain today is a technology that guarantees immutability (and hence traceability) in a value chain. The value chain can be suppy chain, document and transaction chain, etc. For AI, the value chain is "knowledge" it has accumulated from interacting with its environment.The role of Blockchain in this ecosystem is to establish a system of global truth - so that the knowledge and experience developed by the AI is immutable. Also, it ensures that the ecosystem never forgets how it has come about to its current state so that it can avoid pitfalls in the future or revert to a better "self" if required.
Blockchain enables permanent memory that is incorruptible. It never forgets!
Cloud Computing
Cloud technology allows an "always on" mode of operation for AI machines. Even if a part of the broader AI ecosystem is offline or destroyed, the machine still lives on. Cloud also allows heavy processing to be outsourced to the experts who have the technology and resources to provide the level of computing power required to run AI algorithms.As it stands today, Cloud is comprised of massive data storage and computing capacity distributed across multiple locations (called "data farms"). The data of end-users are stored and processed at these farms - though data is replicated to introduced redundancies, thus avoiding service failure to end-user if a farm was to fail. The end-user also is not able to precisely point out which data farm or computer is processing or storing his data - thus the location of physical infrastructure is anonymous to a user. Given that the data is redundantly spread across several computers and/or locations, Cloud algorithms ensure that the end-user is served from the best possible location to provide optimal speed and access times to the users. Theoretically, data on Cloud cannot be destroyed and is always available.
Cloud computing gives AI machines the ability to live for ever. It is always online, because it is in the Cloud.
Big Data Analytics
Big data analytics is about uncovering information such as hidden patterns, unknown correlations, trends, causal relationships using complex, multi-dimensional data. With growing digitization of the world around us, immense amount of data is being generated every minute. IoT promises to increase this exponentially. This data is useful only if we can extract information from it. Data mining is used to sift through the data in search of patterns and relationships. This is helped by artificial intelligence technologies like machine learning and deep learning. While humans are accustomed to visualizing (and hence understanding) data in maximum three dimensions, these techniques can interpret data in multiple dimensions which may not be comprehendible to the human minds. Indeed, most patterns in nature and several scenarios is business have multitude of variables interacting with each other and articulating the patterns in which those variables interact is not possible through the traditional mathematical models. Once the data is mined, predictive analytics is then used to forecast future behaviours.
Big data analytics is used to extract useful information from data. The information may or may not be easily comprehended by humans.
Quantum Computers
The processing power today is not sufficient to run the algorithms of today and achieve the true potential of what AI can do. Either the algorithms need to become more efficient or the processing power needs to rise exponentially. While the algorithms will their due course of time to be designed by humans (until we have AI creating better AI), there is light at the end of tunnel for computing speed. Quantum computing promises to simply multiply the computing power which can then enable the current AI algorithms to achieve much more. The Idea to Reality machine (as described in the Ideas page) is just not possible with today's computing power.Quantum computers gives speed to AI machines to make complex decisions quickly. It increases the speed at which it thinks.