Applied vs generalized AI where do we fit and what is the difference
This is a republish of an earlier released article (circa-2018) referenced here,
Note: ‘xxx’ referenced the ceased project which is now decentralised across the Enix Blockchain — purpose is to provide a link to the prior content at request and for academic referencing.
There is no shortage of AI related news making headlines these days, however the hype/buzz of last year is thankfully declining, with the waiting period for many a investor in equities awaiting to see the results of their investment(s), the push/buzz/hype for 2018–2019 is specific to the Internet of Things (IoT’s) — Actually a segment xxx fits into, but another altogether different article. Going back to AI, predominantly articles about the rise of evil robots, others envision a post-work world where AI renders human workers obsolete. These futuristic stories are entertaining, however they can make it difficult to rationalize the current state of AI.
The reality is that, today, all AI systems rely heavily on data. Contrary to what the media would lead us to believe, AI systems don’t operate in an endless vacuum, though in part xxx does, via automation (ANI). How these systems are designed, constructed and maintained determines their success, by all means, good data = good results, we could just as easily fit in the bad data = bad results bracket, however this is where ANI, AGI comes into play with our bespoke hoop-loop-process created by xxx that is built specifically to cleanse and multi-confirm data relevancy and accuracy, and specifically xxx core backend interest in blockchain, with the recording of such at scale on its own bespoke blockchain framework commonly referred to as Enix blockchain.
All AI is not created equal
Amidst the assault of industry jargon introduced over the past few years, even some of the greatest visionaries in the field of AI have struggled to clearly define the technology due to its broadness and crossovers. However, most experts agree that AI can be split into the categories of Artificial General Intelligence (AGI), Artificial Applied Intelligence (AAI) Artificial Narrow Intelligence (ANI) and of course for the future Conscious AI (CAI — True AI), based on its level of intelligence.
Below I’ve provided brief explanations of the various and where xxx fit’s in.
Artificial General Intelligence
The AGI Society defines artificial general intelligence as “an emerging field aiming at the building of ‘thinking machines’; that is, general-purpose systems with intelligence comparable to that of the human mind.
This unconfined form of AI is intended to demonstrate understanding and reasoning skills with a breadth and depth of knowledge that allows it to easily traverse between vastly unrelated topics and use cases, just as a human can.
Where does AGI fit in with xxx? — We use what we consider as Mimic AGI extensively across all of our modules, however keenly through our core comprehension AI to interpret raw content on the fly or generic processing, both to comprehend (context) and match against seemingly unconnected content on first process, or secondary…
Artificial Narrow Intelligence
The majority of currently active Artificial Intelligence is actually Narrow AI. Narrow AI is usually code that is automating a human activity, and in most cases it outperforms humans in efficiency and endurance due to multi-core processing and cloud power/supercomputer power — or even system power, we only have two eyes, two hands… .
Where does ANI fit in with xxx? — Just about everywhere, due to the nature of the sheer volume scale of data processing we have in place, we’ve essentially pushed ANI into every backend segment of our system, it’s what keeps everything at XXX ticking along, If we could automate employee tasks further we would!
Artificial Applied Intelligence
Artificial Applied Intelligence is commonly defined as an application of artificial intelligence to enable a high-functioning system that replicates and, perhaps, surpasses human intelligence for a dedicated purpose. With a rather vague definition, not only are rule-based chatbots and Robotic Process Automation technologies often miscategorized as being applied AI, there are also significant variations between the capabilities found across applied AI systems in general.
By incorporating machine learning (ML), natural language processing (NLP) and tight integrations with external systems of record, more advanced forms of applied AI go beyond scanning knowledge bases and automating routine tasks to intelligently correlate and expose valuable information and services to users in real time. What’s more, the unification of these capabilities enables autonomous agents to become increasingly conversational, allowing them to make recommendations and/or trigger workflows to fulfill requests for both support agents and business users.
For example, an advanced applied AI system can gather employee background information such as user name, title, location and authorized resources from across disparate data sources. With the ability to efficiently surface, analyze and aggregate this type of information, support departments can more effectively apply service delivery best practices and automate routine tasks such as resetting passwords, enabling system access, answering benefits and/or payroll questions as well as escalating users to the appropriate support channels.
Additionally, autonomous agents using ML can be trained using internal and external data to not only answer questions, but also self-learn and improve responses to users’ inquiries.
Conscious AI (True AI) — Early phase to ASI.
As a company, we stringently believe that conscious AI (Sentient) is far off, possibly achieved in the midst of conflict and a depression, due the sheer volume of funds that would be pushed into such development and the total intellectual minds committed to a common goal — survival, as grim as this sounds, the greatest achievements within scientific remit are generally created amidst chaos and misery.
At this time, we do not fully grasp (scientific community), or understand what makes us as humans tick along — some believe the use and advancement of Deep Learning/Machine Learning will lead to the generation of True AI, i agree and disagree, both will result in a more true-variation of AI however i feel conscious (sentient AI) is still far away off, in addition no chips in the market today are as scaling or on par with the human neural network capabilities/speed, in addition the consciousness of humanity is still in early research in terms of the ability to both comprehend and create, until then, we essentially mimic consciousness (ourselves) in different capacities, for instance in the middle of our systems is a comprehension AI which is mimicking conscious decision, determining all content directives feeding through each and every limb of our system, this by no means, makes xxx “True AI”, however the neural framework of our overall system is built to mimic humanity as a individual but in a collective fashion, with sensory outreach, internal ethical/moral, intellectual, experience based judgments based on baseline parameters, and deduced by that of our comprehension vehicle named Jean.
Artificial Super Intelligence (ASI)
So above we hit on Conscious AI, considered true AI, however then the next step is ASI, this is essentially the holy grail in AI Theory, no longer mimicking Human behavior or actions, comprehension or other, this is the area that AI surpasses that of human intelligence, or that of mimic, we believe firmly this is decades away, we first need to surpass the early consciousness stages, xxx in no manner at this time is within this remit.
— -
AI companies in Blockchain Comparison
1)Nebula AI https://www.nebula-ai.com/
Predominantly blockchain based however with features of ANI & AGI, though not in anyway can be compared to xxx, specifically on the grounds that the full offering of Nebula, is equivalent to a mere few modules within the xxx backplane, and in addition, a few data-sets. However their monetisation capabilities via the blockchain smart contract is not something xxx offers (at this time).
QUANT AI
Analyzing time series and training deep learning models based on AI algorithm to forecast the real-time trend and implement automatic trading of cryptocurrencies.
SENTIMENT ANALYSIS
Building natural language processing algorithms on the blockchain allows the user to classify the polarity of a given text and extract the attitude of the writer. All sentences sent on the chain would be recorded.
WRITING RECOGNITION
By using multiple neural networks, Nebula AI receives and recognizes intelligible handwritten inputs such as characters, words, and phrases.
These would be using both ANI (collection, sorting, cleaning), and AGI (Machine Learning & Deep Learning) from comprehension-context to decision making through the automation process.
2) Singularity.Net — https://singularitynet.io/
The most interesting of the AI Blockchain ICO’s i came across, lead by renowned scientist Dr Ben Goertzel, when i first looked at singularity i was lead by the marketing to believe it was specific to the robotic push of Sophia, which in terms of technology flow at the time, i wrote off as standard content analysing with AGI with baseline rule responses.
Still there appears to be a lot of emphasis on Sophia though i’m not entirely sure in what context Sophia experiment is involved in the project.
However post ICO i’ve noticed the site now has been restructured inline with product deliveries and their offerings are more apparent (without having to read the extensive technical/white paper) to determine the forms of Artificial Intelligence they offer, their scope and in what context they are pushing our solutions, services or other.
Ok so what is Singularity offering, a scaling environment at the heart of it, however let’s break down their solutions highlighted here.
https://singularitynet.io/services-roadmap/
Network Analytics
The SingularityNET Network Analytics will comprise of statistical, neural net, information-theoretic, probabilistic, and logical AI tools to discover patterns and properties in complex networks.
Social Media Analytics
For analyzing social media content, networks, and relationships, we will develop SingularityNET Social Media Analytics.
= xxx completes this automatically and extensively.
Biomedical Data Analysis
The SingularityNET Biomedical Data Analytics tools will be developed in partnership with Hong Kong-based Mozi AI Health.
= Biometric data analysis is something we have an interest within (analysing) however Biomedical is something xxx certainly have no interest within.
GHOST-Powered Chatbots
SingularityNET GHOST-Powered Chatbots allow users to create conversational Agents capable of verbal and non-verbal interactions. By using our custom developed programming language and authoring tools, users can build their own conversational agents like Sophia.
= Bots is something xxx is refraining from involving itself in.
Social Robotics
Our SingularityNET Social Robotics tools and services are used to control Hanson Robotics’ Sophia Robot. Developers will be able to integrate our technology into their own robotics projects in order to create social and emotional robots like Sophia.
Essentially similar to the API’s to be offered by xxx but in another directive, analysing of content (data) extraction of reasoning (context) and actioning of, however something xxx does not hold any interest in.
DNN Model Evaluation & Tuning
The SingularityNET DNN Model Evaluation and Tuning will give developers simple services for evaluating deep neural nets on standard image, audio, video, and natural language problems. We will also allow clients to generate their own models by tuning existing models with their own data.
Model side of Neural Networks, although we utilise our own frameworks / environments, xxx at this time currently possesses no interest in opening our architecture for 3rd parties outside of state.
DNN Model Training & Neuroevolution
By using our SingularityNET DNN Model Training and Neuroevolution, we will allow clients to train neural networks from scratch, using both standard NN techniques and evolutionary computing.
Second Phase of DNN Model Evaluation & Tuning, again out of scope of xxx’s offerings, i’d compare this against Google’s or IBM’s offerings.
By far Singularity is the most complete offering i’ve come across within the ICO sphere, i see them as long-term subject to additional funding rounds, moving their architecture from the cloud to their own data-centres worldwide as a challenger against the offerings of Google Vision, Microsoft Vision, Amazon AI offerings (API’s), in addition xxx commercial offerings, however only within API segment of xxx offerings. Thus we do see this as a possible long term healthy competitor.
— — -
http://www.rayadvisors.com/blog/artificial-intelligence-in-depth
https://codeburst.io/four-types-of-ai-6aab2ce57c19
Now let’s look at the considered four types of AI
Reactive.
The manner in which reactive algorithms work is in the context that data flows in and reactions are made, Reactive algorithms do not in any capacity utilise past experience (Memory, Machine Learning) in addition due to the manner in which the reactive process works, the flow is unable to to refer to prior experiences, thus unable to improve with practice.
Best manner in which to understand this is within the following linear approach.
Trump tweets about BA (NYSE), the content of which is pulled up by a tweet grabber (scraper), to which it is analysed via keyword and sentiment analysis, determining the impact on the stock price, immediately actioning a api script to a trading console to either long/short the company.
This is handy, however without the use of prior experience (example : historical similar data) taken into account the impact of the tweet, is not taken into account, thus actions .
Limited Memory
Limited Memory is the method in which data is retained for a short period of time for actioning processes off. While the machine can use this data for a specific period of time, it is built in an manner where it can not add to a library of their experiences.
Theory of Mind
As humans our behaviour is believed to be driven predominantly by thought (calculated), emotions (calculated), memories (taken into account — pulled into the calculation), and ofcourse models driven by mental-model process, in generic terms the smartest people in the world, process, thus deduce and make decisions based on mental models within their brain, naturally, every human is different, intelligence levels are different, emotional levels are different, memories can be personified biased driven or different recordability levels thus different, and finally thoughts are the end results of the prior stated thus different, however generically speaking humanity as a whole processes and makes decisions within the same process, some harness it more than others ands control input, others do not, it depends in the most easiest sense on wiring, this is very important when it comes to AI, as AI theory of mind is the encroachment towards ASI, as such going back to my earlier statement, bad data in = bad decisions above, at the heart of it, that is purely from a data standpoint, but you also have to consider a machine standpoint, in essence the person creating AI ToM is injecting parts of their self into the system, either intentionally or not. Which in return has a host of knock on effects, as the AI is a mimic of humanity, and humanity is by no means perfect, i myself have many flaws, one being my commitment to this project at the cost of my own health or that of my close ones, a AI that is built to mimic my traits would utilise the same decision flow if such were built around my process flow, as an example.
At this time (Sophia an example i believe) as long as not a bot, is able to mimic, learn, and imitate our mental models, with this system (in public does not exist, in any form / phase) the relatability to humanities is perceived and with impact’s understood because everything is linked or has a knock-on effect.
What i can say is, that at xxx, we are able to read the events as they unfold, we are able to determine the facial expressions, the sentiment and emotional impact to events as they happen, and in part specifically able to pre-judge where the effects would be most felt from the knock on process, however I do not see this as ToM state but through the use of multi-agent AI processes in the Hoop-Loop Process.
Self-Awareness
Self-aware (sentient) i hit on earlier above, this i believe is far of due to technical breakthroughs humanity needs to pass through first.
However the faux mimicking of such is evident in traces across multiple projects…
— -
So hopefully we’ve addressed a few technical elements above, in relative terms people must be curious of my (xxx’s) end goal, after all, when you lay down all of xxx’s offerings, it’s akin to looking at Google minus the billions of dollars, infrastructure, staffing, greatest minds or other, after all there’s not many projects in the AI ICO space, let alone the ICO space in its entirety with so many arms, the truth really comes down, to the people that i brought under my wing through the last two years prior to entering into an ICO, being honest, for me, this has never been about wealth, i had that prior and as it stands with the market, i’d be at a loss regardless, no xxx for me has always been towards the mimicking (at first) ToM to the scale of the Machine, an autonomous driven intelligence system, which at the foundational level is part and parcel with what xxx is or rather where it is going, as a group, you may wonder why the limbs of products, the reality is with the system that has not been built over many years, a lot of capabilities thus products were created as byproducts, kinda, like something i am working on currently but not at liberty to release at this time, the financial elements, it should be noted, as stated above are not an area of interest i personally have bar the idea of funding further developments through their success should they be actioned, i’ve got family in finance, seems that it runs in the blood if at Christmas, however, my intentions are more modest, the ability through data, calculated decision making at scale, in masses to safeguard and the long term protections it affords us…
