Blog is now fully self-building, self-optimising, self-evaluating and self-learning continual improvement Deep Learning AI

Fennaio | blog post


09-10-2021 is now fully self-building, self-optimising, self-evaluating and self-learning continual improvement Deep Learning AI

We built to be as genuinely intelligent and automated as possible so that actual AI is accessible to all; not just coders, developers, Data Scientists, Machine Learning experts, and AI monopolies.

This was done so that anyone can quickly get AI up and running with no/minimal knowledge and to of course massively reduce the cost of AI implementation.

We achieved partial automation early on in's evolvement - a user just had to provide data, choose default settings and tell eldr's dynamic Artificial Neural Network to learn.

This is great, obviously due to the simplicity of creating extremely complex and powerful AI automatically, but secondly because it saves users pots of money and tonnes of time.

We've now made it even more automatic, and powerful - literally upload your data, and if you want, you can select "full automation" and eldr will self build, self learn, self optimise and continually get better and better.

Here's what, why and how we created true and actual Artificial Intelligence that's one click, turnkey no code AI technology.

Self-Building AI

What is Self Building AI?

Artificial Intelligence that's truly no code means that it builds and manages itself; no development is required at any level.

Why use Self Building AI?

Building AI is complex, costly, timely and to do it correctly, accurately and efficiently requires specialist knowledge, technical ability and patience - there are literally billions, actually an infinite number of ways, of building and configuring AI, and all require code.

How is eldr AI self building? learns using artificial neurons arranged in an artificial neural network (ANN) of layers - roughly mimicking how a human brain works at the basic level.

Artificial means these ANNs are created with a machine or computer - in the form of code. has been trained to write its own code and builds its own ANNs using self-governing learning and completely dynamic self-coding.

You can opt to instruct to build itself completely automatically or you can specify all the parameters yourself e.g. neuron number, layer count, ANN size, activation, learning rate, drop out, batch normalisation and others.

To make truly and honestly self-building and no-code, we combined four key features: self-optimisation, self-building, self-learning and continual improvement.

Self Optimising AI

What is Self Optimising AI?

For AI to be truly intelligent it needs to do what natural intelligence does; adapt. Adaptation includes optimisation, or in the case of Deep Learning AI, making the learning process optimal and more efficient each time, fundamentally increasing efficiency of learning and the accuracy of predictions, recommendations, insights and all round decision intelligence.

Why use Self Optimising AI?

As a global community we've embraced AI to take over tasks that usually require human intelligence, and in most cases we want AI to perform these tasks faster, smarter and autonomously. To do this is costly in terms of computing power, time, money and culture shift. This is why self-optimises; to work faster and smarter at the same time as reducing resource demand - and without the need for human involvement as best as possible.

How does eldr AI self optimise?

In traditional software development, performing self-optimisation involves stepwise code. It's not intelligent, it's just doing the standard "if, when, do, this, trial, error" etc. It could eventually get there, but it's not efficient and fundamentally, this is why AI is far superior - it makes all links and judgements between data (in this case learning parameter data) without explicitly telling it what to do.

Using AI for self-optimisation could be achieved 3 core ways: (1) Data Science, (2) Machine Learning or (3) Deep Learning/Artificial Neural Networks. Data Science involves Mathematicians/Statisticians looking for patterns in data - this isn't AI. Machine Learning is getting there in terms of AI; it uses code (the infamous term - algorithms) to look for specific patterns - and is therefore not fully AI as it requires specific rules (in general). Deep Learning AI on the other hand, using Artificial Neural Networks is AI proper as no specific instructions are required.

Luckily for us is itself a Deep Learning AI entity that self builds so is perfectly suited to self-optimisation. And that's exactly what eldr does - actually it did and continues to do so in the background. self-built an ANN that continually optimises itself. Within the ANN are all the times it has learnt anything, the parameters, the outcomes, the predictions, the accuracy - everything. It continually uses this ANN in conjunction with any newly-built task-specific ANN to work out how best to approach any given learning task.

In the deep depths of its learning core, somewhere within the server, will start thinking differently, applying various and a pretty much infinite combination of ways of learning and work out what works best given a particular problem to solve - this is effectively Reinforcement Learning.

This makes one of the most powerful - if not the best AI in the world.

Self Evaluating AI

What is Self Evaluating AI?

We want to exhibit genuine intelligence as much as possible. People often talk about self-awareness in AI and one important aspect of this is self-evaluation. The continual question of "how am I performing?" and the frequent thought of "I can be better" are normal in human brains but they certainly aren't in computers or machines. Human intelligence generally tackles self-evaluation by comparing and contrasting a desired outcome with the actual outcome, and then making necessary adjustments to reduce the gap between the two - and importantly - do this autonomously/inherently with no outside interference. This is what we require from AI to be really classed as intelligent.

Why use Self Evaluating AI?

Aside from the fact with eldr AI we don't want to use the term AI/Artificial Intelligence when something isn't actually displaying any sign of inherent intelligence (the term AI is banded around a lot these days and more often than not it is used generically to describe Machine Learning as opposed to real Artificial Intelligence), having self-evaluating AI massively improves output and accuracy in terms of predictions, recommendations, insights and effective decision making, and importantly reduces resource demand on every level - you could say the exact same things about how naturally occurring intelligence works.

How does eldr AI self evaluate?

Self-evaluation carried out by eldr AI is similar to how it carries out self-optimisation; it uses a dynamic ANN that continually asks "How did I perform?" and it focusses on accuracy. In self-evaluation, scrutinises two types of learning-derived accuracy - (1) Re-running the core data provided initially and checking if the predicted outcome matches what it was trained to do. (2) Using a partial subset of data it has never seen before to see if it can predict something it hasn't actually learnt. That's just the basics. If the accuracy isn't quite right, will inherently try out different ways of thinking, linked with the self-optimisation ANN, constantly striving to get an accuracy of 100%. Each and every time learns something, it adjusts its own self-evaluation ANN - it literally gets better and better at self-evaluating and feeding back its own critique for the next AI task - exactly how a human would do it.

Self Learning AI

What is Self Learning AI?

At its heart, or brain, AI needs to behave like a human, and that means learning. A human is fortunate enough to be born with the incredible machinery that is the brain, made up of neurons all linked to each-other in various domains that crucially have the innate ability to be "plastic". This neuro-plasticity is key to how humans grow their intellectual, social, movement, memory and emotional abilities - it means that neurons and the circuits they are involved with can change and respond in order to learn via stimuli, inhibitory factors and feedback. Neuro-plasticity is therefore something that AI needs to display in order to be classed as intelligence as we know it. This neuro-plasticity is the basis of learning and this is what Self Learning AI is.

Why use Self Learning AI?

A system cannot describe itself as intelligent if it can't exhibit artificial neuro-plasticity. This is why generic software isn't AI - it will always produce the same output with the same given input because it's told what to do. Self Learning AI, or AI that boasts dynamism in the way it learns, and can literally learn via change, dynamic behaviour and plasticity with no outside involvement, is about as close to human intelligence as we can hope to achieve.

This raises a debate as to what AI actually is, and if when we hear the term "AI" is it truly intelligent? My opinion is that only a very small number of "AI" systems or solutions are actually intelligent; they are generally Machine Learning algorithms trained to look at a specific dataset using Data Science - this is not AI, so let's not pretend they are.

eldr AI on the other hand can confidently define itself as Artificial Intelligence due to its ability to self-learn via dynamic Deep Learning AI.

How does eldr AI self learn?

Each time eldr is given a task to do, it consults its own self-optimisation and self-evaluation Artificial Neural Networks to begin to learn via self-building another ANN it decides is best for the particular task given to it. In practical terms, you just need to provide the data to learn from and eldr AI will use its own intelligence to work out how best to tackle learning your data so that it can provide you with the best predictions, insights, recommendations and decision intelligence.

Continuous Improvement AI

What is Continuous Improvement AI?

In intelligent animals, including humans, most of us like to think we improve as time goes by - we have to learn a massive amount of stuff - walking, talking, skills, relationships etc.

We therefore should expect nothing less of an AI system. If what we're selling is intelligent then it has to by design continually improve.

Continuous Improvement AI is real life Artificial Intelligence that is able to consistently get better and better by itself. How many so-called AI solutions can do this? Not many.

Why use Continuous Improvement AI?

As someone who has developed my own AI system, to me the most compelling reason to use Continuous Improvement AI is that Continuous Improvement should be an inherent part of any AI system that calls itself intelligent. If it can't display the key attribute of self improvement then it isn't worth bothering with. A static system is not intelligent. A static non-intelligent AI system will give static output meaning commercial growth and decision making could also become stagnant.

On the flip side, an AI system that continually improves is exponentially more valuable and productive in terms of learning capacity, resource use, predictions, insights, recommendations and decisions. You can realistically expect an AI solution that consistently improves itself without any human involvement to be a far better investment than one that stands still.

How does eldr AI continuously improve?

Continuous Improvement is the full automation pinnacle of eldr AI's arsenal of AI capability. We've added a further Deep Learning Artificial Neural Network that combines the self-building, self-optimising, self-evaluating and self-learning ANN components of eldr's intelligence meaning that when put in full automatic mode eldr will literally spend its entire time getting better and better and better - meaning you get smarter and smarter decisions and bigger and bigger growth.


Post a Comment

Post Comment

You are one step closer to getting Artificial Intelligence into your organisation

Fennaio has the expertise to get you up and running with AI, Machine Learning, Deep Learning and Data Science in your new or existing systems, software and operations.

Get Started
no to gdpr