Artificial intelligence is not a new concept but our ability to create these systems have vastly increased in recent years. The automation capability that they enhance is a process that has been taking place for thousands of years. As an AI Artist, I enjoy exploring the potential of what this technology can do now and in the future.
Automation is a noun defined as “The use or introduction of automatic equipment in a manufacturing or other process or facility.”
Let’s take a deep dive into automation and briefly explore the history before concentrating on the present day and what we can expect in the future.
The history of automation dates back well over 5000 years and the evolution of it likely spans the whole of human history – from the Mayans with their aqueducts that automated water transportation to Henry Ford’s automation of the mechanical assembly line, right up to the present day with digital remodelling taking place in most modern organizations, assisted by recent technological advances, it is inaugurating a contemporary age of automation commonly known as Intelligent Automation.
Historically, automation has always presented new ways to create value beyond that which existed before it, by ameliorating the balance within the classical model of people, process, and technology; and it will continue to do so. Many people are concerned about the personal costs of ushering in this new world, as they have always done so throughout history, yet, these automation improvements have in most cases made life more bearable for the majority of people bound to the classical paradigm.
The information age made use of data-related tasks whereby people were required to enable processes such as transactions and interactions with technology like keyboards, spreadsheets, and telephones. In the 1960s, automation of data-driven enterprise resource systems helped automate many data-driven tasks, which eventually evolved to include robotic process automation.
Many data-related tasks still weren’t possible and automation could only be used for very basic operations such as web scraping and the sorting of reliable data. The more complicated tasks were still carried out by people, until recently, with the introduction of intelligent automation.
Intelligent automation is an exciting field that has already increased our ability to automate tasks that were not possible to automate previously and as the technology increases in these areas, it will continue to open up endless possibilities of what can be automated.
Current intelligent automation is being employed to manage and improve business processes automatically and continuously by using improvements in technology such as high-density file systems, combined with recent advancements in algorithmic analysis and artificial intelligence tools. These advanced automation systems strike a balance between human and machine task orientation, allowing humans to engage in higher-value tasks.
The following list is what makes up the constituent parts of automatic intelligence systems:
Artificial intelligence/machine learning – The application of systems equipped with software that simulates human intelligence processes, including learning without explicit instructions
Natural language processing – The ability to understand human speech as it is spoken
Robotics – The use of robots that can act on the Internet of Things (IoT) and other data to learn and make autonomous decisions
Predictive analytics – The practice of predicting outcomes using statistical algorithms and machine learning.
Let’s explore some of these technologies associated with AI that are underpinning intelligent automation and how they can help an organization take advantage of this evolution in the modern business environment.
Recommendation engines make it possible to tailor personalized suggestions to an individual market of one! Predictive analytics makes it possible to anticipate outcomes based on collected knowledge. Deep learning is using artificial neural network algorithms to reason and remember. Natural language processing allows machines to interact and understand using human speech as it is spoken. Machine learning gives these systems the ability to learn and improve without explicit instruction. Image analysis gives us the ability to easily interpret visual images and conduct matching analysis.
A very interesting application example is machines that can write! American artificial intelligence research company OpenAI recently took the art of machine learning to the next stage within the writing sphere. Traditionally, machines previously used various tools to achieve a certain level of comprehension but were found to be lacking in many ways.
The three main categories of machine learning models are statistical models, heuristic models, and models based on biology (such as evolutionary algorithms and neural networks).
Heuristic models are based on guidelines. A simple example of this would be how we learn rules for conjugation of verbs: I jump, you jump, he jumps, etc. This type of model is uncompromising and therefore isn’t an approach that is used much nowadays.
Statistical models were seen as the best approach to machine writing for a very long time. Even as I type this article on Google Docs, a statistical model is working in the background to suggest to me which words will likely be required next. Statistical approaches basically count words to guess what the next word is likely to be based on the statistics. Other, more sophisticated approaches use bigrams (two consecutive word sequences) and trigrams (three consecutive word sequences). This modelling gives some context, allowing the current piece of text to inform the next. These models work very well in simple situations.
OpenAI’s approach takes auto-writing to a whole new level of sophistication, namely, deep learning neural networks. A neural network is a system that is designed to act kind of like a human brain. A neural network is a computer model that has a network architecture. The architecture has a network of nodes that act like neurons in the brain.
The purest form of a neural network has 3 different layers: input layer, hidden layer, and output layer. Each one of these layers contains nodes. There can be many layers depending on the requirements of the network’s task. The input layer picks up the input signals and transfers them to the next layer, gathering the data from the outside world. The hidden layer performs all the back-end tasks of calculation. The output layer transmits the final result of the hidden layer’s calculation.
But why is OpenAI’s neural network so good? To answer that, we need to understand what makes it unique. OpenAI’s neural network is known as GPT-3 (GPT stands for Generative Pre-trained Transformer and the numeral 3 stands for the third iteration of the network model). GPT-3 is massive in comparison to other neural networks. The network has 175 billion parameters and 96 layers, making it ten times bigger than the previous biggest neural network. GPT-3 doesn’t just use trigram sequences, it uses sequences of 2,064 words to achieve its lifelike writing skills.
GPT-3 was recently used by the Guardian newspaper to create a very convincing op-ed written by the neural network that was indistinguishable from a piece written by a human–that’s not the whole story, though. The Guardian newspaper writers gave GTP-3 some suggestions about what to write, and then they edited the finished article as they would with a human writer. The polished article was actually a co-creation between humans and machines. GPT-3 is good, but it isn’t perfect, and this is why humans are still required as a final arbiters.
These neural networks and automation intelligence can help us in many fields other than writing. Neural networks can be used to eliminate waste in manufacturing and streamline supply chains, for instance, and a whole host of other tasks. There are many other innovative advances in various industries which haven’t even been thought of yet.
I create art with Artificially Intelligent Neural Networks. I’m not afraid of the progress happening in this space, in fact, I embrace it! Neural networks need me as much as I need them. Let’s go forward and benefit from this new technology.
In conclusion, automation, machine learning, and artificial intelligence are at the nascent stage of development but are certainly here to stay. We’ll become more dependent on what they have to offer and will see industries completely disrupted, challenged, and changed for the better due to these new technologies. We’ll still need humans just as people were still required on the production lines of car manufacturers to assist the robots, but the humans will be doing less of the heavy lifting tasks and focused on higher-value tasks.