AI (Artificial Intelligence)

There are many questions surrounding Artificial Intelligence, its potential, and what it means for the future: what is Generative AI really capable of? How does it really work? Can it compose a symphony? Can it be aware of itself? How is it going to change the business landscape and how companies handle their processes and interact with customers? What tasks are the most likely to be automated over the next few years? Is it going to take our jobs, and how long will it take?

 

We’ll try to address all these questions and give a general overview of what AI really is and means, as well as what can be expected from it in the future, in this article. But to begin with, we should attempt to give a comprehensive definition: what exactly is Artificial Intelligence?

 

AI (Artificial Intelligence)

Artificial Intelligence or AI has, without a doubt, been the topic of the year in 2023, and everything seems to indicate that it will continue to be in 2024 as well. Artificial Intelligence (AI) refers to the simulation of human intelligence processes by machines, primarily computer systems. In a business context, AI is utilized to automate repetitive tasks, analyze vast amounts of data for insights, and make informed decisions autonomously. It empowers businesses to enhance operational efficiency, personalize customer experiences, predict market trends, and optimize various processes across departments, ultimately driving growth and competitiveness.

 

Over the last year, we’ve seen advancements as unprecedented as rapid in the realm of Generative AI, kick-started by the impressive feats of OpenAI’s ChatGPT and other products like Gemini, Microsoft Bing AI, Grok, or Google Bard aspiring to refine and enhance the capabilities of Conversational AI as well as image generation AI models like Midjourney. 

 

There are many questions surrounding Artificial Intelligence, its potential, and what it means for the future: what is Generative AI really capable of? How does it really work? Can it compose a symphony? Can it be aware of itself? How is it going to change the business landscape and how companies handle their processes and interact with customers? What tasks are the most likely to be automated over the next few years? Is it going to take our jobs, and how long will it take?

 

 

Interface of Athena AI, Connex's Artificial Intelligence especialised model
Athena AI, Connex’s Artificial Intelligence especialised model

 

We’ll try to address all these questions and give a general overview of what AI really is and means, as well as what can be expected from it in the future, in this article. But to begin with, we should attempt to give a comprehensive definition: what exactly is Artificial Intelligence?

 

What is AI?

A good definition of Artificial Intelligence is the one suggested by University of Stafford professor John McCarthy, considered as one of the founders of Artificial Intelligence as an academic discipline: 

 

“Artificial Intelligence is the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable”

What is Artificial Intelligence? – John McCarthy (2004) 

What does this mean? The first bit of the definition is pretty self-explanatory: AI is about making intelligent machines, and particularly intelligent computer programs. There might be different opinions or theories on how to define intelligence; however, that is well beyond our purposes here. 

 

However, in an abstract sense, some skills we could identify as relevant to the concept of intelligence are problem-solving, creativity, critical thinking, planning and decision making, language use, perception, and the ability to learn from experience. These are all human features that researchers in the field of AI aspire to replicate in their models.

 

But for context, it might be useful to clarify the second bit of McCarthy’s definition: AI “is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.” To understand what this means, it will be useful to understand some key aspects of the history of AI as a field. 

 

The earliest characterisation of AI comes from a 1950 paper by Alan Turing, widely considered to be the “father” of Artificial Intelligence. The paper, entitled “Computer Machinery and Intelligence”, attempts to clarify what it means for a machine to be intelligent, and suggests a thought experiment known as the “Turing test”, originally named the “Imitation Game”, as a way to determine whether a machine or piece of software could be considered intelligent or akin to human agents regarding its ability to use language. 

 

To put it shortly, Turing thought that if a machine is able to hold extensive conversations with humans, consistently deceive them into thinking that they are talking to a human rather than a machine, and do so even when humans ask questions especially purported to find out whether they’re talking to a human or a machine, that means that machine should be considered intelligent.

 

In 1950, Turing said about the implications of this game:

 

“I believe that in about fifty years’ time it will be possible to program computers (…) to make them play the imitation game so well that an average interrogator will not have more than a 70% chance of making the right identification after five minutes of questioning. I believe that at the end of the century (…) one will be able to speak of machines thinking without expecting to be contradicted.”

 

Alan Turing, the "father" of AI
Alan Turing, the “father” of AI

 

Clearly, Turing’s paper frames the question in a way that’s relative to a human skill: the ability to use language in a “human-like” way. To Turing, Artificial Intelligence encompassed, first and foremost, technologies aiming to replicate abilities that we considered as intrinsically human. Remember McCarthy’s definition, where AI is “related to the similar task of using computers to understand human intelligence”.

 

In Turing’s time and all throughout the 60s and 70s, Artificial Intelligence was usually connected to the broader discipline of Cognitive Science, a branch of psychology which aims to understand the processes and mechanisms underlying human cognition. During this period, the focus was on developing computer programs that could simulate human intelligence by mimicking cognitive functions like problem-solving, learning, and language processing. Researchers aimed to unravel the intricacies of the human mind and translate these insights into computational models.

 

In the 1980s, however, there was a paradigm shift as AI researchers began exploring more specialised and task-oriented approaches. Expert systems, which focused on capturing the knowledge and reasoning abilities of human experts in specific domains, gained prominence. This departure marked a transition from the broader aspirations of replicating general human intelligence to addressing practical challenges in specific areas. Again, remember McCarthy: “AI does not have to confine itself to methods that are biologically observable”

 

As technological advancements continued, particularly in the fields of machine learning and neural networks, AI evolved further. The late 20th century witnessed the resurgence of interest in developing systems that could learn from data and adapt to new information. This shift led to breakthroughs in pattern recognition, language understanding, and decision-making, paving the way for the sophisticated AI applications we encounter today.

 

Even if we’re still not where the most optimistic predictions from decades ago hoped we would be, Artificial Intelligence has evolved in leaps and bounds over its (roughly) 70 years of history. Today, AI can be categorised in 2 ways: either by its stage of development or by whether it’s weak or strong AI. Let’s explore what this means in the next sections.

 

The 4 Stages of Development of Artificial Intelligence

Arend Hintze, an assistant professor of integrative biology and computer science and engineering at Michigan State University, outlined four categories of AI, starting with the task-specific intelligent systems widely used today and progressing to hypothetical sentient systems. The classifications are as follows:

 

Reactive Machines

A reactive machine adheres to fundamental AI principles and, as the name suggests, can only utilise its intelligence to perceive and respond to the immediate environment. These machines lack the ability to store memories, preventing them from relying on past experiences for real-time decision-making.

 

Direct perception of the world confines reactive machines to a specific set of specialised tasks. Despite this intentional limitation, there are advantages: these AI systems are more dependable and consistent, reacting in a predictable manner to the same input consistently. One example of an AI program falling into this category would be Deep Blue, the IBM chess program that won over Garry Kasparov in the 1990s. While Deep Blue can recognize chessboard pieces and make predictions, its lack of memory prevents it from drawing on past experiences to influence future decisions.

 

Limited Memory AI

AI with limited memory possesses the capability to retain past data and predictions while gathering information and evaluating potential decisions. Essentially, it delves into the past to glean insights into future possibilities, offering more complexity and expansive opportunities compared to reactive machines. Most modern AI models fall within this category.

 

The development of limited memory AI involves continuous training of a model by a team or the creation of an AI environment where models can be automatically trained and updated.

When implementing limited memory AI in machine learning, six essential steps should be followed:

 

  1. Establish training data.
  2. Create the machine learning model.
  3. Ensure the model can make predictions.
  4. Ensure the model can receive human or environmental feedback.
  5. Store human and environmental feedback as data.
  6. Iterate through the above steps as a cyclical process.

 

Theory of Mind AI

Theory of mind refers to the ability to understand that others have beliefs, intentions, and perspectives different from one’s own. It involves recognizing and attributing mental states to others, allowing individuals to comprehend and predict the behaviour of those around them. This cognitive skill is essential for social interactions and empathy.

 

Applied to AI machines, this suggests that they could comprehend the feelings and decision-making processes of humans, animals, and other machines through self-reflection and determination. Subsequently, machines would use this understanding to make decisions autonomously. 

 

In essence, for this to occur, machines must adeptly grasp and process the concept of the “mind,” the emotional dynamics in decision-making, and various other psychological concepts in real-time, establishing a dynamic, two-way relationship between people and AI. To this date, no one has achieved to program an AI model that can be considered to have theory of mind. 

 

Self Aware AI

Achieving theory of mind in AI, possibly in the distant future, precedes the ultimate phase where AI attains self-awareness. In this advanced state, AI would possess human-level consciousness, comprehending its existence, as well as the presence and emotions of others. This self-aware AI would discern others’ needs not solely from explicit communication but also from the nuances of how it is conveyed. 

 

Building self-awareness into AI necessitates a dual understanding: researchers comprehending the fundamentals of consciousness and then replicating it effectively in machines. Just like Theory of Mind AI, AI with true self-awareness does not currently exist.

 

Weak AI vs Strong AI: What’s the difference?

While, as we have mentioned earlier, it might be difficult to give a general, comprehensive account of what intelligence is, the fundamental difference between strong AI (Artificial General Intelligence) and weak AI (Narrow or Specialized AI) lies in their scope of intelligence and task capabilities.

 

Strong AI, often referred to as artificial general intelligence, envisions a machine with human-like cognitive abilities. It can understand, learn, and apply intelligence to any task, much like a human being. It possesses adaptability, handling unfamiliar problems without specific programming or training for each task. However, as of now, strong AI remains a theoretical concept, and we do not have machines with true human-level general intelligence.

 

On the other hand, weak AI, also known as narrow AI, is designed to excel in specific tasks or a limited set of tasks. It simulates human intelligence within a predefined context and lacks the broad cognitive abilities of a human. Examples of weak AI include virtual assistants like Siri and Alexa, recommendation systems, speech recognition, and self-driving cars. Unlike strong AI, weak AI is prevalent and widely used in various practical applications.

 

Deep Learning vs Machine Learning

While deep learning and machine learning are often used interchangeably, it’s essential to recognize the subtle distinctions between the two. Both fall under the umbrella of artificial intelligence, with deep learning operating as a subset of machine learning, specifically involving neural networks.

 

The term “deep” in deep learning signifies a neural network with more than three layers, encompassing inputs and outputs. This distinction is visually depicted in the accompanying diagram.

 

The primary divergence lies in how these algorithms learn. Deep learning streamlines the feature extraction process, reducing the need for manual human intervention and facilitating the use of more extensive datasets. It can be conceptualised as “scalable machine learning,” a term echoed by Lex Fridman in the same MIT lecture mentioned earlier. In contrast, traditional machine learning relies more on human expertise to determine the feature hierarchy, typically requiring structured data for effective learning.

 

“Deep” machine learning can make use of labelled datasets through supervised learning but doesn’t mandate a labelled dataset. It can process unstructured raw data (e.g., text, images) autonomously, determining feature hierarchies to distinguish between various data categories. Unlike machine learning, it operates without human intervention in data processing, offering opportunities for more innovative scaling of machine learning processes.

 

What is Generative AI?

Generative AI is a term used to encompass deep-learning models capable of processing raw data, ranging from extensive sources like Wikipedia to the complete works of artists such as Rembrandt. These models undergo a learning process to generate outputs with statistically probable similarities to the original data, albeit not identical. In summary, generative models simplify the representation of their training data, drawing from it to create new works.

 

While generative models have long been employed in statistical analysis of numerical data, the rise of deep learning has enabled their extension to more complex data types like images and speech. 

 

"Theatre D'Opera Spatial", an image created using the generative artificial intelligence platform Midjourney by Jason Michael Allen. The image won the 2022 Colorado State Fair's annual fine art competition.
“Theatre D’Opera Spatial”, an image created using the generative artificial intelligence platform Midjourney by Jason Michael Allen. The image won the 2022 Colorado State Fair’s annual fine art competition.

 

Early instances of models, like GPT-3, BERT, or DALL-E 2, have showcased the potential of what can be achieved. The future envisions models trained on extensive sets of unlabeled data that prove versatile across various tasks, requiring minimal fine-tuning. The transition is evident from task-specific systems in a single domain to broad AI that learns comprehensively, spanning multiple domains and addressing diverse problems. Foundation models, trained on large, unlabeled datasets and fine-tuned for various applications, spearhead this transformative shift.

 

In the realm of generative AI, the anticipation is that foundation models will significantly hasten AI adoption in enterprises. The reduction in labelling requirements will facilitate easier integration for businesses, and the precision and efficiency of AI-driven automation will enable a broader range of companies to deploy AI in critical situations.

 

Hopefully, by now you have a clear understanding of what Artificial Intelligence is, what it means, how it works, and how different types of AI can be categorised. Now, let’s move to more practical matters: what can you use AI for?

 

In the following section, we’ll discuss several applications for Artificial Intelligence, as well as different industries and fields that can benefit from their use.

 

What is AI used for?

 

Examples of AI Applications

Automation

Artificial Intelligence (AI) plays a pivotal role in revolutionising automation across various industries. Through the integration of sophisticated algorithms and machine learning capabilities, AI enhances automation tools, enabling them to perform complex and diverse tasks with unprecedented efficiency. 

 

One prominent example is Robotic Process Automation (RPA), a form of software designed to automate rule-based, repetitive data processing tasks traditionally carried out by humans. With the infusion of AI, automation systems can go beyond routine operations, adapting to dynamic environments, learning from data patterns, and making intelligent decisions. This symbiotic relationship between AI and automation not only streamlines and accelerates processes but also opens up new possibilities for innovation, transforming the way businesses operate and increasing overall productivity.

 

Another example is the automation of customer interactions with the use of AI Chatbots. Intelligent Conversational AI Bots, like Athena AI, can be trained and fine-tuned to respond to customer queries with unmatched levels of precision.

 

Machine Vision

Machine Vision involves the capture and analysis of visual information through a camera, analog-to-digital conversion, and digital signal processing. While often likened to human eyesight, machine vision surpasses biological constraints and can be programmed for tasks such as seeing through walls. 

 

Its applications are diverse, spanning from signature identification to medical image analysis. It’s important to note that computer vision, concentrated on machine-based image processing, is frequently interchanged with the term machine vision.

 

NLP

Natural Language Processing (NLP) involves the computational analysis of human language by computer programs. Modern NLP approaches heavily rely on machine learning techniques, evolving beyond rule-based systems. NLP tasks extend to intricate processes such as text translation, deciphering sentiment from written content with Sentiment Analysis, and even speech recognition. 

 

AI Chatbots like Athena AI leverage NLP and Machine Learning to offer detailed responses to specific queries
AI Conversational Bots like Athena AI leverage NLP and Machine Learning to offer detailed responses to specific queries

 

Conversational AI, a facet of NLP, has gained especial prominence over the last couple of years, enabling machines to understand and respond to human language in dynamic interactions. The realm of AI Chatbots is a notable application of conversational AI; intelligent chatbots use NLP to engage in natural, text-based conversations with users, offering a seamless and interactive experience across various platforms.

 

Robotics

Robotics engineering is focused on the creation and production of robots, machines designed for tasks challenging or consistently demanding for humans. Common applications include robotic involvement in car assembly lines, where precision and consistency are paramount, and in space exploration by entities like NASA, utilising robots to manoeuvre substantial objects. 

 

Beyond mechanical design, researchers employ machine learning to develop socially interactive robots, showcasing the evolving intersection of robotics and artificial intelligence in enhancing machines’ capabilities for diverse applications.

 

Text, Image and Audio Generation

The widespread adoption of generative AI techniques marks a significant trend in various industries, as these methods harness the capability to generate diverse forms of media based on text prompts. This application extends across businesses, allowing the creation of an extensive array of content types. 

 

From producing photorealistic art to crafting email responses and even generating screenplays, generative AI showcases its versatility in generating creative and functional content. The technology’s capacity to transform textual input into varied media formats offers businesses a tool with seemingly limitless possibilities for content creation and innovation.

 

 

AI for Businesses: What industries use Artificial Intelligence?

Business Communications and Customer Service

The infusion of machine learning algorithms into analytics and customer relationship management (CRM) platforms is experiencing a substantial uptick, with a primary focus on extracting valuable insights to optimise customer service, as well as enabling customer service automation. Concurrently, websites are increasingly embracing the integration of chatbots, streamlining customer interactions and elevating user experiences. This wave of technological progress is further catalysing a significant emphasis on workflow automation within customer service.

 

AI-driven workflow automation is reshaping the operational landscape, offering a streamlined approach to handling routine tasks and enhancing overall efficiency. Automation is becoming a cornerstone in customer service processes, allowing for the swift resolution of queries, automated data entry, and seamless integration with CRM systems. This not only accelerates response times but also contributes to a more personalised and efficient service delivery.

 

Athena AI routing an inbound interaction based on Live Chat responses
Athena AI routing an inbound interaction based on Live Chat responses

 

Moreover, the continuous evolution of generative AI technologies, such as the advancements witnessed in ChatGPT, is set to unleash transformative consequences. This trajectory encompasses a spectrum of changes, ranging from the potential restructuring of job roles to a revolutionary overhaul of product design methodologies and the disruptive reconfiguration of established business models. 

 

In this dynamic landscape, the integration of Artificial Intelligence, including advanced capabilities like Automatic Speech Recognition (ASR) Interactive Voice Response (IVR) systems, is steering industries towards a future where workflow automation becomes increasingly integral to customer service excellence.

 

Education

Artificial Intelligence (AI) has the potential to revolutionise education by automating grading processes, freeing up educators to focus on other essential tasks. It goes beyond mere assessment, adapting to individual student needs and allowing them to progress at their own pace. AI tutors offer additional support, ensuring students remain on course. 

 

This transformative technology might even reshape the traditional classroom dynamic, with the possibility of AI playing a more central role, potentially impacting the conventional teaching model. Notably, tools like ChatGPT, Google Bard, and other advanced language models demonstrate how generative AI can assist educators in crafting engaging course materials and interact with students in innovative ways. 

 

Healthcare

The primary focus in healthcare innovation revolves around enhancing patient outcomes and cost reduction. Companies are leveraging machine learning to surpass human capabilities in making more accurate and rapid medical diagnoses.

 

Beyond this, AI applications extend to virtual health assistants and chatbots, aiding patients in accessing medical information, scheduling appointments, comprehending billing processes, and handling administrative tasks. Moreover, a diverse range of AI technologies plays a crucial role in predicting, combating, and comprehending pandemics, exemplified by their application in addressing challenges posed by diseases such as COVID-19.

 

Finance

The presence of AI in personal finance tools like Intuit Mint or TurboTax is causing a transformation in the financial sector. These applications, by gathering personal data, offer financial guidance to users. Additionally, in activities like home purchasing, programs like IBM Watson are employed. 

 

Notably, artificial intelligence software has taken a prominent role in executing a significant portion of trading activities on Wall Street today. This shift underscores the substantial impact AI is making in disrupting traditional financial institutions and reshaping various aspects of financial processes.

 

Law

In the legal realm, the exploration phase, involving the examination of documents, can be immensely challenging for humans in the legal field. The incorporation of AI to automate labour-intensive tasks in the legal industry proves to be a time-saving strategy, ultimately enhancing client service.

 

Law firms leverage machine learning for data description and outcome prediction, utilise computer vision to categorise and extract information from documents, and employ Natural Language Processing (NLP) to decipher requests for information. This integration of AI technologies in the legal domain not only streamlines processes but also demonstrates a commitment to improving efficiency and client support within the legal profession.

 

Media

AI techniques play a significant role in the entertainment industry, impacting targeted advertising, content recommendation, distribution, fraud detection, script creation, and film production. The integration of Artificial Intelligence extends to journalism, where automated processes streamline media workflows, reducing time, costs, and complexity. Newsrooms leverage AI for routine tasks, including data entry and proofreading, as well as for researching topics and generating headlines. However, the reliable use of generative AI tools like ChatGPT for content creation in journalism raises questions about its feasibility and ethical considerations.

 

Software coding and IT

Emerging generative AI tools have the capability to generate application code guided by natural language prompts. However, it is still in the early stages for these tools, and it’s improbable that they will replace software engineers in the near future. 

 

Additionally, AI is actively applied to automate various IT processes, encompassing tasks such as data entry, fraud detection, customer service, predictive maintenance, and security measures.

 

Cybersecurity

AI techniques are proving effective across various facets of cybersecurity, addressing tasks such as anomaly detection, mitigating the false-positive issue, and implementing behavioural threat analytics. 

 

Within organisations, machine learning is integrated into Security Information and Event Management (SIEM) software and related domains to detect anomalies and discern suspicious activities indicative of potential threats. Through the analysis of data and logical comparisons to known malicious code, AI can promptly alert to emerging attacks, outpacing the capabilities of human employees and preceding technology iterations.

 

Logistics and transportation

Beyond its essential function in operating autonomous vehicles, Artificial Intelligence plays a crucial role in transportation by overseeing traffic management, forecasting flight delays, and enhancing safety and efficiency in ocean shipping. 

 

Moreover, within supply chains, AI is supplanting conventional approaches to demand forecasting and disruption prediction. The accelerated adoption of AI in supply chain management, particularly spurred by the unforeseen impacts of the global COVID-19 pandemic, underscores the technology’s ability to address dynamic challenges and optimise operations in the realm of transportation and logistics.

 

AI Agent: the next step

Interacting with Artificial Intelligence has traditionally involved inputting prompts for models to generate responses. However, the landscape is shifting with the rise of AI agents, which operate autonomously, driven by objectives rather than prompts. These agents independently devise task lists and adapt based on feedback, continuously evolving to optimize goal achievement.

 

Unlike conventional automation, which relies on predetermined triggers, AI agents excel in navigating unpredictable environments, representing a dynamic form of intelligent automation. They perceive surroundings, process information, and execute actions, ranging from straightforward systems to sophisticated entities capable of learning and adjusting.

 

At the core of an AI agent lies its function, translating data into actions. Percepts convey sensory inputs, while actuators execute decisions. The knowledge base provides initial knowledge, and feedback drives continual improvement.

 

Various types of AI agents exist, from simple reflex to belief-desire-intention agents, finding applications in diverse domains such as autonomous vehicles, virtual assistants, healthcare, finance, customer service, robotics, cybersecurity, and education.

 

These agents revolutionize industries by enhancing decision-making processes, streamlining operations, and providing personalized experiences. However, challenges such as technical optimization and ethical regulation must be addressed.

 

The future of AI agents holds immense potential to augment human capabilities, driving increased productivity and innovation. Businesses must invest in technology and training to fully leverage these benefits, fostering collaboration between AI systems and human workers to unlock their full potential.

 

As AI agents continue to evolve, they represent a significant opportunity to revolutionize industries while affirming the importance of human contribution in the workforce.

 

 

If you enjoyed reading this, you might also be interested in…

NLP (Natural Language Processing)

Chatbot

AI Chatbots

Call Centre AI

AI Automation

Customer Service AI