What is Artificial intelligence (AI)?

Artificial intelligence or AI: Artificial intelligence (AI) is a set of technologies that enables computers and machines to be able to perform a variety of advanced functionalities that would normally require human intelligence (such as learning, reasoning, perceiving, and problem-solving capabilities with quality) or that involves data whose scale exceeds what humans can analyze, including the ability to analyze data, translate spoken and written language, see and understand, then make recommendations, and many more. These activities might be simple, repetitive or complex, cognitive in nature.

Understanding AI in 2024

Artificial intelligence (AI) is the backbone of innovation in modern computing, unlocking value for individuals and businesses. Artificial intelligence (AI) or combined with other technologies (e.g., sensors, geolocation, robotics) can perform tasks that would otherwise require human intelligence or intervention. Digital assistants, GPS guidance, autonomous vehicles, and generative AI tools (like Open AI’s Chat GPT, Google’s Gemini, among others) are just a few examples of AI in the daily news and our daily lives

As a field of computer science, artificial intelligence encompasses (and is often mentioned together with) machine learning and deep learning. These disciplines involve the decision-making process modeled after the development of AI algorithms that can ‘learn’ from available data and make increasingly more accurate classifications or predictions over time like that of the human brain.

The release of ChatGPT seems to mark a turning point even though Artificial intelligence has gone through many cycles of hype. Today, generative AI can learn and synthesize other data types including images, video, software code, and even molecular structures and not just human languages.

Applications for AI are growing day by day. Ethics and responsible AI has become critically important as the hype around the use of AI tools in business takes off.

The Four Types of Artificial intelligence (AI)

As researchers strive to develop more advanced forms of artificial intelligence, they also aim to refine their understanding of intelligence and consciousness. To clarify these concepts, researchers, including Professor Arend Hintze of the University of Michigan, have identified four types of AI:

1. Reactive Machines

• Description: The most basic form of AI, reactive machines do not possess knowledge of past events. They can only “react” to present situations.

• Capabilities: These machines perform specific tasks within a narrow scope, such as playing chess, but cannot operate outside their limited context.

2. Limited Memory Machines

• Description: These machines have a limited understanding of past events, allowing them to interact more dynamically with their environment.

• Capabilities: An example is self-driving cars, which use limited memory to make decisions about turns, speed adjustments, and observing other vehicles. However, their understanding of the world is still constrained to short-term memory.

3. Theory of Mind Machines

• Description: Representing an early form of artificial general intelligence (AGI), these machines can create representations of the world and understand other entities within it.

• Capabilities: Although this type of AI has not yet been realized, it would require the ability to recognize and interpret the emotions, beliefs, and intentions of others.

4. Self-Aware Machines

• Description: The most advanced theoretical form of AI, these machines would possess self-awareness, understanding themselves, others, and the world.

• Capabilities: Achieving this level of AI would mean reaching true AGI. However, this remains a distant reality.

Types of Artificial Intelligence: Weak AI vs Strong AI

1. Weak AI —

• It is known as “Narrow AI” or Artificial Narrow Intelligence (ANI)—is AI trained and focused to perform specific tasks, such as remembering things, perceiving things, or solving simple problems. Weak AI models intelligent human behavior, which allows machines to solve complex problems.

• Examples of weak AI include image and facial recognition systems, fraud detection software, and predictive maintenance models. Weak AI drives most of the AI that surrounds us today.

• “Narrow” might be a more apt descriptor for this type of AI as it is anything but weak: it enables some very robust applications, such as Apple’s Siri, Amazon’s Alexa, IBM watsonx™, and self-driving vehicles.

2. Strong AI —

• It is made up of “Artificial General Intelligence” (AGI) and “Artificial Super Intelligence” (ASI). AGI, or general AI, is a theoretical form of AI where a machine would have an intelligence equal to humans; it would be self-aware with a consciousness that would have the ability to solve problems, learn, and plan for the future.

• While there are no clear examples of strong artificial intelligence, the field of AI is rapidly innovating. ASI—also known as “superintelligence” — would surpass the intelligence and ability of the human brain. While strong AI is still entirely theoretical with no practical examples in use today, that doesn’t mean AI researchers aren’t also exploring its development. Strong AI would be able to learn from experience and apply that learning to other situations, as well as use life experiences to plan for the future. Strong AI systems don’t actually exist yet, but deep learning algorithms are most strongly associated with human-level AI.

• In the meantime, the best examples of ASI might be from science fiction, such as HAL, the superhuman and rogue computer assistant in 2001: A Space Odyssey.

Deep learning vs Machine learning

“Machine learning and Deep learning” are sub-disciplines of AI, and deep learning is a sub-discipline of machine learning.

Both machine learning and deep learning algorithms use neural networks to ‘learn’ from huge amounts of data. These neural networks are programmatic structures modeled after the decision-making processes of the human brain. They consist of layers of interconnected nodes that extract features from the data and make predictions about what the data represents.

Machine learning and deep learning differ in the types of neural networks they use, and the amount of human intervention involved. Classic machine learning algorithms use neural networks with an input layer, one or two ‘hidden’ layers, and an output layer. Typically, these algorithms are limited to supervised learning: the data needs to be structured or labeled by human experts to enable the algorithm to extract features from the data.

Deep learning algorithms use deep neural networks—networks composed of an input layer, three or more (but usually hundreds) of hidden layers, and an output layout. These multiple layers enable unsupervised learning: they automate extraction of features from large, unlabeled and unstructured data sets. Because it doesn’t require human intervention, deep learning essentially enables machine learning at scale.

What is Artificial General Intelligence (AGI)?

Artificial general intelligence (AGI) refers to a theoretical state in which computer systems will be able to achieve or exceed human intelligence. In other words, AGI is “true” artificial intelligence as depicted in countless science fiction novels, television shows, movies, and comics.

As for the precise meaning of “AI” itself, researchers don’t quite agree on how we would recognize “true” artificial general intelligence when it appears. However, the most famous approach to identifying whether a machine is intelligent or not is known as the Turing Test or Imitation Game, an experiment that was first outlined by influential mathematician, computer scientist, and cryptanalyst Alan Turing in a 1950 paper on computer intelligence. There, Turing described a three-player game in which a human “interrogator” is asked to communicate via text with another human and a machine and judge who composed each response. If the interrogator cannot reliably identify the human, then Turing says the machine can be said to be intelligent.

For example, while a recent paper from Microsoft Research and OpenAI argues that Chat GPT-4 is an early form of AGI, many other researchers are skeptical of these claims and argue that they were just made for publicity.

The rise of Generative AI Models

Generative AI refers to artificial intelligence systems that are capable enough to create new content, such as text, images, music, or other data, often mimicking human creativity. These systems leverage machine learning, particularly deep learning, to generate outputs that are novel and relevant to the input data they were trained on. These systems leverage machine learning, particularly deep learning, to generate outputs that are novel and relevant to the input data they were trained on.

At a high level, generative models encode a simplified representation of their training data and draw from it to create a new work that’s similar, but not identical, to the original data.

Generative models have been used for years in statistics to analyze numerical data. The rise of deep learning, however, made it possible to extend them to images, speech, and other complex data types. Among the first class of AI models to achieve this cross-over feat were variational autoencoders, or VAEs, introduced in 2013. VAEs were the first deep-learning models to be widely used for generating realistic images and speech.

Early examples of models, including GPT-3, BERT, or DALL-E 2, have shown what’s possible. Systems that execute specific tasks in a single domain are giving way to broad AI systems that learn more generally and work across domains and problems. Foundation models, trained on large, unlabeled datasets and fine-tuned for an array of applications, are driving this shift.

As to the future of AI, when it comes to generative AI, it is predicted that foundation models will dramatically accelerate AI adoption in enterprise. Reducing labeling requirements will make it much easier for businesses to dive in, and the highly accurate, efficient AI-driven automation they enable will mean that far more companies will be able to deploy AI in a wider range of mission-critical situations.

Types of Generative Models:

1. Generative Adversarial Networks (GANs):

• Consist of two neural networks, a generator and a discriminator, that compete with each other to create realistic data.

2. Variational Autoencoders (VAEs):

• Encode input data into a latent space and then decode it back to generate new data.

3. Autoregressive Models:

• Generate data one step at a time, such as GPT-3 for text.

Applications:

• Text Generation: AI models like GPT-3 can write essays, articles, or even poetry.

• Image Creation: GANs can generate realistic images of faces, objects, or scenes.

• Music Composition: AI can create original music compositions.

• Game Development: Generative AI can create game levels, characters, and narratives.

• Healthcare: Synthesizing medical data for research and training purposes.

Challenges:

• Quality Control: Ensuring generated content is high-quality and free from errors.

• Bias and Ethics: Addressing biases in training data to avoid harmful or biased outputs.

• Intellectual Property: Determining ownership and copyright of AI-generated content.

Future Directions:

• Improved Models: Enhancing the capabilities and efficiency of generative models.

• Interdisciplinary Applications: Expanding generative AI applications across various fields.

• Ethical Guidelines: Developing frameworks to guide ethical use of generative AI.

Generative AI is rapidly evolving and has the potential to transform numerous industries by automating creative processes and generating novel content that can be utilized in diverse ways.

Artificial Intelligence Training Models

Artificial intelligence (AI) training models involve a process where AI systems learn from data to perform specific tasks. This process is known as “training” and it is crucial for developing AI models that can recognize patterns, make decisions, and generate content.

Here’s a detailed look at the key aspects of training AI models:

1. Data Collection

1. Data Types: Text, images, audio, video, and other sensor data.
2. Data Sources: Databases, online repositories, user-generated content, and proprietary data.

2. Data Preprocessing

1. Cleaning: Removing noise and correcting errors in the data.
2. Normalization: Scaling data to a standard range.
3. Augmentation: Generating new data samples by altering existing data (e.g., rotating images).
4. Splitting: Dividing data into training, validation, and test sets.

3. Choosing a Model

a) Supervised Learning:

• Models trained with labeled data (e.g., classification, regression).

• A machine learning model known as supervised learning uses labelled training data, or structured data, to translate a given input into an output. Put simply, you input the algorithm images that have been labelled as cats in order to train it to recognise pictures of cats.

b) Unsupervised Learning:

• Models trained with unlabeled data (e.g., clustering, dimensionality reduction).

• A machine learning technique known as “unsupervised learning” makes patterns out of unlabeled (or “unstructured”) data. In contrast to supervised learning, the outcome is unpredictable. Instead, the algorithm classifies the input into groups according to qualities as it gains knowledge from it. Unsupervised learning, for example, excels in descriptive modelling and pattern matching.

c) Semi-supervised Learning:

• Combines labeled and unlabeled data.

• Semi-supervised learning is a type of machine learning that combines aspects of supervised and unsupervised learning.

d) Reinforcement Learning:

• Models learn by interacting with an environment to maximize rewards.

• A machine learning model known as “learn by doing” is called reinforcement learning. Through trial and error (a feedback loop), a “agent” learns how to carry out a specified task until its performance falls within a desired range. When the agent completes the task successfully, it gets positive reward; when it doesn’t, it gets negative reinforcement. Teaching a robotic hand to pick up a ball is an example of reinforcement learning.

4. Model Architecture

1. 1. Linear Models: Simple models like linear regression.

1. 1. Neural Networks: Multi-layered networks for complex tasks (e.g., CNNs for images, RNNs for sequential data).

1. 1. Ensemble Methods: Combining multiple models to improve performance (e.g., random forests, gradient boosting).

5. Tools and Frameworks

1. 1. Libraries: TensorFlow, PyTorch, scikit-learn, Keras.

1. 1. Platforms: Google Cloud AI, AWS AI, Microsoft Azure AI.

Training AI models is a complex, iterative process that requires careful consideration of data, model selection, training techniques, and evaluation metrics to develop effective and reliable AI systems.

Common types of Artificial Neural Networks

Artificial neural networks (ANNs) are a cornerstone of deep learning, designed to mimic the way human brains process information. There are various types of neural networks, each suited to different types of tasks. Here are some of the most common types:

1. Feedforward Neural Networks (FNN)

• Structure: The simplest type of ANN where connections between nodes do not form cycles. It consists of an input layer, one or more hidden layers, and an output layer.

• Use Cases: Basic tasks like image recognition and simple classification problems.

2. Convolutional Neural Networks (CNN)

• Structure: Contains convolutional layers, pooling layers, and fully connected layers. Convolutional layers apply filters to the input to capture spatial hierarchies in data.

• Use Cases: Image and video recognition, image classification, object detection, and visual data processing.

3. Recurrent Neural Networks (RNN)

• Structure: Includes loops in the network allowing information to be carried across nodes, making them ideal for sequential data. They have memory cells that capture information about previous inputs.

• Use Cases: Time series analysis, language modeling, speech recognition, and text generation.

4. Long Short-Term Memory Networks (LSTM)

• Structure: A type of RNN designed to handle long-term dependencies. LSTMs use gates to control the flow of information, making them effective at remembering information over long sequences.

• Use Cases: Tasks requiring memory over long sequences, like language translation, time series prediction, and handwriting recognition.

5. Gated Recurrent Units (GRU)

• Structure: A simplified version of LSTM networks, with fewer gates and thus fewer parameters. They are computationally more efficient while still addressing the vanishing gradient problem.

• Use Cases: Similar to LSTMs, including sequence prediction and natural language processing.

6. Autoencoders

• Structure: Comprises an encoder to compress the input into a latent space and a decoder to reconstruct the input from the latent space. It can be used for unsupervised learning.

• Use Cases: Dimensionality reduction, image denoising, and anomaly detection.

7. Generative Adversarial Networks (GANs)

• Structure: Consists of two neural networks, a generator and a discriminator, that are trained simultaneously. The generator creates data, and the discriminator evaluates its authenticity.

• Use Cases: Image generation, video generation, and creating realistic simulations.

8. Radial Basis Function Networks (RBFN)

• Structure: Includes an input layer, a hidden layer with radial basis functions as activation functions, and an output layer. They are used for function approximation.

• Use Cases: Function approximation, time-series prediction, and control systems.

9. Self-Organizing Maps (SOM)

• Structure: An unsupervised learning algorithm that uses a grid of neurons to map high-dimensional data into a lower-dimensional space.

• Use Cases: Data visualization, cluster analysis, and feature extraction.

10. Transformer Networks

• Structure: Uses mechanisms called attention to process input data in parallel rather than sequentially. It consists of encoder and decoder blocks and is highly effective for sequential data.

• Use Cases: Natural language processing tasks such as translation, text summarization, and question answering.

Each type of neural network has its unique structure and applications, making them suitable for different kinds of tasks in AI and machine learning.

Benefits of Artificial intelligence (AI)

Artificial Intelligence (AI) offers numerous benefits across various sectors. Here are some key advantages:

1. Automation

• Increased Efficiency: AI systems can automate repetitive and mundane tasks, increasing efficiency and freeing up human workers to focus on more complex activities.

• Cost Reduction: Automating tasks can lead to significant cost savings in labor and operational expenses.

2. Enhanced Decision Making

• Data Analysis: AI can analyze large datasets quickly and accurately, providing insights that help in making informed decisions.

• Predictive Analytics: AI algorithms can predict future trends based on historical data, aiding in strategic planning.

3. Improved Customer Experience

• Personalization: AI can provide personalized recommendations and services, enhancing customer satisfaction.

• 24/7 Support: AI-powered chatbots and virtual assistants can offer round-the-clock customer support.

4. Increased Productivity

• Task Management: AI can optimize workflows and manage tasks efficiently, leading to higher productivity levels.

• Resource Allocation: AI can allocate resources more effectively, ensuring optimal use of available assets.

5. Innovation and Creativity

• New Products and Services: AI can aid in the creation of innovative products and services that were previously unimaginable.

• Creative Assistance: AI tools can assist in creative processes such as content creation, music composition, and design.

6. Healthcare Advancements

• Medical Diagnosis: AI can analyze medical images and data to assist in diagnosing diseases accurately.

• Personalized Treatment: AI can tailor treatment plans based on individual patient data, improving health outcomes.

7. Safety and Security

• Fraud Detection: AI can detect fraudulent activities by analyzing transaction patterns.

• Surveillance: AI-enhanced surveillance systems can identify suspicious behavior and potential threats.

8. Economic Growth

• New Markets: AI can create new markets and industries, driving economic growth.

• Job Creation: While AI can automate certain jobs, it also creates new job opportunities in developing and managing AI systems.

9. Environmental Impact

• Resource Management: AI can optimize the use of natural resources, leading to more sustainable practices.

• Climate Modeling: AI can improve climate models, helping predict and mitigate the effects of climate change.

10. Education and Learning

• Personalized Learning: AI can create customized learning experiences tailored to individual student needs.

• Tutoring and Support: AI-powered educational tools can provide additional support and tutoring to students.

Danger of Artificial intelligence (AI)

While artificial intelligence offers numerous benefits, it also presents several potential risks and dangers. Here are some of the main concerns associated with AI:

1. Job Displacement

• Automation: AI systems can perform tasks traditionally done by humans, leading to job losses in various sectors, particularly in routine and repetitive jobs.

• Economic Disruption: The rapid pace of automation could outstrip the ability of economies and workers to adapt, leading to significant economic and social upheaval.

Bias and Discrimination

• Training Data: AI systems learn from data, and if this data contains biases, the AI can perpetuate and even exacerbate these biases, leading to discriminatory outcomes in areas like hiring, lending, and law enforcement.

• Algorithmic Bias: Inherent biases in the algorithms themselves can also lead to unfair and unethical decisions.

2. Privacy Concerns

• Data Collection: AI systems often require vast amounts of data, raising concerns about how this data is collected, stored, and used.

• Surveillance: AI can enhance surveillance capabilities, leading to potential misuse by governments and organizations to monitor and control individuals.

3. Security Risks

• Hacking and Cyberattacks: AI can be used to develop sophisticated cyberattacks, and AI systems themselves can be vulnerable to hacking.

• Autonomous Weapons: The development of AI-powered weapons could lead to new forms of warfare, with significant ethical and safety implications.

4. Lack of Transparency

• Black Box Problem: Many AI models, particularly deep learning models, are complex and not easily interpretable, making it difficult to understand how they make decisions.

• Accountability: Lack of transparency can lead to challenges in holding AI systems accountable for their actions and decisions.

Ethical Concerns

• Decision-Making: AI systems making critical decisions (e.g., in healthcare, law enforcement) raise ethical questions about responsibility and the appropriateness of machine-led decision-making.

• Autonomy: As AI systems become more autonomous, ensuring they act ethically and align with human values becomes more challenging.

5. Existential Risks

• Super intelligent AI: The theoretical development of AI that surpasses human intelligence poses existential risks if such AI acts in ways that are detrimental to humanity.

• Control Problem: Ensuring that super intelligent AI remains under human control and aligned with human interests is a significant concern.

Addressing these dangers requires a multi-faceted approach, including robust regulatory frameworks, ethical guidelines, ongoing research into AI safety, and public dialogue about the societal impacts of AI.

Artificial Intelligence Applications

There are numerous, real-world applications for AI systems today. Below are some of the most common use cases:

1. Speech recognition

Also known as automatic speech recognition (ASR), computer speech recognition, or speech-to-text, speech recognition uses NLP to process human speech into a written format.

2. Image recognition

Identify and categorize various aspects of an image.

3. Customer service

Online virtual agents and chatbots are replacing human agents along the customer journey. They answer frequently asked questions (FAQ) around topics, like shipping, or provide personalized advice, cross-selling products or suggesting sizes for users, changing the way we think about customer engagement across websites and social media platforms.

Examples include messaging bots on e-commerce sites with virtual agents, messaging apps, such as Slack and Facebook Messenger, and tasks usually done by virtual assistants and voice assistants.

4. Computer vision

This AI technology enables computers and systems to derive meaningful information from digital images, videos and other visual inputs, and based on those inputs, it can take action. This ability to provide recommendations distinguishes it from image recognition tasks. Powered by convolutional neural networks, computer vision has applications within photo tagging in social media, radiology imaging in healthcare, and self-driving cars within the automotive industry.

5. Supply chain

Adaptive robotics act on Internet of Things (IoT) device information, and structured and unstructured data to make autonomous decisions. NLP tools can understand human speech and react to what they are being told. Predictive analytics are applied to demand responsiveness, inventory and network optimization, preventative maintenance and digital manufacturing. Search and pattern recognition algorithms—which are no longer just predictive, but hierarchical—analyze real-time data, helping supply chains to react to machine-generated, augmented intelligence, while providing instant visibility and transparency.

6. Weather forecasting

The weather models broadcasters rely on to make accurate forecasts consist of complex algorithms run on supercomputers. Machine-learning techniques enhance these models by making them more applicable and precise.

7. Anomaly detection

AI models can comb through large amounts of data and discover atypical data points within a dataset. These anomalies can raise awareness around faulty equipment, human error, or breaches in security.

8. Cybersecurity

Autonomously scan networks for cyber-attacks and threats.

9. Predictive modeling

Mine data to forecast specific outcomes with high degrees of granularity.

10. Translation

Translate written or spoken words from one language into another.

History of Artificial Intelligence: Key dates and names

The concept of “a machine that thinks” dates back to ancient Greece. However, significant milestones in the evolution of artificial intelligence, particularly since the advent of electronic computing, include the following:

Key Dates and Names in Artificial intelligence AI History

1. 1950: Alan Turing publishes Computing Machinery and Intelligence. Turing, known for breaking the German ENIGMA code during WWII and often called the “father of computer science,” poses the question: “Can machines think?” He introduces the Turing Test, where a human interrogator tries to distinguish between a computer and human text response. Despite much scrutiny, the Turing Test remains a foundational concept in AI and philosophy, exploring ideas around linguistics.

2. 1956: John McCarthy coins the term “artificial intelligence” at the first AI conference at Dartmouth College. McCarthy later invents the Lisp programming language. In the same year, Allen Newell, J.C. Shaw, and Herbert Simon develop the Logic Theorist, the first AI software program.

3. 1967: Frank Rosenblatt creates the Mark 1 Perceptron, the first computer based on a neural network that learns through trial and error. In 1968, Marvin Minsky and Seymour Papert publish Perceptrons, a landmark work on neural networks that also casts doubt on future neural network research.

4. 1980s: Neural networks utilizing backpropagation algorithms become widely used in AI applications.

5. 1995: Stuart Russell and Peter Norvig publish Artificial Intelligence: A Modern Approach, a leading AI textbook. They discuss four potential goals or definitions of AI, differentiating systems based on rationality and thinking vs. acting.

6. 1997: IBM’s Deep Blue defeats world chess champion Garry Kasparov in a chess match.

7. 2011: IBM Watson defeats Jeopardy! champions Ken Jennings and Brad Rutter.

8. 2015: Baidu’s Minwa supercomputer employs a convolutional neural network to identify and categorize images with greater accuracy than humans.

9. 2016: DeepMind’s AlphaGo, powered by a deep neural network, defeats world champion Go player Lee Sedol in a five-game match. The victory is notable due to the vast number of possible moves in Go. Google subsequently acquires DeepMind for approximately USD 400 million.

10. 2023: The rise of large language models (LLMs), such as ChatGPT, significantly enhances AI performance and its potential to drive enterprise value. These generative AI models are pre-trained on vast amounts of raw, unlabeled data.

11. 2024: Businesses worldwide are leveraging generative AI to drive significant business value. According to a McKinsey Global Survey on AI, 65% of organizations report regular use of generative AI, nearly doubling the adoption rate from ten months earlier.

Future of AI:

Future developments in machine learning and artificial intelligence (AI) have the potential to drastically change several fields, including how humans live, work, and interact with technology. The possibilities for AI and ML are enormous, and they will present both obstacles and opportunities that need to be carefully managed. These technologies will have a significant impact on society as they develop, therefore in order to fully realize their potential in a morally and responsibly manner, technologists, legislators, and the general public will need to continue talking and working together.

• Applications of AI in healthcare, finance, and other industries.
• Recent advancements in deep learning and neural networks.

• Ethical implications and challenges of AI.