Skip to main content

The Comprehensive Beginners’ Guide to AI

Artificial intelligence (AI) is arguably the hottest topic in the tech world today. But what is AI, really? And what are its implications?

AI really started in the 1950s with Alan Turing and later the Dartmouth Conference. You can read more details here. IBM’s Deep Blue computer beating the reigning world chess champion, Gary Kasparov, over six games in 1997 brought AI back into the public’s attention. Fast forward to today, and AI is permeating every area of our lives, so I thought it would be useful to go back to basics and understand the building blocks to AI.

A Glossary of AI Terms

Let’s start with a definition. AI is an attempt to give machines some of the capabilities that intelligence gives humans. It uses algorithms or rules to guide the software in what it must do.
Next, let’s look at some of the terms used with AI.

Machine Learning

Machine learning uses algorithms that can learn from data and improve their performance, but don’t require programming by humans. Training a program involves feeding it data and letting it try to recognize patterns and then make decisions or predictions. This can be done with:

  • Reinforcement learning: Advanced Reinforcement Learning is a way to reward the computer when it gets something right. It’s used in gaming, robotics, and autonomous vehicles.
  • Unsupervised learning: The computer explores the data finding patterns and relationships of its own.
  • Supervised learning: The software is given labelled data allowing it to check its answers against the ones given. That way, it can modify its algorithms appropriately.

Deep Learning and Neural Networks

Deep learning uses neural networks to allow the software to recognize patterns and then make decisions. A neural network is composed of interconnected neurons that process information by passing it through layers of neurons. The input layer receives the data, and the other layers process and transform the data, finally giving it to the output layer in the form of a prediction. Neural networks can learn from data—a process called backpropagation—which improves its predictions. They can be used for things like facial recognition and translation.

Deep learning is a subset of machine learning that uses artificial neural networks and can learn from unstructured data, such as images, audio and text. Convolutional neural networks (CNNs) are great for vision tasks and can be used to recognize objects or detect faces. Recurrent neural networks (RNNs) are excellent at processing sequential data. That makes them good at language processing (e.g., NLP) and time-series analysis. Newer RNNs have long short-term memory (LSTM) modules, which allows them to remember relevant sequences for longer.

Natural Language Processing

Natural language processing (NLP) means that the software can understand human speech and respond. Think virtual assistants, chatbots and even translators. They can also be used for sentiment analysis—determining the emotion expressed in written text.

Computer Vision

Computer vision aims to give machines the ability to see and interpret visual information from images or video. It can be used for object detection, image classification, facial recognition and autonomous driving.

Generative AI

Generative AI can produce or generate high-quality content, which could be text, images or videos. It uses a large language model (LLM), which is initially given large amounts of data, i.e., its pre-training data set content. The LLMs uses transformers to convert sentences and data sequences into numbers known as vector embeddings. The data is then classified. The model then learns. Supervised learning makes the output seem natural.
ChatGPT and Gemini are examples of generative AI.

AI Subtypes

There are also different types of AI, including:

  • Narrow AI (weak AI): Brilliant at one task (like being a virtual assistant) or a range of similar tasks.
  • General AI (strong AI): This aims to replicate human intelligence. It can think, learn and understand. This is what researchers are working toward now.

Common AI Applications

Popular AI libraries include:

  • TensorFlow, for building and training AI models.
  • Scikit-learn, for machine learning algorithms and data pre-processing.
  • NumPy, for working with numbers and arrays.
  • Pandas, for data manipulation and analysis.

What examples of AI are you likely to find on a mainframe? There’s watsonx Code Assistant for Z, which is a generative AI product built on the watsonx enterprise AI platform. It can help developers translate mainframe COBOL applications into Java for cloud deployment. It offers improved testing, faster rewriting of functionality into Java, and lower costs associated with updating the old COBOL code. Code Assistant also provides automated testing processes and can be used in each step of the modernization process, converting the existing COBOL code to Java. watsonx Code Assistant for Red Hat Ansible Lightspeed is also available.

AI software from IBM includes:

  • Watson Assistant: Build and deploy AI-powered chatbots and virtual assistants that understand natural language and can interact with users in a conversational way.
  • Watson Discovery: Extract insights from unstructured data like text, images and audio to gain a deeper understanding of your customers, products and operations.
  • Watson Studio: A cloud-based platform for building, training and deploying machine learning models.
  • Watson Knowledge Catalog: Organize and discover your organization’s knowledge assets to improve decision-making and collaboration.
  • AutoAI: Automate the machine learning process to make it faster and easier to build and deploy AI models.
  • A studio environment for training, validating, tuning and deploying generative AI and machine learning models.
  • Foundation Models: Pre-trained AI models that can be customized for specific tasks, such as language translation, text summarization and image recognition.
  • Watsonx.governance: Tools and services for managing and monitoring your organization’s AI projects to ensure responsible and ethical use of AI.

In addition, IBM is offering the following facilities:

  • AI Toolkit: Use AI open-source software on IBM Z and LinuxOne.
  • Python AI Toolkit: Access a library of relevant open-source software.
  • AI embedded into real-world apps: Build and deploy machine learning models.
  • Accelerate TensorFlow Inference: Import TensorFlow models trained anywhere and deploy them. IBM ZDNN Plugin for TensorFlo
  • In-memory computing performance: An in-memory compute engine and analytics run time.
  • IBM Z Deep Learning Compiler: Compile .onnx deep learning AI models into shared libraries and run them on IBM Z.
  • Popular open-source tools: Use Anaconda on IBM Z and LinuxONE, and leverage industry-standard packages such as Scikit-learn, NumPy and PyTorch with zCX containers.

The Future of AI

AI has a huge future. CICS programmers, for example, will be looking to call out from CICS to generative AI to help people perform their job faster and more easily. There’s lots of talk of AI being used in ransomware attacks on mainframes, and mainframes running a security AI to defend against such attacks. If you haven’t already, now is the time to get in on the ground floor and understand AI and its implications.