Where does GenAI fit within the AI landscape
Generative Artificial Intelligence, abbreviated as GenAI or Generative AI, refers to a type of artificial intelligence that is designed to generate new content, data, or information that is similar to, or indistinguishable from, what a human might produce. One of the prominent models in generative AI is the Generative Pre-trained Transformer (GPT), which is capable of generating coherent and contextually relevant text based on the input it receives.
Generative AI systems learn real-world data to generate seemingly new content, such as text, images, audio, video, tabular data, synthetic data, and code, with similar probabilistic distributions and characteristics of the real-world data. Generative AI can produce highly realistic and complex content that mimics human creativity, making it a valuable tool for many industries such as art, writing, software development, product design, health care, finance, marketing, fashion, and more. Generative AI is an exploding area within the field of artificial intelligence. Therefore, it is important to explore the AI backdrop and see where generative AI fits in.
Artificial Intelligence (AI)
SAS describes Artificial Intelligence (or AI) as “the science of designing ethical and transparent systems to support and accelerate human decisions and actions.”
AI is a branch of computer science that deals with the creation of intelligence agents, which are systems that can reason, learn, and act autonomously. Essentially, AI has to do with the theory and methods to build machines that think and act like humans. Advancements in computer processing and data storage made it possible to ingest and analyze more data than ever before. Around the same time, we started producing more and more data by connecting more devices and machines to the internet and streaming large amounts of data from those devices. All these advancements brought artificial intelligence closer to its original goal of creating intelligent machines, which we're starting to see more and more in our everyday lives. From recommendations on our favorite retail sites to auto-generated photo tags on social media, many common online conveniences are powered by artificial intelligence. Most artificial intelligence examples that you hear about today, from chess-playing computers to self-driving cars, rely heavily on machine learning that gives the computer the ability to learn without explicit programming.
Machine Learning (ML), Deep Learning (DL) and Beyond
Artificial intelligence is the broad science of mimicking human abilities, and machine learning is a specific subset of artificial intelligence that trains a machine how to learn. Machine learning automates analytical model building. Early work with neural networks stirred excitement for "thinking machines." A neural network is a kind of machine learning inspired by the workings of the human brain. It's a computing system made up of interconnected units (like neurons) that processes information by responding to external inputs, relaying information between each unit. The process requires multiple passes at the data to find connections and derive meaning from undefined data.
Machine learning and deep learning are subfields of artificial intelligence, and deep learning is a specific field of machine learning. Whereas machine learning mostly dealt with huge tabular data, deep learning generally dealt with images, audio, and text as input data that results in even larger tables. Recent deep learning breakthroughs are driving an AI boom. Deep learning uses huge neural networks with many layers of processing units, enabling them to process more complex patterns than machine learning in large amounts of data. Deep neural networks leverage artificial neurons, backpropagation, weights, and biases to identify patterns based on the inputs.
Artificial Intelligence applies machine learning, deep learning, and other techniques to solve problems in areas such as computer vision (CV), natural language processing (NLP) and internet of things (IoT). In addition, several technologies enable and support AI including GPUs, advanced algorithms and APIs.
- Graphical processing units, or GPUs, are key to AI because they provide the heavy compute power that's required for iterative processing. Training neural networks requires big data plus compute power.
- Advanced algorithms are being developed and combined in new ways to analyze more data faster and at multiple levels. This intelligent processing is key to identifying and predicting rare events, understanding complex systems, and optimizing unique scenarios.
- Application programming interfaces, or APIs, are portable packages of code that make it possible to add AI functionality to existing products and software packages. They can add image-recognition capabilities to home security systems and Q&A capabilities that describe data, create captions and headlines, or call out interesting patterns and insights in data.
In summary, the goal of AI is to provide software that can reason on input and explain on output. AI will provide human-like interactions with software and offer decision support for specific tasks, but it's not a replacement for humans. And it won't be any time soon.
Generative AI: Subset of All
Select any image to see a larger version.
Mobile users: To view the images, select the "Full" version at the bottom of the page.
Now we finally get to where generative AI fits into this AI landscape. Generative AI is a specific field of artificial intelligence that uses several techniques that continue to evolve. Foremost are deep learning algorithms such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), which consist of multiple layers of interconnected nodes that accept large quantities of data as inputs and identify patterns or structures within the data set. CNN is mostly used for image recognition and computer vision tasks. It works by pooling important features from an input image. RNN is used in advanced fields like natural language processing, handwriting recognition, time series analysis, and so on.
Generative AI can process both labeled and unlabeled data using many machine learning algorithms like neural network, k-nearest neighbor, and more. Generative AI also uses some of the reinforcement learning algorithms that focuses on sequential decision-making problems. The goal for such algorithms is to train an agent to perform actions in order to optimize an objective based on a reward signal.
A recent breakthrough of transformer models has revolutionized how we use machine learning to analyze unstructured data. The transformer in NLP is an innovative architecture that aims to solve sequence-to-sequence tasks while handling long-range dependencies with ease. One of the key innovations in transformer models is the use of the self-attention mechanism, which enables the model to weigh the importance of different parts of the input and output without using sequence-aligned RNNs or CNNs. Large language models (LLMs) are also a subset of deep learning. The use of deep learning models which will employ a Variational autoencoder or Generative Adversarial Network model uses methods for generating synthetic data.
AI systems can be used to do different things for us, such as forecast the demand for a product or the weather, recommend customized product suggestions for e-commerce sites, identify objects from images and video, detect fraudulent bank transactions, or optimize large scale production systems. On the other hand, generative AI is a type of AI whose job is simply to generate content.
Arguably, because machine learning and deep learning are inherently focused on generative processes, they can be considered types of generative AI, too. Broadly speaking, whereas traditional AI/ML systems recognize patterns and make predictions, GAI systems learn real-world data to generate data – like text, images, audio, video, synthetic data, code – with similar probabilistic distributions and characteristics of the real-world data.
Find more articles from SAS Global Enablement and Learning here.
