What is Generative AI?

When OpenAI’s ChatGPT generative artificial intelligence (AI) service hit the internet, it spread so quickly that, to many people, generative AI meant AI. Generative AI performs a subset of AI tasks, and not only are they the most attention-getting, they’re the ones that have the potential to create the most immediate benefit for companies. Here are the details on the GenAI category:

Generative AI Explained

What is Generative AI?
What are the Features of Generative AI?
How Does Generative AI Work?
Why is Generative AI Important?
What are Generative AI Use Cases?
Which Industries are Adopting Generative AI?
Generative AI Companies

What is Generative AI?

Generative AI is called that because, well, it generates. It creates new content based on existing content. What’s particularly noteworthy about generative AI is that it does this through natural language requests, and it "doesn’t require knowledge of or entering code,” Gartner says.

Generative AI isn’t actually all that different from the auto-complete functions present in some applications, such as Google Chrome and Microsoft Word. It’s just a matter of scale. Auto-complete looks back at just a few words at a time and generates just a word or two. Generative AI takes a large amount of input — including the terabytes of data on which it was originally trained — and consequently can generate a large amount of output.

GenAI Adoption

Generative AI is being accepted and used more quickly and broadly than personal computers and the internet were when they were introduced, according to the report "The Rapid Adoption of Generative AI." In August 2024, 39% of the U.S. population aged 18-64 used generative AI, more than 24% of workers used it at least once in the week prior to being surveyed and nearly one in nine used it every workday.

And it’s not stopping. Gartner predicts that by 2026, more than 80% of enterprises will have used generative AI APIs or models or deployed GenAI-enabled applications in production environments, up from less than 5% in 2023. Gartner compares generative AI to having an impact similar to the steam engine, electricity and the internet. “The hype will subside as the reality of implementation sets in, but the impact of generative AI will grow as people and enterprises discover more innovative applications for the technology in daily work and life,” Gartner predicts.

This video by IBM offers an additional explanation of generative AI technology:

See more: 10 Top Generative AI Certifications

What are the Features of Generative AI?

While not all generative AI apps use foundation models, many do, such as OpenAI’s Generative Pre-Trained Transformer (GPT) that ChatGPT uses, the Gemini model that Google’s Gemini uses and the DALL-E model that OpenAI’s DALL-E image creator uses, according to the Ada Lovelace Institute. Depending on the data it was trained on and the request it’s given from the user, a generative AI app could generate text, images, sounds, animation, 3D models and other types of data, according to NVIDIA. Previously known as transformer models, they are machine learning (ML) models based on large amounts of data.

Depending on its type, a foundation model performs tasks, such as language processing, visual comprehension, code generation, human-centered engagement and speech to text, according to Amazon.

Large Language Models and Natural Language Processing

A specialized kind of foundation model is a large language model (LLM), which is dedicated to text in some way, either taking input from text, generating text or both, according to Amazon. Typically they include another important feature of generative AI, natural language processing (NLP), which converts human language into something the LLM can understand and vice versa. For example, ChatGPT is an NLP interface to the GPT LLM; it converts the human input into prompts and then converts the LLM outputs into text.

NLP is among the most useful AI functions, because it makes it easier for humans to communicate with an AI system without having to know programming. The AI system can communicate back to the user without programmers specifically having to create the responses. NLP enables AI functions, such as processing data collections of voice recordings and written texts, automating interactions with human users, interpreting user queries, sentiment analysis, translation and content moderation, according to Cloudflare. It’s also the power behind virtual assistants, such as Siri and Alexa, and search engines, such as Google.

In addition, NLP is used for preprocessing data or turning data into a form that’s easier for a model to understand, Cloudflare says. Examples include converting all text to lowercase; removing endings, plurals and possessives; breaking text into smaller pieces; and removing words, such as definite and indefinite articles and prepositions, that the system doesn’t need.

Types of GenAI Models

There are three key kinds of generative models, each with advantages and disadvantages, according to NVIDIA.

Generative adversarial networks (GANs): GANs, one of the first types of generative models, pit two neural networks against each other: a generator that generates new examples and a discriminator that learns to distinguish the generated content as either real (from the domain) or fake (generated). “The two models are trained together and get smarter as the generator produces better content and the discriminator gets better at spotting the generated content. This procedure repeats, pushing both to continually improve after every iteration until the generated content is indistinguishable from the existing content,” NVIDIA says. The advantage of this method is that it produces high-quality quickly but isn’t very diverse. They’re most often used for domain-specific data generation
Variational auto-encoders (VAEs): “VAEs consist of two neural networks typically referred to as the encoder and decoder,” NVIDIA says. “When given an input, an encoder converts it into a smaller, more dense representation of the data. This compressed representation preserves the information that’s needed for a decoder to reconstruct the original input data, while discarding any irrelevant information. The encoder and decoder work together to learn an efficient and simple latent data representation. This allows the user to easily sample new latent representations that can be mapped through the decoder to generate novel data.” But while they’re faster, the results aren’t as detailed as those of diffusion models.
Diffusion models: Generative models use a two-step process during training, forward diffusion and reverse diffusion. “The forward diffusion process slowly adds random noise to training data, while the reverse process reverses the noise to reconstruct the data samples,” NVIDIA says. “Novel data can be generated by running the reverse denoising process starting from entirely random noise.” The advantage is that they can offer the highest quality of output but take longer to train and to run. This model is what’s used for popular generative AI products, such as ChatGPT for text and DALL-E for imagery.

How Does Generative AI Work?

In its most basic form, generative AI works by using neural networks to analyze data and then generate new and original content based on it, according to NVIDIA. AI systems can also work unsupervised, or semi-supervised, to create and use foundation models.

Large language models are a particular kind of foundation model that is specifically focused on language-based tasks, such as summarization, text generation, classification, open-ended conversation and information extraction, according to Amazon. They’re trained on extremely large amounts of data — essentially, the entire internet. Through that exposure, they learn to apply their knowledge in a wide range of contexts, which means they can consider billions of parameters and go on to generate content from very little input.

“What all of these approaches have in common is that they convert inputs into a set of tokens, which are numerical representations of chunks of data,” MIT says. “As long as your data can be converted into this standard, token format, then in theory, you could apply these methods to generate new data that look similar.”

Why is Generative AI Important?

There are several why generative AI represents a seismic shift in technology, according to Salesforce:

It improves efficiency and productivity, because it generates so much content so quickly.
It improves customer relationships, because it responds quickly with personalized information.
The personalization is much richer, because of how much data the generative AI system can absorb.
It helps sales teams focus on sales, using it to create notes and follow-ups.
It makes creative teams more productive, because it brings ideas to life more quickly.
Developers are more productive, because they can use it to automate repetitive tasks.

GenAI is Growing Markets and Production

In dollars, McKinsey & Company predicts that generative AI could add the equivalent of $2.6 trillion to $4.4 trillion annually just across the 63 use cases it analyzed. “By comparison, the United Kingdom’s entire GDP in 2021 was $3.1 trillion,” McKinsey says. “This estimate would roughly double if we include the impact of embedding generative AI into software that is currently used for other tasks beyond those use cases.”

Goldman Sachs predicts that generative AI could drive a 7% or almost $7 trillion increase in global gross domestic product (GDP) and lift productivity growth by 1.5 percentage points over 10 years.

Statista predicts the global market for AI is likely to reach up to nearly $2 trillion by 2030.

But is generative AI proving to be worthwhile? According to Gartner this summer, it is. “Earlier adopters across industries and business processes are reporting a range of business improvements that vary by use case, job type and skill level of the worker,” Gartner says. Survey respondents reported a 15.8% revenue increase, 15.2% cost savings and 22.6% productivity improvement on average.

What are Generative AI Use Cases?

Any job that creates content could be a use case. In fact, according to the summer 2024 Deloitte CEO Survey, 43% of respondents said they were actively experimenting with or regularly incorporating generative AI into their daily operations. Even more were using it in their organizations to identify growth opportunities, discover new insights (45%), and accelerate innovation (43%).

In addition, 49% said they were experimenting with the tool themselves to become familiar with it to help them in their own jobs, particularly for communication, content generation and information synthesis support, Deloitte reports. What’s more, 25% of CEOs said they were actively experimenting with it, while 19% said they were using it regularly to support their own work. Only 12% said they had never used it.

On a more company-wide basis, here are some key GenAI uses cases in different departments:

Code generation: software developers and programmers use generative AI to write code, update it, look for bugs, test it and document it.
Product development: product managers use generative AI to evaluate and adjust designs, optimize them and collect user feedback.
Sales and marketing: marketing managers use generative AI to look for approaches; generate personalized communication across email, social media and text messages and for blog posts and websites; and analyze the performance of content.
Project management and operations: project managers use GenAI for the detailed work of summarizing documents, generating tasks, predicting risk and forecasting timelines, using historical data.
Graphic design and video: creatives use GenAI to create videos and imagery more quickly.
Business and employee management: GenAI gives employees documentation, reviews and other feedback.
Customer support and customer service: GenAI allows chatbots to handle routine interactions at any time of the day while humans handle more complicated scenarios.
Fraud detection and risk management: underwriters and claims adjusters use GenAI tools to look at policies and claims, examine data for patterns associated with fraud and create reports on potential risks.
Generating synthetic data for training and testing: organizations that are concerned about privacy and bias quickly create all the testing data they need, using the specific parameters they require.

Highly regulated industries, such as health care, insurance and education, may be more limited because of their legal and compliance requirements.

Which Industries are Adopting Generative AI?

The following industries demonstrated the biggest interest in GenAI based on the share of their organization’s digital budget invested in generative AI, according to a McKinsey & Company report:

Technology
Energy and materials
Financial services
Media and telecommunications
Consumer goods and retail
Advanced industries
Business, legal and professional services
Health care, pharmaceuticals and medical products

Gartner breaks the type of work that industries are expecting generative AI to do into two main camps. Generative AI will augment core processes with AI models in industries, such as pharmaceutical, manufacturing, media, architecture, interior design, engineering, automotive, aerospace, defense, medical, electronics and energy. GenAI will also augment the supporting processes that span many organizations in marketing, design, corporate communications and training and software engineering, Gartner predicts.

Generative AI startup costs also vary per industry, Gartner notes. For example, building a custom generative AI LLM for the medical, insurance or financial services industry from scratch could cost from $8 million to $20 million in startup costs, plus ongoing costs ranging from $11,000 to $21,000 per user per year. On the other hand, a commercial generative AI coding assistant could incur startup costs of just $100,000 to $200,000, plus ongoing costs ranging from $280 to $550 per user per year, Gartner says.

Generative AI Companies

With generative AI playing such a major role in the AI marketplace in general, it’s not surprising that many of the leading generative AI companies are also considered leading AI companies in general. Here are some of the top GenAI companies:

Adobe: Firefly is the company’s image creation and editing program
Amazon: The company’s Q apps became available in July; in addition, the company has invested $4 billion in Anthropic
Anthropic: Claude is a generative AI intelligent assistant; one distinction is that Anthropic is a public benefit company, meaning its purpose is “the responsible development and maintenance of advanced AI for the long-term benefit of humanity”
Google: Google’s AI line, model and chatbot go by the name Gemini; Gemini has access to more recent data than some competitors
IBM: The company released Watson in 2010
Meta: Meta AI offers an intelligent assistant as well as image generation
Microsoft: As one of the major investors in OpenAI, Microsoft’s Copilot is similar to ChatGPT
OpenAI: Its release of ChatGPT pioneered the boom in generative AI, and it continues to evolve the product
Stability AI: The company is focused more on imagery and video
XAI: Founded by Elon Musk, the company’s model and chatbot Grok, and it generates images – without the limitations imposed by other generative AI graphics programs — as well as text