GenAI is transforming how we create, analyze, and discover - pushing the boundaries of what's possible across an astounding range of fields.

General Modalities

The following table provides an overview of the general modalities in which Generative AI can be applied:

Modality	Examples
Language	Spoken and Written
Time series	Music, Speech, Finances
Visual 2D	Images, Diagrams
Visual 3D	3D Models, Virtual Reality
Visual 2D with time	Animated Graphics, Videos
Visual 3D with time	3D Animations, Simulations
Graphical	Relation and Influence Networks
Generally linear sequences	Genome, Proteome
Multidimensional Temporal sequences	Weather, Brain Recordings, Stock Market
Multimodal variants	Combination of the above methods

For a more detailed description of these modalities, refer to this section.

General Activities

Because at its core, GenAI works on Information, there are several fundamental ways in which Generative AI can be used. The application often depends on the field. Here are the core activities that can be used across many, if not all, fields of applications:

Creating Information

At its base, Generative AI is used to create information, such as new text or images. This creation can take several forms:

Expansion

Generating larger outputs from small inputs
Writing detailed documentation or articles
Brainstorming and ideation
Explaining complex concepts in detail
Creating training data for other AI systems

Reasoning

Evaluating trade-offs between different approaches
Analyzing complex scenarios and providing recommendations
Conducting risk assessments
Problem-solving with multiple variables
Strategic planning and decision-making

Converting Information

Generative AI can generate content in one domain with input from another. This includes:

Translating between languages (natural or programming)
Converting data formats (e.g., JSON to CSV)
Transforming natural language into structured queries
Converting visual information into textual descriptions
Transforming textual descriptions into visual representations

Compactifying Information

Generative AI excels at information compression and summarization:

Creating concise summaries of lengthy documents
Extracting key points from meetings or discussions
Distilling research papers into core findings
Generating executive summaries
Creating bullet-point highlights from detailed content
Even creating lossless compression at a fundamental level!

At a fundamental level Language Modeling Is Compression demonstrates 3x lossless compression of text and images.

Uses either newly trained 200K-3M transformer models or pre-trained Chinchilla models and achieves impressive compression rates. Details on implementation are somewhat hidden.

### Finding Information

Generative AI can understand and locate specific information:

Searching through documents for precise data points
Querying knowledge bases or databases
Finding relevant information in large datasets
Semantic search and relationship mapping
Answer extraction from complex documents

Taking Action

Generative AI can trigger and coordinate actions:

Generating executable commands
Orchestrating API calls
Managing workflow automation
Coordinating tool interactions
Implementing decision outcomes

Classifying and Predicting Information

While traditionally the domain of AI/ML, Generative AI can also perform:

Sentiment analysis and classification
Pattern recognition and prediction
Trend analysis and forecasting
Risk assessment and evaluation
Multi-label classification tasks

These activities can be combined to create more complex workflows, such as:

Finding relevant information, reasoning about it, and taking appropriate action
Converting information, compactifying it, and presenting insights
Creating new information based on patterns found in existing data

GenAI Use Cases