GenAI Use Cases
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