Prerequisite: None
Enterprises are now incorporating a wider range of data types in applications of all sorts. Over the next several years, more enterprise applications will be built and run on multimodal data sets. Multimodal data management is designed to capture, correlate, and manage the text, images, audio, video, clickstream, sensor, and biometrics data modalities that might be associated with a particular entity, such as an individual’s profile in a customer database.
Multimodality is coming to such enterprise applications as search engines, which are evolving beyond simple keyword searches to support inputs and results in multiple formats. For deep learning and other advanced analytics, multimodal data can reveal deeper insights and support more nuanced annotation and labeling than is possible with any one type of data in isolation. Where generative AI is concerned, a growing range of sophisticated LLMs support text-to-image, image-to-text, and other multimodal capabilities.
As enterprises incorporate multimodality into generative AI and other applications, they will need to evolve their data platforms, pipelines, tools, and organizations. Join TDWI Senior Research Director James Kobielus for this presentation on the trends that are fueling multimodal data and the emerging practices that enterprises will need to adopt to keep pace. He will outline such emerging practices as:
- Controlling the costs associated with storing and processing vast amounts of multimodal data
- Discovering and acquiring the necessary data modalities when data is incomplete, unavailable, or too expensive
- Implementing multimodal integration, processing, storage, and governance into enterprise data lakehouse architectures
- Integrating new data modalities into enterprise data architectures that have been built to support structured, unstructured, and semistructured textual data
- Managing vectorized data embeddings within enterprise data infrastructures in support of unified search, query, and analytics across multimodal data sets
- Incorporating multimodal learning workflows into enterprise training of generative AI and other models that incorporate disparate data types
- Evolving enterprise metadata infrastructures to support real-time adaptive fusion, correlation, synchronization, alignment, and semantics across multimodal data within enterprise applications