#268 Scaling AI in the Enterprise with Abhas Ricky, Chief Strategy Officer at Cloudera
DataFramed
Deep Dive
Why is data quality becoming more critical in AI applications?
Data quality is critical because it directly impacts the trustworthiness and fidelity of AI outputs. Enterprises are increasingly focusing on structuring and managing their data to ensure it can be used effectively for training large language models and other AI frameworks.
What are the key challenges enterprises face in managing data for AI?
Enterprises face challenges such as data sprawl, ensuring data security and governance, and managing the total cost of ownership. Data sprawl occurs due to various teams creating their own systems, leading to scattered data that is hard to manage and secure.
What does 'data as an asset' mean in the context of AI?
Data as an asset means treating data as a valuable resource that can be used to train AI models and generate insights. It involves structuring data so it can be easily accessed and applied to AI frameworks, enabling faster and more efficient use cases without needing extensive upskilling.
How do enterprises typically differ in their AI adoption journeys?
Enterprises can be categorized as leaders or laggards. Leaders have already solved data engineering problems and are scaling generative AI applications, while laggards are still in the early stages, focusing on building their core data platforms before diving into AI.
What steps should enterprises take to prepare their data infrastructure for AI?
Enterprises need to solve data engineering problems first, including data ingestion, preparation, and serving, before applying AI models. This ensures the data is ready for use in AI applications, enabling a smooth transition from data to insights to action.
Why is compute cost a significant factor in AI adoption?
Compute cost is a major factor because running AI workloads, especially with large language models, is energy-intensive and expensive. Organizations need to plan for increased IT budgets to account for the compute requirements of AI applications.
How can enterprises decide which AI models to use for their use cases?
Enterprises should evaluate their use cases based on the complexity and importance of the outcomes. Simple tasks like text summarization can be handled by smaller models, while more complex tasks like drug discovery require larger, more compute-intensive models.
What are the key steps in scaling AI applications in enterprises?
Scaling AI involves three main steps: scaling the generative AI infrastructure, embedding AI into business processes, and adopting advanced technologies like agent-based applications. This ensures AI is integrated into daily operations and can handle high-volume tasks efficiently.
Why is hybrid AI important for large organizations?
Hybrid AI allows organizations to run AI applications in the form factor of their choice, whether public or private cloud, depending on data residency and security needs. This approach minimizes costs and ensures sensitive data remains within secure environments.
What final advice does Abbas have for organizations looking to improve their AI capabilities?
Abbas advises that AI is a team sport, requiring collaboration across teams. Enterprises must also ensure they trust their data and embed AI into their normal processes to scale effectively, avoiding it becoming a mere science experiment.
- Increased focus on data quality and trustworthiness for AI applications
- Data sprawl creates issues around data security and cost
- Data as an asset is becoming increasingly important
Shownotes Transcript