Snowflake
BUILD 2024: Agents & multi-modal apps are the tip of the
“Iceberg” | LinkedIn
Developers came together virtually and globally from November
12th through the 15th for BUILD, The Dev Conference for Apps & AI. The event
featured product announcements, luminary talks, technical sessions, hands-on
labs, and Gen AI bootcamps. Here are some highlights:
- **Snowflake Intelligence**: Enables business users to ask questions across
structured and unstructured enterprise data through data agents you create.
Check out the live demo
https://www.linkedin.com/pulse/build-2024-agents-multi-modal-apps-tip-iceberg-snowflake-computing-xiqnc/
- **Unistore**: Powered by Hybrid Tables, now generally
available.
- **Snowflake Open Catalog**: Uses the Apache Iceberg™ REST protocol to
integrate external data engines and apply your organization’s governance
controls across them.
- **Multi-Modal Models**: Includes Llama 3.1 and 3.2 models, Anthropic’s Claude
3.5 models, more languages, context window choices, and embedding model sizes.
Supports container runtime for notebook-based AI workflows.
- **Secure Model Sharing**: Develop, distribute, and monetize AI models to third
parties on Snowflake Marketplace.
**Luminary Talks**
Catch up with Dr. Andrew Ng, founder and Executive Chairman of LandingAI, on the
rise of AI, agentic reasoning, and the role of unstructured data such as text,
images, video, and audio on automation and application building across
industries.
Here’s a summary of Andrew Ng's keynote on the rise of
AI agents and agentic reasoning -
https://www.youtube.com/watch?v=KrRD7r7y7NY 27 mins - Andrew Ng Explores
the Rise of AI Agents and Agentic Reasoning | BUILD 2024 Keynote
- AI as General Purpose Technology (00:01-00:43)
: Andrew compares AI
to electricity, emphasizing its versatility and broad potential for new
applications.
- Focus on the Application Layer (01:16-01:45)
: While much buzz
surrounds foundational technologies, Andrew believes the real opportunities
lie in the application layer, where AI can create significant value.
- Faster Machine Learning Development (02:19-02:53)
: Generative AI has
accelerated the development process, allowing teams to prototype and deploy
applications in days instead of months, encouraging rapid experimentation.
- New Prototyping and Iteration Models (03:27-03:58)
: Fast iteration
and testing new prototypes quickly are becoming key strategies for AI teams,
facilitating the creation of new products through trial and error.
- AI Evaluation Bottleneck (04:29-05:01)
: The shift to generative AI
has created challenges in evaluation, requiring new approaches to test and
refine models efficiently.
- Pressure on Non-ML Aspects of AI Development (05:01-05:30)
: While
machine learning modeling has sped up, other stages like software
integration, deployment, and operations need to catch up.
- Responsible Innovation (06:05-06:39)
: Teams are moving fast but in a
responsible way, emphasizing quick prototyping without compromising on
safety and ethics.
- Agentic AI Workflows (06:39-07:14)
: The rise of agentic AI, where AI
agents perform tasks iteratively (e.g., drafting, researching, revising),
represents a significant advancement in AI capabilities.
- Key Agentic Design Patterns (09:18-14:44)
: Four design patterns
emerge in agentic AI: reflection (self-critique and improvement),
task-oriented use (e.g., API calls), planning (sequencing actions for
complex tasks), and multi-agent collaboration (multiple agents solving parts
of a task together).
- Visual AI and Data Processing (17:54-20:21)
: Agentic workflows are
also being applied to image and video data, allowing teams to process and
extract meaningful information from large visual datasets more effectively.
Watch the replay featuring Katie Bauer (Head of Data at
GlossGenius), Kasia Sitkiewicz (Staff Product Manager at GitHub), and Holden
Karau (Open Source Engineer at Netflix)
In case you missed it, you can still access the sessions
on-demand or view them from your existing registration. Watch now
https://www.linkedin.com/pulse/build-2024-agents-multi-modal-apps-tip-iceberg-snowflake-computing-xiqnc
.
12/07/2024
|