APC 技術ブログ


株式会社 エーピーコミュニケーションズの技術ブログです。

Part 3: Introduction to Data + AI Summit 2024 (Compound AI System)


This is May from the Lakehouse Department of the GLB Division.

This year, Databricks' Data + AI Summit 2024 (DAIS) will be held in San Francisco from June 10th to 13th.

This article explains the highlights and key points of DAIS.



In this article, I would like to share a topic that I am personally interested in from the DAIS sessions: Compound AI Systems. Increasingly, we are building Compound AI Systems (systems that combine multiple AI models or use external components) to tackle complex AI tasks. The "COMPOUND AI SYSTEMS WORKSHOP" will bring together researchers to discuss challenges in the design, efficient implementation and training, and evaluation methods and metrics of compound AI systems.


Table of Contents

What is a Compound AI System?

It refers to the integration of various AI technologies into a single solution in a systematic way to approach complex business problems.

Reference: https://www.sbbit.jp/article/cont1/125680

Why Compound AI Systems?

  • Ability to handle a variety of data types (structured data, unstructured data, etc.)

  • Ability to solve one problem in multiple ways

  • Creativity is generated by combining various components, resulting in innovative functions

  • Enabling human-like decision-making from multiple perspectives

  • Improves data analysis and diversifies business based on the analysis results

The global market size for complex AI is expected to grow from US$900 million in 2023 to US$4.4 billion in 2028, at a CAGR (compound annual growth rate) of 36.5%.

Reference: https://www.gii.co.jp/report/mama1287803-composite-ai-market-by-offering-software-hardware.html

Challenges for Compound AI Systems

  • It is difficult to select the optimal combination of components and effectively manage their interactions
  • It is necessary to effectively allocate resources such as computing time and budget to each part
  • New standards and evaluation models are needed to measure overall performance, reliability, etc.
  • There is a possibility of unexpected security risks

Challenges such as these are expected.


This is the end of this article, but as with last year, we will introduce a site that will provide on-site reports later, stay in tuned!

About the special site during DAIS

This year, we have prepared a special site to report on the session contents and the situation from the DAIS site! We plan to update the blog every day during DAIS, so please take a look.