Welcome to the AI jungle! Now what? by Kevin Dubois

The AI landscape is growing rapidly with over 1 million models available on platforms like Hugging Face, creating challenges for organizations to navigate effectively

Open source is crucial for AI adoption, but many models claim to be open source while having limitations on contributing knowledge back or unclear training sources

Multiple personas need to work together to implement AI successfully:

• Data Scientists for model development
• ML Engineers for platform setup
• Application Developers for integration
• Operations teams for infrastructure

Projects like Open Data Hub and InstructLab help democratize AI by:

• Providing opinionated platforms for model training and serving
• Enabling knowledge contribution back to models
• Offering taxonomy-driven data curation
• Supporting local machine training and experimentation

Kubernetes provides a good foundation for serving AI models with benefits like:

• Scalability
• Easier deployments
• Resource management
• Integration with existing infrastructure

Tools like LangChain4j and Quarkus help Java developers integrate AI into applications without deep ML expertise

Organizations face common challenges with AI adoption:

• Limited access to cloud resources/GPU compute
• Need for local development capabilities
• Lack of standardization across providers
• Requirements for model governance and security

Open standards and collaboration between vendors is essential, similar to how Linux and cloud computing evolved

The path to AI adoption typically involves:

• Initial experimentation with existing models
• Refining models for specific use cases
• Operationalizing with proper infrastructure
• Integrating into applications
• Monitoring and governance

Focus should be on democratizing AI access while maintaining open source principles of transparency and collaboration