Data and AI Governance

Last week marked a significant milestone for Google, showcasing not only the launch of Gemini Ultra but also the AI Innovation session as part of their Cloud Roadmap series. This session was a focal point for articulating Google’s vision and strategy for enabling customers to harness the power of AI within their operations.

This blog post will look at the content shared in the AI Innovation session, and the next post will look into Gemini Ultra.

Google AI Innovation session

The AI Innovation session introduced their business model for customers to leverage AI. At its core, their approach is made up of three components:

Model Garden

Google’s Model Garden serves as a foundational entry point for companies embarking on their AI journey. Within this expansive repository, Google offers access to more than 130 AI models. This collection encompasses Google’s proprietary foundation models, including the notable Gemini, alongside specialized pre-trained models for specific tasks like translation. Additionally, it features a selection of open-source models, such as Meta’s Llama 2, broadening the scope of available AI tools. As the field of AI continues to evolve, we can anticipate the addition of new models to the Model Garden, further enriched by strategic partnerships and the development of new AI technologies.

Vertex AI

Once you’ve chosen a model from the Model Garden, Vertex AI provides the environment for its customization and enhancement. Google has designed Vertex AI to simplify the adaptation of models to align with your organization’s specific requirements. This platform comes equipped with a comprehensive set of tools for model training and fine-tuning, enabling you to optimize performance. Additionally, Vertex AI includes capabilities for monitoring model performance and assessing models for both safety and explainability. This ensures that you can not only tailor the models to your needs but also maintain oversight on their operational effectiveness and ethical considerations.

Duet AI

After tailoring the model to fit your organization’s specific requirements, Duet AI steps in to facilitate the creation of AI-powered applications using these customized models. Duet AI is designed to be inclusive, catering to both technical experts and those without coding expertise. It achieves this by offering AI-assisted development tools, enabling users who may not have programming skills to participate in the application development process. This approach democratizes the development of AI applications, ensuring that a broader range of users can contribute to and benefit from AI innovations within your organization.

So what’s my view on this?

Google have created an easy to understand deployment model for organizations to follow. Allowing their customers to select their base model, customize it as needed, then host it for further development is a powerful combo and its likely we will see other providers create similar models.

This approach also gives plenty of room for other organizations to partner with Google to provide their own LLM’s, customization services, or even app creation on Google hosted LLM’s. I expect we will see many startups offering AI services that are being hosted by Google in the near future.

On the flip side, this model, underpinned by Google’s vast size and structural rigidity, raises concerns about stifling innovation, fostering vendor lock-in, and complicating regulatory compliance. The dominance of Google could potentially lead to a monolithic ecosystem, limiting the diversity of AI development approaches by prioritizing compatibility over disruptive innovation

Stay tuned for our next article, where we’ll take a closer look at Gemini Ultra and its implications for organizations interested in the latest AI advancements.