Jun 02, 2023

Model Citizens, Google Updates Vertex AI

Like human intelligence, AI models are evolving.

Intelligence evolves. Across human knowledge landscapes and in the Machine Learning (ML) that we use to drive Artificial Intelligence (AI), we start with a base level of understanding and then extend our ability to comprehend, infer, reason and calculate. Because the same basic learning constructs apply to both machine brains and our own, we need to develop systems to deliver knowledge as it expands over time.

For us complex humans, we might think about our school and college systems, the use of books and the limitless expanses of wiki pages that now populate the web. For machines and AI, we talk about the use of ‘models’ i.e. a means of architecting the relationship between different pieces of information to give it order, structure, value and a denoted relationship to other elements of data. The woven fabric of connections inside an AI model resembles a spaghetti-like maelstrom of logic - and in geometry, we call a meeting point between any of those lines and edges a vertex.

The software engineers at Google paid homage to their favorite geometry teachers when they named the company’s Machine Learning (ML) platform Google Vertex. Built to enable data scientists and data engineering specialists to train and deploy ML models and AI applications, Google Vertex works to customize Large Language Models (LLMs) used in AI-powered applications.

Remembering that ML models start as foundational models (sometimes written as foundation model, or base model), this intelligence needs to evolve like any other. As previously explained here, The Stanford Institute for Human-Centered Artificial Intelligence (HAI) described foundational models as ‘critically central yet incomplete [in] character. So then, what has the Google Cloud team been doing to make Vertex smartness smarter?

Like almost every other enterprise technology vendor of any reasonable size or stature, Google added Generative AI support on Vertex AI (in this case the company did this around six months ago). Now, the company is expanding Vertex AI’s capabilities with a view to making it easier to experiment and build with foundation models, customize them with an enterprise’s own data sets and make them simpler to integrate and deploy them into applications. All of that also needs to happen with privacy and safety features within the realms of what we now like to call responsible AI, so this is all built-in.

Where does a data engineer keep their machine learning models if they are going to be organic and capable of nurturing new sprouts and fronds? In a model garden, right? Google Vertex Model Garden is a ‘curated collection’ of machine learning models and software tools. It currently features over 100 enterprise-ready foundation model Application Programming Interfaces (APIs), open source models and task-specific models that come both from Google, plus also from third parties.

“Many customers start their generative AI journey in Vertex AI’s Model Garden, accessing a diverse collection of curated large models available via APIs. Developers and data scientists can navigate Model Garden to select the right models for their use cases, based on capabilities, size, possibility for customization, and more - ensuring they have not only access to powerful models, but also the choice and flexibility required to tune and deploy models at scale,” noted Google’s Amin Vahdat, VP/GM ML for systems and cloud AI & June Yang, VP for cloud AI & industry solutions, in a technical blog released in line with the Google Cloud Next event.

Vahdat and Yang point to new models in Model Garden designed to further the company’s customer commitment to providing choice with a diverse and open ecosystem. As well as updates to several of Google’s first-party foundation models, the company is also bringing the expertise of Google DeepMind (the company’s deep neural intelligence service) to its users at this level. Significantly here, Vertex AI Extensions will enable models to retrieve real-time data and take real-world actions. Also, Vertex AI data connectors offers offer data ingestion and read-only access across various sources.

Model Garden is designed by Google to offer enough variety to allow enterprises to match models to their specific operational needs. There is also the ability to get what Google calls ‘full transparency’ into a model’s weights (how much one element of data is defined to have a relationship with another) and artefacts (the input or output of a model, or an interim result produced by the software tools in the model) for compliance and auditing support purposes.

"While foundational models are powerful, they are frozen after training, meaning they are not updated as new information becomes available, and thus may deliver stale results,” explain Vahdat and Yang. "Vertex AI Extensions is a set of fully-managed developer tools for extensions, which connect models to APIs for real-time data and real-world actions. With Extensions, developers can use pre-built extensions to popular enterprise APIs, or build their own extensions to private and public APIs. Developers can use extensions to build powerful gen AI applications like digital assistants, search engines and automated workflows.”

For example, a developer can use pre-built extensions for a Human Resources (HR) database and Vertex AI Search to create a chatbot that helps workers complete HR tasks in natural language, such as looking up benefit deadlines or travel policies that may be subject to change over time. Another example might be an application designed to to analyze software code for vulnerabilities. Software application development professionals can use extensions to ingest internal codebases and lookup evolving security threats in real-time.

“Google Cloud’s out-of-the-box AI and API support has revolutionized our workflow. The integrated AI environment offered by Google Cloud is a key ingredient in our application architecture that blends foundation and proprietary ML models to solve the scalability challenges of real-time content personalization,” said Tommaso Vaccarella, co-founder at Connected Stories. “Beyond the innovation, the stringent data safety measures reassure us and our clients that sensitive information remains protected. Google Cloud provides the power and speed we need to bring cutting-edge enterprise-ready solutions to market.”

All of these developments hopefully give us some extra insight into not just why AI is changing, but how its mechanics are working, why we have the ability to do new things with new techniques and what it will mean for the now altogether smarter applications that we will all be using on a daily basis.

CEO of Google Cloud Thomas Kurian calls it, “An entirely new era of cloud, fuelled by generative AI."

It may have been more accurate to suggest that this is an entirely new era of data engineering architecture and application intelligence science, which itself is fuelled by cloud backbones and their ability to support it. But hey, it essentially comes to the same thing i.e. smarter, slicker, smoother technology that can make our lives better.

As in fashion, it is in AI, this year’s model may not be next year’s model, so let’s all stay on trend.