Intro: Generative AI on Google Cloud

In the last year, we have seen a remarkable change in the performance of GenAI and its potential to drive enterprise value significantly. As such, Google Cloud has released a set of GenAI tools and platforms for end users, app builders, and AI developers to accelerate the enterprise adoption of GenAI. This post will discuss these tools and how they can be leveraged to improve enterprise performance and drive enterprise value. Generative AI can be a game changer in the future because of its capability to generate new content instantly based on training data. As such, it can play a critical role in improving the productivity and efficiency of individuals and businesses. GenAI models are capable of advanced tasks, empowering enterprises to do the following:

Accelerate content creation and discovery to enrich marketing efforts.
Speed up research process.
Deliver efficient round-the-clock customer service.
Improve developer efficiency.

This will bring forth problem-solving at a rate that supersedes human capabilities and at a fraction of the time and cost.

What is GenAI – From Traditional AI to GenAI

GenAI is a subset of deep learning that can generate new content such as text, images, audio, and even code. While traditional AI models are trained on labeled data, GenAI models are trained on small amounts of labeled data, accompanied by large amounts of unlabeled and unstructured data (semi-supervised learning). This way, GenAI models learn the patterns in unstructured content so that they can generate new content based on these patterns. Traditional approaches in building AI models require explicit programming of rules, logic, or training machine learning (ML) models using labeled data (supervised training). This limits the ability of traditional ML models to handle complex and diverse datasets.

Advancements in the field of AI brought about the rise of artificial neural networks and deep learning. Inspired by the human brain’s architecture, artificial neural networks comprise numerous interconnected nodes capable of learning from data and adapting to new patterns without explicit programming. This formed the foundation for Generative AI.

How GenAI came into the picture

The real breakthrough in GenAI came from the discovery of transformers – a type of neural network architecture that could handle sequential data, such as text, in a highly efficient manner. This architecture brought a paradigm shift in the work of AI where ML models could take prompts and generate new content, hence the name Generative AI. The transformer architecture gives GenAI models natural language processing capabilities which makes them perfect for a variety of use cases, such as the following:

Text-, image-, audio-, and video generation
Machine translation
Sentiment analysis
Question answering
Document summarization

Instead of creating individual ML models for specific tasks, GenAI allows the creation of large multi-purpose ML models, called Foundation Models, that can be finetuned for specific use cases.

GenAI on Google Cloud

Creating GenAI solutions can be challenging for any business, regardless of its scale. Doing this requires enterprises to have robust computational resources, infrastructure, and large amounts of high-quality training data to produce competent GenAI models. However, most enterprises lack such resources that are available on Google Cloud, including a set of ready-to-use GenAI products to enable users to build transformative applications based on their data.

Vertex AI Search and Conversation (formerly Gen App Builder)- Enterprise Search and Conversational AI
Vertex AI – Generative AI Studio and the Model Garden.

Vertex AI Search and Conversation (formerly Generative AI App Builder)

Vertex AI Search and Conversation provides a way to combine organization data with Google Cloud’s large language models (LLMs) to allow end users to find relevant information in an interactive and personalized manner. It enables users to create GenAI-powered chatbots and search engines using Enterprise Search and Conversational AI.

Enterprise Search

Organizations can create powerful search experiences for their customers and staff by developing GenAI-powered search engines using Enterprise Search. Built with NLP capabilities, Enterprise Search allows semantic search to simplify the retrieval of accurate information from an organization’s database or external resources. Industry applications built with Enterprise Search can take multi-modal inputs, such as natural language, map coordinates, etc., and produce grounded responses and personalized recommendations. In addition, they can conduct multi-turn conversations that allow follow-up questions. Enterprise search empowers business search engines to answer nuanced questions like:

“Which of your products best suits my needs for [specific use case]?” – customers.
“What are the main drivers behind our website traffic and conversion rates?” – employees.

This allows organizations to not only search for solutions but also explore their data and external resources to discover new insights.

Conversational AI

This technology allows organizations to create chatbot apps capable of simulating human-like conversations with users looking for information from business websites. The integration with Google Cloud’s LLMs extends the capabilities of these chatbots beyond simple conversations only. They can also authenticate users, check the order status, make payments, and so on. As a developer, Conversational AI gives you granular control over the responses that will be generated by the chatbot. You can choose to include or exclude generated responses.

Apps built with Vertex AI Search and Conversation include relevant and useful recommendations in the responses generated. For instance, if a customer searches for a mountain bike, a helmet, bike safety lights, or a floor pump can be included in the responses. This is important in product discovery, upselling, or cross-selling. In internal use cases, recommendations can help employees discover new information and insights that could help push the business forward.

Vertex AI

Vertex AI is Google Cloud’s end-to-end machine learning platform. You can build, evaluate, version, and deploy ML models to production environments on this platform. Vertex AI has two main generative AI capabilities included in it: the Model Garden and Generative AI Studio.

Model Garden

Model Garden contains numerous pre-trained multi-task foundation models that you can import into your project and finetune to suit a specific use case. This includes Google’s large language models (LLMs) like PaLM, LaMDA, Imagen, open-source, and third-party models from Google’s partners. These models come enterprise-ready and are suitable for tasks like text, image, code, and video generation, as well as code completion.

Apart from foundation models, the Model Garden contains task-specific pre-trained models for simple tasks like object classification and detection. Users will be able to add these capabilities to their applications via simple API calls without worrying about infrastructure provisioning for the AI models.

Generative AI Studio

Under the umbrella of Vertex AI, Google also offers a tool for rapid prototyping and testing of Generative AI models. After selecting a suitable model for your use case in the Model Garden, you can set up the input instructions given to the model to elicit the desired output. This is especially important when working with language models like PaLM. You can also improve the quality of the model’s output by finetuning it using your organization’s data.

GenAI models in the Model Garden are hosted by Google and other third-party providers and are accessible via API. Because of this, organizations that intend to combine these models with their data would be concerned about losing control over their data. This is a huge problem for businesses operating in regulated industries like healthcare and finance. Fortunately, Google Cloud provides a secure environment for your data by ensuring that it can only be accessed by you. As a result, Google Cloud increases your application of data privacy while making large GenAI models available via simple API calls. This makes Google Cloud’s GenAI tools suitable for several use cases.

Implementing GenAI on Google Cloud

Google Cloud has prepared GenAI for commercial use and enterprises can use them to generate more value from their businesses in many different ways:

Content creation in digital marketing

Businesses can use Imagen, a text-to-image foundation model from the Model Garden to generate suitable images for their digital marketing campaigns. You can finetune Imagen in Generative AI Studio using your organization’s data. This helps to integrate your branding elements into the images that the model will generate including: brand mascot (such as Colonel Sanders for KFC), color palette, typography – where the text will be included in the image, logo, etc. As businesses create content that has to be consistent with all their brands, they can improve brand identity and differentiate themselves from their competitors.

A virtual assistant for ecommerce using Conversational AI

Using Conversational AI in the Vertex AI Search and Conversation, e-commerce stores can create chatbots capable of holding conversations with customers. With the ability to replicate human-like interactions, these chatbots can act as virtual assistants, guiding customers through their shopping experience. They can make it easy for customers to find products that accurately match their requirements, creating personalized shopping experiences. Since they can handle additional tasks – like user authentication – they can provide round-the-clock customer support and boost overall customer satisfaction.

Creating new product designs

Enterprises dealing in fashion can use the diffusion models in Model Garden to generate new product design ideas. One such model is ControlNet – a foundation model that allows you to control image generation with text prompts. You can combine this model with data on your previous product designs, and use it to generate new design ideas.

Improving information discovery on websites and databases

Companies can simplify the information discovery on their digital assets (such as their website) by creating GenAI-powered search engines using Enterprise Search. They can set up the search engine in Generative AI Studio, and define the data sources, and their output. Furthermore, they can import this feature into their app using APIs. With this search engine, employees will improve their productivity by finding information not just by keywords, but using meaning as well (semantic search).

Summary

Google Cloud provides a variety of GenAI tools that lower the barrier for organizations looking to integrate GenAI into their operations. This includes pre-trained models that you can readily use in your projects or finetune them to meet your specific use case. With little to no code, you can create GenAI-powered applications and experiences thanks to GenAI on Google Cloud. This will help to improve your business’s productivity, allowing you to get more value from it. Although it is still in its early stages, GenAI on Google Cloud has shown great promise with the products that have been released so far. Moving forward we expect Google Cloud to pursue GDPR compliance to ensure companies in Europe can use these tools to improve productivity. In addition, Google Cloud is expected to improve the language coverage in Generative AI Studio to give non-English speakers a better experience when using it.

Google Cloud NEXT ‘23 has also just ended and in the event, Google Cloud announced a huge bunch of news on GenAI. This includes several new GenAI models added to Model Garden such as Llama 2 (from Meta), upgrades to existing foundation models such as PaLM2 and Imagen, new tools for tuning foundation models, and the official release of Vertex AI Search and Conversation. More on this in our next piece.

Intro: Generative AI on Google Cloud

What is GenAI – From Traditional AI to GenAI

How GenAI came into the picture

GenAI on Google Cloud

Vertex AI Search and Conversation (formerly Generative AI App Builder)

Enterprise Search

Conversational AI

Vertex AI

Model Garden

Generative AI Studio

Implementing GenAI on Google Cloud

Summary

Ressourcen

Further Reading

You may also like…

Press Release: happtiq Wins Google Cloud Sales Partner of the Year

Vertex AI – Model training, Features and more

What is T-Systems Sovereign Cloud?