Generative AI

Generative AI (GenAI) suddenly got a lot of attention after Chat-GPT was made available to the public. It was mesmerizing to see it generate human-like outputs across various domains without explicitly providing detailed instructions to the system. GenAI creates new content and ideas such as images, text, videos, music, stories, and so forth, all of which are powered by large models that are pre-trained on vast amounts of data. These pre-trained models are called foundational models (FMs).

GenAI’s capabilities and possibilities have excited businesses too in terms of how they can use this technology to help their customers and make their experiences even better.

Some of the use cases that GeAI can solve are as follows:

• Text: Content writing, chat, taking notes, and sales and support
• Code: Code generation and documentation, text to SQL
• Image and video: Image/video generation, media and ads, social media, and design
• Speech: Voice synthesis...

How does generative AI help different industries?

Since we will not be able to dive deep into every use case across many industries, the least we can do is highlight how GenAI can disrupt the conventional ways of solving use cases across many sectors. Every organization wants to ensure that it can transform its business outcomes by incorporating GenAI into its operations. Every industry has many low-hanging use cases where GenAI can accelerate its business outcomes.

Financial services

Since our book revolves around GreatFin, a financial conglomerate, let’s start with how GenAI helps solve several important but tedious use cases within the financial services industry. Here are some examples:

• Fraud detection: GenAI can play a crucial role in identifying and preventing fraudulent activities in financial transactions. By analyzing extensive datasets, generative models can detect patterns and anomalies, enabling the identification of suspicious transactions and the...

Fundamentals of generative AI

The fundamental of GenAI always revolves around FMs. These FMs are pre-trained on vast amounts of unstructured data and contain a large number of parameters, sometimes in the billions, which makes the FMs capable of learning new complex concepts. FMs that are used for natural language processing, such as the ones from OpenAI’s GPT-3 and GPT-4, which are used in Chat-GPT, are pre-trained on a diverse range of internet text, enabling them to learn patterns, grammar, and general knowledge from vast amounts. These FMs are also called large language models (LLMs).

FMs differ from other ML models in several ways:

• Scale: FMs are trained on massive amounts of data, often involving billions of parameters. This large scale allows them to capture complex patterns and relationships in the data.
• Pre-training and fine-tuning: FMs undergo a two-step training process. First, they are pre-trained on a large corpus of publicly available text from the...

Generative AI on AWS

Ever since the art of the possibility, using GenAI has become obvious, and almost all cloud service providers and software vendors have shown a sense of urgency in providing new services/tools to help organizations build GenAI-based applications for their use cases. AWS also provides a few services that directly assist with this. Keep in mind that new services, along with new features in existing services, will continue to roll out going forward, so keep an eye on new ways of solving business use cases in the future.

To unlock the potential of GenAI, AWS focuses on a few considerations that organizations care for. Let’s look at them and introduce AWS services that support these considerations.

Firstly, building ML models is never trivial; in our predictive analytics chapter, we discussed the many stages that need to be addressed before and after training an ML model. Building and using FMs at scale needs a lot of work. AWS recently announced a new...

Analytics use case with GenAI

It’s just a matter of time before GenAI makes its way into most of the services used to build a modern data platform. In this section, we will just provide a glimpse of how GenAI can make analytics on AWS even easier. As always, here is a use case from GreatFin.

Use case for GenAI for data analytics

GreatFin has built a modern data platform on AWS and uses multiple purpose-built stores such as Amazon Redshift as a data warehouse, Amazon RDS as a transactional database, and a data lake on S3. To get data from these systems, complex SQL queries need to be written. Recently, there has been a steady request from non-technical users to provide them with a mechanism by which they can just converse in natural language and get the results from these data systems.

GreatFin has been asked to invest in mechanisms that allow anyone in the organization to request data using plain English questions. Under the covers, this is translated into the necessary...

Summary

In this chapter, we explored the latest trend around GenAI and how it is changing the ways businesses think about solving their use cases. We went through a range of possible use cases that each industry can solve using GenAI. We also looked at how FMs and LLMs are the core drivers for achieving GenAI outcomes.

We then pivoted toward how AWS helps organizations use GenAI for their use cases. Amazon Bedrock is a service that simplifies building and deploying GenAI applications using FMs in AWS. Purpose-built accelerators, such as AWS Trainum, are used for cost-effectively training LLMs, and AWS Inferentia and Inferentia2 are used to achieve best-in-class price performance to draw inferences from the FM models. Dedicated GenAI applications such as Amazon CodeWhisperer can help boost the productivity of the development team by auto-generating code. AWS also provides the flexibility to search for and choose from many of the FMs available in the market by using SageMaker JumpStart...

References

• Building GenAI applications workshop with SageMaker: https://catalog.workshops.aws/building-gen-ai-apps-with-found-models/en-US
• GenAI workshop on SageMaker: https://catalog.us-east-1.prod.workshops.aws/workshops/972fd252-36e5-4eed-8608-743e84957f8e/en-US