The 2014 movie Interstellar left a lasting impression on my mind. There were several movies before and several after, but none intrigued me about AI as much as Interstellar did. There are 2 scenes in that movie which I love to this day.
1. The scene where the crew is in space and Cooper (Matthew McConaughey) adjusts the “honesty and humor quotients” of TARS/CASE/KIPP the AI Robot. A similar one later when Dr Mann (Matt Damon) says KIPP does not understand fear.
2. The scene when the crew is on the water planet and TARS/CASE/KIPP saves Dr Brand’s (Anne Hathaway) life by rescuing her from the tidal wave.
In my mind, the movie drove the point of how humans can benefit from AI and Generative AI technologies by using and leveraging these technologies to achieve things which require capabilities beyond the human physical and physiological limitations.
There are several AI and Generative AI offerings available for general use now. Before we look at what is available, let us first sort the definitions of AI and Generative AI. In simple terms, Artificial Intelligence (AI) is the ability of machines to recognize, process and act on the information. They may also be able to learn and incorporate the learnings in a continuous learning mode. Generative AI is a type of AI which is capable using all the information and learnings, to generate new information such as text, images, audio and code. Generative AI has emerged as a disruptive technology and has gained prominence in a very short span.
In this blog let us focus on the different Generative AI offerings from AWS and take a quick look at what each service can offer.
AWS Generative AI Ecosystem services
Developer Productivity
- Amazon CodeWhisperer:
- Uses comments and existing code to help build applications faster providing code suggestions in form of snippets and even full functions.
- Can scan and list the vulnerabilities in the code and suggest best practices.
- Works with most of the commonly used IDE’s and most of the commonly used programming languages.
- CodeWhisperer detects if a code is similar to any open-source training data. It can flag such instances with the link repository URL and project license information or optionally filter them out. All the code you generate is owned by you unless it called out by the service.
- As of today, its free for individual use.
Generative AI – Use case Enablers
- Amazon Bedrock:
- Simplifies the process of building the Generative AI applications using Foundational models (Pretrained model trained ready to be used).
- Helps privately customize the foundational models using your own data.
- Customize the foundational models using your own data, test them using Amazon SageMaker ML Experiments and deploy at scale using the familiar AWS services.
- Currently, Amazon Bedrock offers models such as –
- Amazon’s Titan family which comprises of two models at present – a text-generating model and an embedding model.
- 3rd party models such as –
- AI21 Lab’s Jurassic-2 family, which are multilingual and can generate text in German, French, Spanish, Italian, Portuguese and Dutch.
- Anthropic’s Claude, which can perform a range of conversational and text-processing tasks.
- Stability AI’s Stable Diffusion, which can generate images, art, logos and graphic designs.
- From a Data Security and Privacy standpoint, it is important to note that, the customer is always in control of their data and the data will not leave the VPC.
- As of this writing, Amazon Bedrock is announced to be to be released soon in preview; so, some of this information is subjected to change.
- Amazon SageMaker JumpStart:
- Helps to easily build, train and deploy machine learning models with a fully managed infrastructure.
- Helps deploy foundational models with a click.
- Access, label and process structured and unstructured data.
- Support many of the leading frameworks, toolkits and programming languages out of the box.
- Helps with training, fine tuning and deploying pre-trained models.
- Since these models are pre-trained, you can use them as is or customize them for specific tasks.
Generative AI – Hardware accelerators
- AWS Trainium:
- Purpose built machine learning accelerators for deep learning training.
- Optimized for training natural language processing, computer vision, and recommender models used in a broad set of applications, such as text summarization, code generation, question answering, image and video generation, recommendation, and fraud detection.
- AWS Inferentia
- Used for Model hosting and inference.
- ML inference accelerator designed for running high-performance foundational model inference at a lower cost.
Illustrative: high level architecture to show a simple use case where Bedrock responds to a user input using prompt data.
There is an enormous interest in customers across the industries to incorporate and leverage Generative AI capabilities to elevate their business functions. For example – App development customers are actively evaluating services such as Amazon CodeWhisperer to reduce their time to market and cut costs; customers dealing with high traffic call centers are looking to optimize agent workload by efficient call summarization; customers working with submitted document are looking to use intelligent search capability of semantic AI etc.
For customers who are new to Generative AI and wondering how they can benefit from Generative AI, and where to look for additional information; Infosys has launched Infosys Topaz. Infosys Topaz is an AI first set of services, solutions and platforms using Generative AI technologies. It can help accelerate business value, unlock efficiencies at scale and build connected ecosystems. Link to Infosys Topaz – https://www.infosys.com/newsroom/press-releases/2023/enterprises-generative-ai.html
I do not think we are still at a point today where we can implement Generative AI in our day to day lives, as imagined in the movie. However, looking at the progress so far, I think the day is not far where we have not only implemented Generative AI in our day to day lives but also depending on it for furthering the human potential.
Date: June 01, 2023
Reviewed by: Ashutosh Dubey, Amazon Web Services