Generative AI: Exploring Key Concepts with AWS PartyRock
Learn about Generative AI, its features and capabilities, in fun, no-code playground presented by AWS PartyRock.
Published Feb 5, 2024
Last Modified Feb 19, 2024
Just not long ago, I could not answer that question. However, recent introduction of AWS BedRock and AWS PartyRock have cleared the convoluted realm of Generative Artificial Intelligence.
Recent evolution of technology has introduced a new type of applications based on Generative AI and Foundation Models (FM), which are large-scale, pre-trained deep-learning models adapted to a wide range of tasks. These models are capable of creating new data and insights from a large amount of existing data. Large Language Models (LLM) is a subset of FMs created for natural language understanding and processing. LLMs are capable of following natural language instructions, accessing external APIs for more data, and analyzing text from data streams.
According to AWS website, PartyRock is a fun and intuitive hands-on, shareable generative AI app-building tool in no-coding environment, yet flexible enough to build cool Generative AI-based applications.
When creating your application, AWS PartyRock will ask you to describe your app. The user input is served as a prompt for a Generative AI model to suggest basic features of the App.
Next, your application can be updated using five types of widgets:
- User Input - input textbox (prompt)
- Static text - any text that may serve as a description
- AI Powered Text Generation - text response generated by LLM model
- AI Powered Image Generation - image generated by FM model
- AI Powered Chatbot - conversational chatbot
For each AI powered widget you have the ability to select most suitable Foundation Model. There is an option to adjust Temperature and Top P parameters in Advanced settings. Temperature is a value between 0 and 1. Use lower temperature if you want more deterministic responses, and use higher temperature if you want more creative or different responses. Top P is based on probability of the potential choices. If you set Top-p below 1.0, the model considers the most probable options and ignores less probable options, leading to more stable result.
My first introduction to AWS Bedrock and AWS PartyRock was rather amaizing. I was a part of AWS re:invent 2023 when both services were introduced to the public. The most enjoyable part of the re:invent experience was my participation in AWS Gameday event that offered hands-on opportunity to experience PartyRock and Bedrock in action. This led to creation of Unicorn Space Suit Builder** app. AWS PartyRock assisted in understanding the main concepts of Generative AI applications as well as provided me with the first experience.
On the other hand, AWS Bedrock provides an opportunity to utilize generative AI capabilities to build powerful customized applications with "generative" functionality. AWS Bedrock unleashes the full power of generative AI by providing not only a great selection of Foundation Models but an ability to fineturn specific model to your data thus improving accuracy of generated output. Looking into the future, AWS Amplify, AWS Q, and AWS Bedrock, combined together, would allow us to develop powerful custom applications.
For instance, GoalTracker: Reach Your Goals application developed on AWS PartyRock can be improved by developing additional custom features with AWS Bedrock in combination with AWS Q and AWS Amplify services:
- Additional goal settings: goal description, target date, and key metrics to measure progress.
- Goal categorization into different areas, such as health, finance, career, personal growth, or education, to help users organize their objectives.
- Progress indicators and charts to visualize the user's advancement toward their goals with option to record achievements and milestones.
- Integration of a daily planner (calendar) to help users allocate time for working toward their goals.
Check out Unicorn Space Suit Builder App
Olga Perera is a PhD candidate (ABD) at DSU. She is proficient with machine learning algorithms, natural language processing, relational and NoSql databases, AWS GovCloud. Recipient of MS Applied Data Science Award from Syracuse University, recipient of Graduate Research Award from Dakota State University, recipient of Most Innovative Solution Award, Machine Learning Hackathon from NASA GRC.