Generative AI technology has become a battleground for various tech companies, especially language generation, which is a well-known technology. AWS released the Amazon BedRock large-scale language model in 2023, causing a stir, and recently released two updates, including the “new Llama language model” and “responsible AI,” allowing you to run generative AI applications safely and easily. Let’s follow along with the cloud computing blog to understand the new features of AWS on Amazon BedRock.
Table of Contents
Table of Contents
Amazon BedRock Introduces Meta’s New Llama 3 Language Model
AWS recently announced that Meta’s Llama 3 model is now officially available on Amazon Bedrock. The Llama 3 model is primarily designed to help users build, experiment, and responsibly expand their generative AI applications. The Llama 3 model series includes two sizes: 8B and 70B, which have been trained on over 15 trillion data tokens. This dataset is seven times larger than that of the second-generation Llama 2 language model, with a broader range of data, allowing for more precise training.
Users can now utilize two new Llama 3 models on Amazon Bedrock, each suited for different applications:
- The Llama 3 8B model is suitable for devices with limited computational resources, such as edge devices. Additionally, it excels in tasks like text summarization, text classification, sentiment analysis, and language translation.
- Llama 3 70B Model is designed for content creation, conversational AI, language understanding, research and development, and enterprise applications. It particularly shines in tasks such as text summarization and accuracy, text classification and fine-grained differences, sentiment analysis and reasoning, language modeling, dialogue systems, code generation, and command execution.
- Meta is currently training Llama 3 models with parameters exceeding 400B. These models will feature new capabilities, including multimodal support, multilingual support, and longer context analysis.
Novice AI users can use the Amazon Bedrock console to apply for the Llama 3 8B Instruct or Llama 3 70B Instruct models, facilitating easy experimentation and evaluation of top-tier foundational models. This update gives AI developers more powerful tools and flexibility for various industry applications.
Responsible AI – introducing Guardrail feature to enhance security.
In addition to launching new Llama language models, recent concerns about security in generative AI have intensified. Trend Micro’s research reveals that since the surge in generative AI popularity last year (2023), social engineering, fraud, and cybercrime have evolved significantly. Hackers are now leveraging AI to extract personal data from stolen information for resale to other criminal groups, targeting specific individuals.
To address these challenges, Amazon BedRock has introduced the “Guardrail” feature. It acts as a vital link between enterprise applications and large language models, automatically assessing and blocking content that falls into restricted categories. This enhancement not only bolsters the security of foundational language models but also ensures that over 85% of harmful content is effectively blocked in practical applications. It represents a significant step toward responsible AI adoption, safeguarding user privacy in applications utilizing generative AI capabilities.
The generative AI craze is surging, and AWS is launching new language model applications and responsible AI to capture the market. Want to know how your business can leverage AI to boost efficiency and enhance applications? Contact Nextlink now to learn about AWS generative AI for enterprises!
