Security

AVA is built on Azure OpenAI, a service provided by Microsoft that offers access to a vast library of advanced language models. These models can be utilized through Azure's robust and scalable cloud platform. As AVA is built on Azure OpenAI, your data—used for input or output—remains secure within your company. This applies to data from:

• Prompting
• Office 365 files
• Your own personal files

For more information on data security, read here:

https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy

<aside> 💡 Maybe add some comparisons for each model and explain which one to use for different scenarios

</aside>

Models

AVA’s main service is OpenAI but, it also offers a diverse range of AI models to choose from. It can leverage models from these services:

• Azure OpenAI
• Anthropic Claude API
• Google API
• Meta Llama 2/3
• And others, based on specific customer requirements!

Model Comparison

Azure OpenAI

• Seamless integration with Microsoft's suite of products like Azure, Office 365, and Dynamics 365.
• Azure's robust infrastructure allows for easy scaling of AI models.
• High emphasis on security and compliance, including GDPR and HIPAA.
• Extensive support and detailed documentation available.

OpenAI Use-Case:

• Data Analytics and Reporting
  • Why: Advanced analytics capabilities and integration using Microsoft products

Anthropic Claude

• Strong emphasis on ethical AI and safety, reducing risks of harmful outputs.
• Uses advanced techniques to improve model alignment with human values.
• Open about research and methodologies, promoting trust and collaboration.

Claude Use-Case:

• Ethical AI for healthcare
  • Why: Strong emphasis on ethical AI and safety, reducing risks of harmful or biased outputs in sensitive applications like healthcare.

Google Gemini

• Backed by Google’s extensive research in AI, resulting in cutting-edge models.
• Seamless integration with Google Cloud services and other Google products.
• Robust infrastructure allows for easy scaling.
• Generally user-friendly with extensive documentation and support.

Gemini Use-Case:

• Natural Language Processing for Search Engines
  • Google's expertise in NLP and seamless integration with Google Cloud services make it a strong choice for search engine enhancements.

Meta Llama

• Being an open-source model, it allows for extensive customization and transparency.
• Strong community support and collaboration.
• Generally, more cost-effective due to its open-source nature.
• High flexibility for research and development purposes.

Llama Use-Case:

• Academic Research
  • Being open source, it offers extensive customization and flexibility, making it ideal for research and experimentation.

Pros and Cons

• Integration and Ecosystem:
  • Azure OpenAI and Google excel in seamless integration with their respective ecosystems.
  • Meta Llama offers flexibility due to its open-source nature but lacks formal integration.
  • Anthropic Claude might have limited integration options compared to the others.
• Ethics and Safety:
  • Anthropic Claude stands out with its strong focus on ethical AI.
  • Azure OpenAI and Google also emphasize security and compliance but may not have the same level of focus on ethical AI.
  • Meta Llama relies on community-driven ethics and safety measures.
• Cost:
  • Meta Llama is generally more cost-effective due to its open-source nature.
  • Azure OpenAI and Google can be expensive, especially for extensive usage.
  • Anthropic Claude may also be costly due to its specialized nature.
• Scalability:
  • Azure OpenAI and Google offer robust scalability due to their extensive infrastructure.
  • Meta Llama and Anthropic Claude might face more challenges in scaling, depending on the deployment environment.
• Support and Documentation:
  • Azure OpenAI and Google provide extensive support and documentation.
  • Meta Llama relies on community support, which might not be as comprehensive.
  • Anthropic Claude offers transparency but may have limited formal support.

Home

Table of Contents