Generative Pillars, the Toolbox for Value-Driven AI

Authored by Andrea Cappelletti, Business Unit Director, Digital Transformation & Hyperautomation at S2E

What is Generative Artificial Intelligence

Generative AI typically relies on foundation models, which are AI models trained on vast amounts of data. These models can learn the distribution of the training data, study its probability distribution, and generate a content that replicates the most probable characteristics of the training data. Foundation models can be used via conversational interfaces or through APIs (Application Programming Interfaces) invoked by developers in the code they are developing.

SOURCE: Osservatori.net

Generative Pillars - the Appoach

In order to fully exploit the extraordinary opportunities offered by generative AI and extract maximum value from it, an innovative approach is needed. The relentless evolution of tools and technologies must be considered in this approach. Generative Pillars are not specific solutions tied to technologies, but rather a platform that incorporates the best available technologies over time. This approach requires a high level of expertise and experience, which S2E has implemented and continues to implement in this area for its clients.

Generative Pillars

The Generative Pillars framework is a set of skills and tools, a veritable toolkit, that professionals can use to build tailored conversational agents based on different models, data, and use cases. The pillars are modular and composable to address use cases effectively, with methods that are continuously and rapidly evolving. The ability lies in being able to pivot in real-time to the best technology.

S2E responds to this complexity by immediately transferring with Execution capability - greatly appreciated by clients - the experience it gradually realized, without the grand promises that characterize so many players in the field at present. Flexibility and evolution of a "mobile" platform have been favored over a "consolidated" but static product incapable of adapting to continually changing technologies. The ideal companion for Generative Pillars is GenerativeShield, S2E's SaaS solution that mitigates the risks and complexities generated by the use of conversational platforms based on GenAI.

Anthropic Claude3 and Scope of Application

S2E develops generative AI solutions with Anthropic Claude on AWS Bedrock, based on Claude3's foundation models, a multimodal model able to handle text and images. Benchmarks show that it is even faster, more performant, and more cost-competitive than GPT4 and Gemini.

• Analyze documents or reports containing graphs or images in detail, leveraging Claude Opus's capacity.
• Develop a prompt for data extraction of all data present in the graphs, thanks to Claude3's multimodal approach.
• Compare extracted data to provide precise and relevant answers based on reasoning ability, creating a CoT prompt.

Conclusion

With deep knowledge of artificial intelligence, S2E is able to provide tailored solutions for any company looking to accelerate and optimize its generative AI strategy in a scalable, secure, and efficient manner. ChatGPT can make mistakes. Check important information.