Dear All,
How are you?
It is quite impossible nowadays to look at any technology news and not hear of Generative AI. Let us try to first understand the same from the very beginning and basics.
We will first understand the basic concepts and then will drift towards how SAP has incorporated the same in their business model.
What is “intelligence”? The ability to accomplish complex goals.
What is “artificial intelligence”? Intelligence exhibited by non-biological systems.
Now let us understand the Approaches to Artificial Intelligence
Now as we have a basic understanding of AI concepts let us talk about Generative AI.Below are some of the quick facts:
Now let us look at the basic definitions:
- Foundation models are neural networks trained on large volumes of data using self-supervised learning that can be applied to many tasks.
- Large language models (LLMs) are a subcategory of foundation models for text incl. computer code.
- Generative AI can create novel output in text, images, sound, or video based on simple user input (called “prompts”).
The below timeline shows the Evaluation of SAP products in terms of technology and innovation.
Generative AI is not all good and rosy, it does have its fair amount of limitations and hiccups.
The limitations of generative AI models
- Hallucination. Large language models can generate plausible-sounding yet false answers.
- Up-to-date and specific knowledge. The knowledge of a generative AI model is frozen in time from when it was trained. In contrast to world general knowledge, business information changes quickly.
- Inconsistent math abilities and limited notion of time. Although improving, large language models are not calculators. Other kinds of foundation models could conceivably complete forecasts and math.
But there are methods that will help to make it more reliable and efficient.
- Prompt Engineering – Provide more information to describe the task
- Retrieval Augmented Generation(RAG) – Extend to external domain knowledge by
retrieving and injecting information via embeddings (numeric vectors). - Orchestration Tools – Agents, functions, plug-ins, prompt and model chaining, memory.
Now let us try to understand as how SAP has incorporated Generative AI within its business models.
Extending SAP applications with generative AI
SAP has provided the entry point to Generative AI, with a BTP application SAP AI Launchpad which can be found in SAP Discovery Center.
SAP AI Launchpad is a multitenant software as a service (SaaS) application in the SAP Business Technology Platform. Customers and partners can use SAP AI Launchpad to manage AI use cases (scenarios) across multiple instances of AI runtimes (such as SAP AI Core). SAP AI Launchpad also provides generative AI capabilities via the Generative AI Hub.
The below diagram is quite self-explanatory and helps to understand SAP capability in Generative AI.
- At the very top layer, it has an SAP AI launchpad that provides various tools for Prompt engineering and management.
- The bottom layer showcases the Trust and control that SAP has incorporated with the foundation models so that it would be reliable for the business.
- SAP is introducing a vector engine within SAP HANA Cloud so that the it understands, extract or insert data in a natural language.
Now let us talk about the Generative AI business use cases
- Just Ask feature for SAP Analytics Cloud
- Document Information Extraction in SAP Transportation Management Application
- Joule – The copilot that truly understands your business
The below diagram shows some more use cases.
I hope this will spark more interest in the areas of AI and GenAI, I would like to mention that technology is ever-evolving. GenAI seems to have a long road ahead.
Thanks all for reading the article, please do leave feedback, if possible.
Happy Learning !!
