FAQ

Artificial intelligence is a very broad term that encompasses technologies and methodologies that enable machines to perform tasks that would normally require human intelligence. There is traditional AI like Machine Learning (statistical) and Generative AI (creative). Machine learning uses statistical methods to learn from data to make predictions or decisions without explicit programming. Generative AI creates new data, often resembling existing data it has been trained on.

Retrieval-Augmented Generation (RAG) is a technique that allows large language models (LLMs) to retrieve and incorporate data from external sources (i.e. databases, contracts, spreadsheets), with the aim to make the outputs generated more reliable and accurate. The benefits of this are we can get answers and content generated based off up-to-the-second and propriety information, as well as the ability to restrict the LLM to be grounded in only the information we want, when extremely precise or explainable outputs are required.

Agentic workflows can greatly help when AI systems are required to perform more complex tasks that require strategic planning, reasoning, iteration, and the ability to access outside tools and applications. We use powerful frameworks such as LangChain within Python or JavaScript to achieve this. For example, we would like to plan a surf trip in Europe on a budget for the upcoming weekend. This can be broken into four steps: Surf agent - engage with a specialist API to return European locations that have good surf conditions this coming weekend Budget agent - check for flight and hotel prices in these regions within our budget Itinerary agent – Using the LLMs knowledge of these regions plan a fun itinerary for the weekend Summary agent – Provide a well structured summary of the trip options with associated costs, to allow you to make final decision.