aiLLM
Introduction
The AI Large Language Model facet, known as aiLLM, uses large language models to predict columns of a dataset or answer questions. It is most useful with text and descriptive data and can predict column output as text for given text input. In the field of AI this type of problem solving is known as question answering.
aiLLM performs a search on nearest neighboring data to find the most relevant pieces of information from your data and formulates human readable, text-based answers to questions. aiLLM returns information that is primarily based on the training data that you provide instead of the entirety of the dataset used to train the LLM.
Use Cases
aiLLM helps you perform a variety of natural language processing tasks. The section below describes some of the most common LLM use cases and maps each one to a corresponding Ikigai aiLLM submodel. If none of the use cases described in this section address your particular need, you may consider using aiLLM's custom submodel.
- Targeted responses - Answer questions based on specific documentation. For example, you can provide Ikigai documentation as training data and users can ask how they would use the platform to perform a specific task.
- Sentiment Analysis - Perform zero shot classification for text heavy datasets. This is a type of machine learning where a model can categorize a new item without ever having seen it before. For example, you can provide a dataset of film reviews and aiLLM determines if the review reflects a positive or negative sentiment.
- Predict Numerical Values - Predict numerical values based on input data. For example, predict housing prices based on different attributes, like house age and location.
- AI Chatbots - Simulate human conversation with an end user. Use aiLLM's chat submodel with a dashboard to provide instantaneous human-like response to user questions and requests.
- Flow Matching - Describe an action that you want to perform and receive a response from aiLLM with the best Ikigai flow to complete your task. Given flow names and corresponding descriptions, match a new description to the flow the user should run.
- Summarize Data - Input text to be summarized in a format and style that you designate. For example, provide a product description in your catalog and aiLLM can generate product summaries that are two sentences long.
- Investigate a Dataset - Gather information about a particular dataset. For example, given a dataset including salary information about a specific profession, ask questions related to salary and experience level.
Key Concepts
A large language model (LLM) is a type of AI system that can recognize and generate text. An LLM is trained on enormous datasets so that it can recognize and interpret human language or other types of complex data. To peform specific tasks or types of data processesing, you can train an LLM on a selected set of data, or text, so that it can predict and generate responses for a more discrete set of topics.
Using Ikigai's aiLLM model you can further train an LLM on a smaller set of data in order to perform a specific task based on a prompt. At a high-level this process consists of the following steps:
- identifying your use case and its corresponding aiLLM submodel,
- formatting a train and query dataset to provide as input, and
- determining your desired output and the submodel settings required to generate your output.
Refer to Ikigai aiLLM Documentation for more information.
