Exploring the Core Concepts of AI

"The best way to predict your future is to create it." - Abraham Lincoln

Generative AI is a rapidly evolving field, with new breakthroughs happening all the time. Researchers are constantly exploring new ways to create AI systems that can generate more complex and realistic data.

Within the public sector , we are exploring various use cases to improve business productivity and efficiency. However , the key thing is drive some fundamentals with our business stakeholders on the common terminology within the AI world.

What is Generative AI?

Generative AI involves creating models that can autonomously generate new data. This could be in the form of images, text, audio, or videos. The models are trained on a dataset and then use probability distributions to create new data points that closely resemble the training data. These generated samples are then evaluated using a scoring function, which informs the model on whether the generated output is acceptable or not.

? Prompts are the primary input provided to an LLM. In the simplest case, a prompt may only be the user-prompt such as natural language text that request GenAI to perform a specific task.

? Prompt engineering describes the process of adjusting LLM input to improve performance and accuracy. Prompt engineering makes AI applications more efficient and effective. Application developers typically encapsulate open-ended user input inside a prompt before passing it to the AI model.

? User-prompts are whatever you type into e.g. a chat box. They are generally in the everyday natural language you use, e.g. ‘Where can I buy polo necked T-shirts?’.

? Meta-prompts/System prompts are more about the structure and composition of the prompt itself rather than the specific content . They are higher-level instructions that help direct an LLM to respond in a specific way. They focus on how the elements of the prompt are organized to guide the AI in generating the desired output.

? Embedding is the process of transforming information such as words, or images into numerical values and relationships .Embeddings convert real-world objects into complex mathematical representations that capture inherent properties and relationships between real-world data. Embeddings are typically stored in vector databases .

? Retrieval augmentation generation(RAG) is the process of optimising the output of a large language model, so it references an authoritative knowledge base outside of its training data sources before generating a response.

RAG extends the already powerful capabilities of LLMs to specific domains or an organization's internal knowledge base, all without the need to retrain the model. It is a cost-effective approach to improving LLM output so it remains relevant, accurate, and useful in various contexts.

? Vector databases is a type of database that stores data as high-dimensional vectors, which are mathematical representations of features or attributes.?The vectors are usually generated by applying some kind of transformation or embedding function to the raw data, such as text, images, audio, video, and others.?

? Grounding is the ability to connect model output to verifiable sources of information. If you provide models with access to specific data sources, then grounding tethers their output to these data and reduces the chances of inventing content.This is particularly important in situations where accuracy and reliability are significant.

? Chat history is a collection of prompts and responses. It is limited to a session. Different models may allow different session sizes. For example, Bing search sessions allow up to 30 user-prompts. The chat history is the memory of LLMs. Outside of the chat history LLMs are ‘stateless’. That means the model itself does not store chat history. If you wanted to permanently add information to a model you would need to fine-tune an existing model (or train one from scratch).

? Parameter efficient tuning or adapter tuning is the process of optimising the performance of the AI model for a specific task or data set by adjusting configuration settings.

? Model fine-tuning is the process of limited re-training of a model on new data. It can be done to enforce a desired behaviour. It also allows us to add data sets to a model permanently. Typically, fine-tuning will adjust only some layers of the model’s neural network. Depending on the information or behaviour to be trained, fine-tuning may be more expensive and complicated than prompt engineering. Experience with model tuning in government is currently limited and we are looking to expand on this topic in a future iteration of this framework.

? Open-source models are publicly accessible, and their source code, architecture, and parameters are available for examination and modification by the broader community. E.g. PyTorch , Tensor flow,Keras , etc.

? Closed models are proprietary and not openly accessible to the public. The inner workings and details of these LLM models are kept confidential and are not shared openly.E.g. chatGPT

I will discuss with further AI topics in my next artcile.