Why LLMs could be a big deal for HCI

Large Language Models are machine learning models that have been trained on a very large volume of textual data that represent sequences of words in different languages. The goal is to predict the next most probably token (word) given a sequence of tokens (words). With this powerful fundamental attribute, they can be adopted in many use cases including language translation, text summarisation, text classification among other.

A use case which I find particularly interesting is the classification of human-written sentences into actionable intents which I believe can enable humans to communicate with computers in a more natural way.

Computers or computerised systems require a well defined series of steps and instructions to be able to effectively carry out operations. These steps may demand some degrees of technical sophistication from even everyday users which can also negatively affect the experience they have with the system. Humans are notorious for expressing simple ideas in the most complicated ways and language is the most common method of expression. Perhaps it makes sense to have a way for humans to interact with software using natual languages they are already familiar with.

With LLMs, we can extracts intents from a sentence and convert those intents into operations that can be executed by the system as long as they are supported. Here is an example.

    "Send $20 to Henry"
    >>  sendMoney("Henry", 20);

    "Henry asked me for $100 but I only have 20, so just send everything to him"
    >> sendMoney("Henry", 20);

    "Share $40 equally among Henry and Tope please, I am feeling quite generous today😊"
    >> sendMoney("Henry", 20); sendMoney("Tope", 20);

In each instance, a capable LLM (Bloom or GPT3) is able to distill off the noise, extract and infer the intents of the human user and convert them to a predefined internal instruction. This of course will require some coding and clever prompting which I hope to demonstrate in the coming posts.

The significance is that users can be more expressive when using a product that implements this approach which is also a huge boost to the user experience – it’s all about the user afterall.