AI language models: The case for cultivating context

Language models work with patterns in text, but business conversations carry more than words. A deadline, a promise and a tentative suggestion can look similar on the page while requiring very different actions.

That makes context central to useful AI. The model needs the relevant information, and the application needs a way to check what the model produces.

Artificial intelligence (AI) is one recent effort to imbue programs with human-like qualities. It’s a long road to artificial general intelligence, or general AI, which could mimic humans’ general cognitive abilities. However, we are making inroads with specific AI use cases that could change how we interact with the world and with each other.

Large language models, or LLMs, are one class of AI language model. This article explains how they use context, where they help and what changes when a model becomes part of a tool that performs work.

AI Language Models: From Primates to Programs

Early chatbots often relied on rules and matching words or phrases. Modern language models can handle more flexible requests, although a fluent response still needs to be checked against the facts.

Humans, however, go beyond these basics thanks to a key concept: context.

Consider a conversation between two friends. The first asks how the second is doing, and the second replies “fine”. The word is easy to parse: Merriam-Webster defines it as “well or healthy; not sick or injured.” Despite the obvious meaning, however, the first friend replies “I’m here if you need to talk.”

The friend used more than the dictionary meaning of the word. Tone, facial expression and the history of the conversation helped shape the response. An AI application can use only the context made available to it, and that context may be incomplete.

It is also why language analysis technologies have historically struggled to move beyond basic questions-and-answers. However, AI language models have made remarkable advances in recent years.

Can AI Language Models Break Barriers?

Along with context, human language is also unique because it’s compositional. This property allows humans to express their thoughts in sentences that recognize multiple tenses and use a combination of subjects, objects, and verbs. With just 25 subjects, 25 objects, and 25 verbs, it’s possible to create more than 15,000 sentences.

The compositional nature of human language also makes it possible to create strings of seemingly nonsense words that actually make sense in context. Consider the (in)famous buffalo sentence below:

Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.

While this looks like gibberish, it’s grammatically correct. It lacks punctuation and articles such as “which” or “that”, but it is possible to understand. Here’s how.

The capitalized instances of Buffalo refer to the city.

The lowercase instances use buffalo as an animal and as a verb meaning to intimidate or confuse. The sentence can be read as:

Buffalo from the city of Buffalo that are intimidated by other Buffalo buffalo also intimidate Buffalo buffalo.

Sure, it’s not an easy read, but with the ability to parse compositional language, it does make sense. While most human sentences aren’t quite so convoluted, we naturally assume that listeners have the ability to combine the obvious meaning of our words with both contextual and compositional clues.

For a business application, ambiguity is often more ordinary: an unclear pronoun, an outdated price or a promise recorded in one system but missing from another.

Clear instructions and the right source material matter as much as fluent wording.

The Basics of LLMs

AI language models aren’t new. As artificial intelligence moved from potential to possible, researchers quickly realized the potential of AI to analyze and understand language. Historically, however, language models were limited in scope and size.

Earlier language models used approaches such as statistical word sequences and recurrent neural networks. Their limitations were not simply a matter of assigning equal weight to every word.

A chatbot is an interface or application. A language model is one possible component behind it. Some chatbots use fixed rules, while others use LLMs together with retrieval and business tools.

Many modern LLMs use transformer architectures. They process text as tokens and learn patterns during training. The model then uses the context supplied in a request to generate an output.

Two important transformer ideas are positional information and attention. These help the model represent token order and relationships within the context it can access.

Positional information tells the model about token order. Attention lets it weigh relationships between tokens. This is not access to every previous conversation or unlimited memory.

These mechanisms support useful language processing, but they do not guarantee factual accuracy. The application still needs relevant inputs and a way to evaluate the result.

Artificial Intelligence, Real Impact: Large Language Models in Action

It’s one thing to design a language model capable of improving over time to better understand humans and provide relevant responses, but what does this look like in practice?

Here are five examples of language models in use.

Search Results

Search engines such as Google and Bing are already using LLMs to improve search results. By applying large language models to understand the context of user queries, search engines can provide more accurate, relevant hits.

Content Generation

A text assistant can draft email subject lines, outline an article or revise a document from a brief. Image generation is a related generative AI capability, but an image model is not simply another name for a language model.

Customer Service

A support application can combine an LLM with approved help content and tools for retrieving account information. Its usefulness depends on answer quality, reliable sources and a clear path to a person when the request exceeds its scope.

For example, a support assistant might summarize the customer’s issue, retrieve a relevant help article and prepare an escalation. The application must be configured to do those jobs, and its responses should be checked against the underlying account and policy information.

Language Translation

LLMs can assist with translation, including adapting wording to a context. They can still miss nuance or use the wrong terminology. Have a qualified speaker review consequential or customer-facing translations.

Document Analysis

For industries such as finance, law, and healthcare, LLMs can be used to identify key themes, pinpoint repeated themes or ideas, or discover issues with reporting and documentation that must be addressed before reports are filed or businesses take action.

What’s Next?

A model does not automatically become more accurate every time it is used. Training, retrieval, memory and conversation context are different mechanisms, with different limits.

Some applications can search external sources or retrieve internal documents before answering. That gives the model additional context at the time of the request, without necessarily retraining the model itself.

Retrieval can make an answer better grounded, but it cannot guarantee correctness. Check that the source is relevant, current and accurately represented in the response.

An AI agent adds a way to take actions through tools. An AI employee organizes that capability around an ongoing responsibility. Evaluate the complete assignment, including tool access, follow-through and exceptions, rather than judging only the quality of a chat response.

A broader example is Artie, an AI employee designed to take responsibility for ongoing work across apps and company context. A team might evaluate a recurring CRM review rather than ask for an isolated summary. The test should include the source records, the finished review and how unresolved questions reach the right person.

Artisan’s Ava applies AI to outbound sales work such as prospect research and outreach. That specialized workflow is one way to turn a model’s language capabilities into a business process.

It’s also worth noting that as AI language models evolve, so too will the role of human workers. While there’s a growing fear of AI replacing human staff, the far more likely scenario is that roles will simply shift. While AI might handle the heavy lifting of drafting a marketing email or creating a web design, humans in the loop remain essential to review and revise ideas before they go live.

The Bottom Line: AI Language Models and Language 2.0

Language models are useful when they have the right context and their output can be checked. For business work, the stronger evaluation follows the whole process: what information went in, what the system did and whether the finished result meets the team’s standard.