Overcoming a Big Language: How SLM Can Beat Its Bigger, Resource-Intensive Cousins

Two years on from ChatGPT's public launch, conversations about AI are inevitable as companies across all industries try to make use of it. large language models (LLMs) to transform their business processes. However, as powerful and promising as LLMs are, many business and IT leaders have come to over-rely on them and overlook their limitations. That's why I foresee a future where special language models, or SLMs, will play a larger role in enterprise IT.

SLMs are often referred to as “small language models” because they require less data and training time and are “simplified versions of LLMs.” But I prefer the word “specialized” because it better conveys the ability of purpose-built solutions to do highly specialized work with greater accuracy, consistency, and transparency than LLMs. By supplementing LLMs with SLM, organizations can create solutions that take advantage of the strengths of each model.

Trust and the LLM 'black box' problem

LLMs are extremely powerful, but they are also notorious for “losing the plot,” or offering results that seem off course due to their general training and large data sets. That tendency is made more difficult by the fact that ChatGPT at OpenAI and other LLMs are “black boxes” that do not reveal how to arrive at an answer.

This black box problem is going to be a bigger issue in the future, especially for companies and business-critical applications where accuracy, consistency and compliance are of utmost importance. Think of healthcare, financial services and legal as prime examples of professions where incorrect answers can have serious financial consequences and even life or death consequences. Regulatory bodies are already taking note and are likely to start asking clear AI solutionsespecially in industries that rely on privacy and data accuracy.

Although businesses often use a “human-in-the-loop” approach to mitigate these issues, over-reliance on LLMs can lead to a false sense of security. Over time, complacency can set in and mistakes can slip through the cracks.

SLM = more definition

Fortunately, SLMs are better suited to address many of the limitations of LLMs. Rather than being designed for general tasks, SLMs are developed with a narrower focus and trained on domain-specific data. This specialty allows them to handle advanced language needs in areas where accuracy is critical. Instead of relying on extensive, heterogeneous databases, SLMs are trained on targeted information, giving them the contextual understanding to deliver more consistent, consistent and relevant responses.

This offers several advantages. First, they are more transparent, making it easier to understand the origin and rationale behind the product. This is essential in regulated industries where decisions must be traced back to source.

Second, their smaller size means they can often perform faster than LLMs, which can be a critical feature for real-time applications. Third, SLMs offer businesses more control over data privacy and security, especially if they are used internally or built specifically for the enterprise.

Additionally, while SLMs may require specialized training to begin with, they reduce the risks associated with using third-party LLMs controlled by outside providers. This control is invaluable in applications that require strict data handling and compliance.

Focus on developing knowledge (and beware of overbearing salespeople)

I want that to be clear LLM and SLM they are not mutually exclusive. In practice, SLMs can extend LLMs, creating hybrid solutions where LLMs provide broader context and SLMs ensure detailed execution. It's still early days even as far as LLMs are concerned, so I always advise technology leaders to continue to explore the many possibilities and benefits of LLMs.

Additionally, while LLMs can scale well for a variety of problems, SLMs may not scale well to specific use cases. It is therefore important to have a clear understanding in advance of the use cases to address.

It is also important that business and IT leaders devote more time and attention to building the specific skills necessary to train, define and test SLMs. Fortunately, there is a lot of free information and training available through common sources such as Coursera, YouTube and Huggingface.co. Leaders should ensure their developers have enough time to learn and experiment with SLMs as the battle for AI expertise intensifies.

I also advise leaders to carefully vet partners. I recently spoke with a company that asked for my opinion on the claims of a particular technology provider. My opinion was that they were either overstating their claims or simply out of their depth in terms of understanding the capabilities of the technology.

The company wisely took a step back and implemented a controlled proof of concept to verify the vendor's claims. As I suspected, the solution wasn't ready for prime time, and the company was able to walk away with very little time and money invested.

Whether a company starts with a proof of concept or a live deployment, I advise them to start small, test often and build on early successes. I have personally experienced working with a small set of instructions and information, only to find the results looking off course when I then feed more information to the model. That's why slow and steady is a sensible approach.

In summary, while LLMs continue to provide increasingly valuable capabilities, their limitations are becoming increasingly apparent as businesses increase their reliance on AI. Adding SLMs is a way forward, especially in high-stakes areas that demand precision and definition. By investing in SLMs, companies can future-proof their AI strategies, ensuring that their tools not only drive innovation but also meet the demands of trust, reliability and control.

AJ Sunder is the co-founder, CIO and CPO of Responsive.