Two years from the public release of ChatGPT, conversations about AI are inevitable as companies in every industry look to use. major language models (LLMs) to change their business processes. However, as powerful and promising as LLMs are, many business and IT leaders rely too much on them and forget their limitations. This is why I look forward to a future where specialized language models, or SLMs, play a larger, complementary role in business IT.

SLMs are more commonly referred to as “small language models” because they require less data and time to train and “more streamlined versions of LLMs.” But I prefer the word “specialized” because it better conveys the ability of these purpose-built solutions to do more specialized work with more precision, consistency and transparency than LLMs. By adding LLMs to SLMs, organizations can create solutions that take advantage of the strengths of each model.

Trust and the ‘black box’ problem in LLM

LLMs are incredibly powerful, but they’re also known to sometimes “go off the rails,” or offer outputs that veer off course because of their generalist training and large data sets. That tendency is made more problematic by the fact that OpenAI’s ChatGPT and other LLMs are essentially “black boxes” that don’t reveal how they get the answer.

This black box problem will become a bigger issue going forward, especially for companies and business-critical applications where accuracy, consistency and compliance are paramount. Think of health care, financial services and legal as prime examples of professions where inaccurate answers can have huge financial consequences and even life-or-death effects. Regulatory bodies have taken notice and are likely to start making demands explainable AI solutionsespecially in industries that rely on data privacy and accuracy.

While businesses often put in place 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 occur and mistakes can disappear without being noticed.

SLMs = greater clarity

Fortunately, SLMs are better suited to address many of the limitations of LLMs. Rather than being designed for general-purpose tasks, SLMs are developed with a narrower focus and trained on domain-specific data. This specification allows them to handle nuanced language requirements in areas where accuracy is most important. Instead of relying on vast, heterogeneous datasets, SLMs are trained on targeted information, giving them the contextual intelligence to provide more consistent, predictable and relevant responses.

It offers many advantages. First, they are more explainable, making it easier to understand the source and rationale behind their outputs. This is critical in regulated industries where decisions must be traced back to a single source.

Second, their small size means that they can often perform faster than LLMs, which can be an important factor for real-time applications. Third, SLMs offer enterprises more control over data privacy and security, especially if they are deployed internally or built for the enterprise.

Additionally, while SLMs may initially require specialized training, they reduce the risks associated with using third-party LLMs controlled by external providers. This control is essential in applications that demand strict data management and compliance.

Focus on skill development (and beware of salespeople who overpromise)

I want to make that clear LLMs and SLMs not mutually exclusive. In practice, SLMs can complement LLMs, creating hybrid solutions where LLMs provide a broader context and SLMs ensure accurate implementation. It’s still early days though where LLMs are concerned, so I always advise technology leaders to continue exploring the many possibilities and benefits of LLMs.

Additionally, while LLMs can scale well for a variety of problems, SLMs may not transfer well to some use cases. That’s why it’s important to have a clear understanding up front of what use cases to solve.

It is also important that business and IT leaders spend more time and attention on building the distinct skills required for training, maintenance and testing of SLMs. Fortunately, there is a lot of free information and training available through common sources such as Coursera, YouTube and Huggingface.co. Leaders must ensure that their developers have enough time for learning and experimenting with SLMs as the battle for AI expertise intensifies.

I also advise leaders to carefully examine colleagues. I recently spoke with a company that asked for my opinion on the claims of a technology provider. My guess is that they are either overstating their claims or simply out of their depth in terms of understanding the capabilities of the technology.

The company wisely went back and implemented a controlled proof-of-concept to test the vendor’s claims. As I suspected, the solution was not ready for prime time, and the company was able to walk away with little time and money invested.

Whether a company is starting a proof-of-concept or a live deployment, I advise them to start small, test regularly and build on early successes. I personally experienced working with a small set of instructions and information, only to find the results to turn the road when I then fed the model more information. So slow and steady is a prudent approach.

In summary, while LLMs will continue to provide valuable capabilities, their limitations are becoming more apparent as businesses increase their reliance on AI. The addition of SLMs offers a way forward, especially in high-stakes fields that demand precision and ease of explanation. 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.