The Next Big Thing In AI: Small Language Models For Enterprises

Right now the most popular way to build an LLM-based system to handle large amounts of information is called retrieval-augmented generation (RAG). These systems try to find documents relevant to a user’s query and then insert the most relevant documents into an LLM’s context window. Such tools have proved that they can successfully consume vast amounts of data from the internet and summarise the information effectively. But they’re also prone to providing inaccurate answers and they even hallucinate. Satya Nadella, CEO of Microsoft, highlighted the adoption of SLM by companies like AT&T, EY and Thomson Reuters.

• These models comprehend multiple Indian languages, providing features like conversation summarization and sentiment analysis.
• For vertical industries or specialized use, massive general purpose LLMs such as OpenAI’s GPT 4 or Meta AI’s LLaMA can be inaccurate and non-specific, even though they contain billions or trillions of parameters.
• In 2021 alone we saw $29.1B of capital being invested by Venture Capitalists and many of those are adding LLMs to their products.
• Though LLMs are mainly used for legitimate purposes, they can lead to the potential for misuse.

The security risks of implementing AI

After each dance, women would stay where they were while men would rotate to the next woman. The “women” are query vectors (describing what each token is “looking for”) and the “men” are key vectors (describing the characteristics each token has). As the key vectors rotate through a sequence of GPUs, they get multiplied by every query vector in turn. Before a GPU can start doing math, it must move data from slow shared memory (called high-bandwidth memory) to much faster memory inside a particular execution unit (called SRAM).

Enter small language models (SLMs)

The fact that data never leaves the device is a huge benefit for privacy. We already see such applications through ChatGPT, Github Copilot, and many others. Quanta Magazine moderates comments to facilitate an informed, substantive, civil conversation. Abusive, profane, self-promotional, misleading, incoherent or off-topic comments will be rejected. Moderators are staffed during regular business hours (New York time) and can only accept comments written in English.

Best Travel Insurance Companies

This sometimes works better than a conventional search engine, but today’s RAG systems leave a lot to be desired. They only produce good results if the system puts the most relevant documents into the LLM’s context. But the mechanism used to find those documents—often, searching in a vector database—is not very sophisticated. If the user asks a complicated or confusing question, there’s a good chance the RAG system will retrieve the wrong documents and the chatbot will return the wrong answer. Healthcare providers are increasingly utilizing the immense potential of Foundation Models to promote patient engagement and adherence.

• This is why human brains have the capacity for far more complex levels of intelligence.
• Turing proposed that a computer can be said to possess artificial intelligence (AI) if it can create human-like responses to questions.
• To optimize the training process for these small models, researchers use a few tricks.
• In the next three years, the company’s LLMs further increased in size with GPT-4 trained on 1.7 trillion parameters.
• Interestingly, even smaller models like Mixtral 8x7B and Llama 2 – 70B are showing promising results in certain areas, such as reasoning and multi-choice questions, where they outperform some of their larger counterparts.
• SLMs can perform well in dynamic environments where rapid response is key.

In the case of Chegg, which has developed 26 SLMs, the data is based on over 100 million pieces of learning content, created over the past 10 years. Training an LLM like GPT-4 can cost millions of dollars and take months. SLMs, in contrast, require significantly fewer resources, slashing training costs. Maheshwari notes that “SLMs cost just 1/10th of what LLMs require, offering a highly cost-effective solution for many enterprise applications.”

Why Meta’s large language model does not work for researchers

Large language models (LLMs), such as ChatGPT and others, are being integrated into health care … These LLMs are being used to promote patient engagement and adherence, providing personalized, human-centric care. The smaller size of SLMs limits their ability to store lots of factual knowledge. This can lead to incorrect responses, more biases and irrelevant content. These include using selected training data and targeted improvements after training. Additionally, smaller domain specific models trained on more data will eventually challenge the dominance of today’s leading LLMs, such as OpenAI’s GPT 4, Meta AI’s LLaMA 2, or Google’s PaLM 2.

Writer, for example, is a startup that offers a full-stack, genAI platform for enterprises; it can support business operations, products, sales, human resources operations, and marketing. The company offers a range of language models that cater to specific industries. The company’s smallest model has 128 million parameters, the largest — Palmyra-X — has 40 billion. But to really align the potential of the technology with learning outcomes, what we need are education-specific small language models (SLMs), Dan Rosensweig, Executive Chairman at student learning platform Chegg, tells TNW. A report from The Times of India detailed how NoBroker, a real estate platform, developed SLMs to improve customer service interactions. These models comprehend multiple Indian languages, providing features like conversation summarization and sentiment analysis.

In 2021 alone we saw $29.1B of capital being invested by Venture Capitalists and many of those are adding LLMs to their products. They are using AI technology to create personalized messages to help patients. An AI-powered chatbot from a Value-Based Care (VBC) provider that sends tailored advice based on a user’s personal preferences. A knowledge base system explains a potential diagnosis and educates patients on pathways. Those and other solutions will go a long way toward helping patients stay engaged with their health goals over time and remain adherent.