Everyone is talking about Large Language Models (LLMs) these days. It seems like a new LLM comes on the market every week, so the landscape is constantly evolving.
The real hype around LLMs began with OpenAI’s revolutionary ChatGPT tool, and only intensified with the release of GPT-4. This trend really started with Meta’s LLama model. This development laid the foundation for numerous other models, often named after animals, such as Stanford University’s Alpaca, Databricks’ Dolly, TII’s Falcon and Microsoft’s Orca. Those who want to stay up-to-date on these lightning-fast technological advances should check out Huggingface.co’s open source leaderboard(https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard). Here all open source models are weighed on various indicators.
 | Discover amazing ML apps made by the community huggingface.co |
But every hype also has a downside, namely the dreaded phase in the hypecycle, also called the Through of Disillusionment. So, where are we in the hype cycle of LLMs? Do we see LLMs as a temporary fad that has already peaked, or are we instead in the “slope of enlightenment,” where AI is consolidating and becoming an integral part of our daily lives?
At Isatis, we strongly believe in the latter scenario. Variants of ChatGPT focused on coding and copywriting are used daily by professionals. This shows the great value of LLMs for personal professional use, but where is the value at the corporate level?
ChatGPT draws on general Internet knowledge and can sometimes “hallucinate. This means no answers based on actual company documents and the possibility of misinformation that appears very believable. This poses a risk to companies, which need reliable and tailored information.
This is where LangChain comes in. LangChain allows companies to integrate LLMs with their corporate data, whether open source or API-powered LLMs such as GPT-4.
Is your company looking to integrate LLMs with reliable, customized information? We’d love to think with you. Contact us today for a no-obligation consultation!
LangChain offers many possibilities. It can be deployed to create a chatbot to “talk” to your internal PDF files. You can use local open source models to protect privacy and competitive data. In addition, you can use the most powerful models, such as GPT-4, via an API for more complex tasks. Moreover, LangChain allows you to use vector database integrations, which act as both a short- and long-term memory, allowing you to contextually query on a long-term basis.
At Isatis, we recognize the power of LLMs and their potential to add value to both individuals and businesses. Therefore, we have organized a workshops where we let our data scientists learn and explore this new technology. These sessions have been very successful and given the versatility of LLMs, we plan to organize them more often.
Whether you are a data scientist eager to get started with the latest technologies, or a company looking to harness the power of LLMs, we invite you to contact us. We look forward to exploring together what this amazing technology can do for you or your company.
