Despite its potential, the article questions who Arctic LLM is really for, given the abundance of other "open" generative models available. It also points out that Arctic LLM's relatively small context window could be a disadvantage, as models with small context windows are prone to forgetting the content of recent conversations. Furthermore, like all generative AI models, Arctic LLM is a statistical probability machine and may occasionally provide incorrect answers, a phenomenon known as "hallucinations".
Key takeaways:
- Snowflake, the cloud computing company, has unveiled Arctic LLM, an enterprise-grade generative AI model optimized for enterprise workloads, including generating database code.
- Arctic LLM is part of a family of generative AI models called Arctic and is claimed to outperform Databricks’ DBRX and Meta’s Llama 2 70B in tasks of coding and SQL generation.
- Arctic LLM uses a mixture of experts (MoE) architecture, which breaks down data processing tasks into subtasks and delegates them to smaller, specialized models. This efficient design enabled it to train on open public web data sets at roughly one-eighth the cost of similar models.
- Snowflake plans to make Arctic LLM available across a range of hosts, including Hugging Face, Microsoft Azure, Together AI’s model-hosting service, and enterprise generative AI platform Lamini, but it will be available first on Cortex, Snowflake’s platform for building AI- and machine learning-powered apps and services.