The next generation of AI for enterprise is here — don't take it for Granite

Welcome to The Short, IBM Research's recap of the latest innovations in AI, quantum computing, semiconductors, and the cloud. This week, we’re diving into the news out of TechXchange about IBM’s new generation of Granite models.  

Week of October 15 - 21, 2024

In this week's edition:

• Granite 3.0 
• Safer interactions with AI models 
• New use cases for cutting-edge AI  

The third generation of Granite is here 

IBM unveiled its new Granite 2B and 8B language models, designed to be the workhorse models for enterprise AI. These models were trained on 12 trillion tokens in 12 languages, from a combination of public and enterprise data sources. They’re well-suited for typical AI tasks like RAG, classification, summarization, entity extraction, and tool use. But when you use InstructLab to tune them, they can perform as well as much bigger models on many tasks that are important to businesses — at a fraction of the cost.   

🪨 Read the blog here

Safer interactions with Granite Guardian 

IBM also showed off its new Granite Guardian models, which let enterprise app developers implement safety guardrails. They check user prompts and LLM responses for several different types of risks, including social bias, hate, toxicity, profanity, violence, and jailbreaking. And when the models are applied to RAG use cases, they can check for groundedness, context relevance, and answer relevance. These 2B and 8B models offer the most comprehensive set of risk and harm detection capabilities available today.  

🌐 Read the blog here

Making coding simpler, and more models to try out  

Beyond just new models, IBM also unveiled the next generation of its watsonx Code Assistant. Powered by Granite code models, it offers general coding assistance for languages like C, C++, Go, Java, and Python. IBM itself has been using the assistant and saw 90% faster code documentation using in its own software development. IBM also unveiled several other models for emerging use cases. That includes two Granite Mixture-of-Experts architecture models, Granite 1B A400M and Granite 3B A800M, which are lightweight models small enough to be deployed on a laptop, or an edge device. And finally, we also introduced new versions of our Granite Time Series models, which outperform models 10 times their size. 

🖥️ Read the blog here

IBM executive stories

As we close Hispanic Heritage Month 2024, check out this interview with IBM’s Senior VP and Director of Research, Dario Gil to learn more about his perspective on innovation, his motivations and his journey at IBM:

IBM Research history

On October 18, 1995, IBM researchers Praveen Chaudhari, Jerome J. Cuomo, and Richard J. Gambino were awarded the National Medal of Technology for their discovery of special magnetic materials. Check out "This Day" in IBM Research History to learn more about this milestone:

Research Roundup

Highlighting new publications from IBM researchers that we liked the sound of:

• Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series
• Granite-Function Calling Model: Introducing Function Calling Abilities via Multi-task Learning of Granular Tasks
• MediTOD: An English Dialogue Dataset for Medical History Taking with Comprehensive Annotations

What are you most excited to try out from all the Granite updates launching this week? Let us know in the comments!        

Follow IBM Research on LinkedIn and subscribe to our Future Forward newsletter to stay up to date on the latest news each month on breakthroughs in AI, quantum computing and hybrid cloud.