IBM’s Watson Gets a Granite Edge

The big tech giant introduced new generative AI models and capabilities that could make Watsonx more attractiveMost big-tech players today are already a part of the  conversations around artificial intelligence (AI) and its potential to make our lives easier. IBM, once considered a giant amongst pygmies in the world of technology, sought to be back in the reckoning with Watsonx four months ago and has now gone a step further by releasing new GenAI models and capabilities. 

These new models that have been launched under the Granite series, seem to be large language models on the lines of OpenAI’s efforts with ChatGPT. They can summarize, analyze and generate text, but there’s nothing more that IBM is willing to share about Granite. However, it does note that data used to train the series would be revealed in the next quarter. 

What’s the Granite bit?

However, one of the company executives did share some of the details stating that the new models were developed using curated, enterprise-quality data instead of publicly scraped data. An article published on TechCrunch quotes Tarun Chopra, IBM’s VP of product management for data and AI to suggest that the Granite series has data subsets that are specialized from different domains. 

The official explains that there’s a model trained on finance data that allows AI builders to use a much smaller model that can perform as well as the larger ones. They support most enterprise natural language processing tasks including summarizing, content generation and extraction of insights from the data sets. 

Better models and better deployments

This works in tandem with Watson.ai that allows users to test, deploy and monitor models after they get deployed. IBM is also coming up with Tuning Studio that is a tool which allows users to tailor generative AI models to their data.  It allows customers to fine-tune models to new tasks with as few as a 100 to a 1000 examples. 

The report also notes that Watsonx.ai would also get a synthetic data generator for tabular data that is used in relational databases. The company had claimed in a statement that by generating such data and internal datasets, companies can use the generator to extract insights from AI model training and fine tune them with reduced risks. 

Watsonx.data to launch soon

The generative AI capabilities becoming available in Watsonx.data later this month will allow users to simplify and accelerate the way they interact with their data … To help illustrate how this experience might play out, say a user is looking for specific data, says Chopra in the report. 

He also adds that by using the AI chat assistant interface in Watsonx.data, the Watsonx.ai conversational model can generate a text response alongside API calls and parameters to support that user’s request. It’s also possible to import external data using the same interface, and the AI model will perform semantic data enrichment.