DataStax uses ThirdAI to bring artificial intelligence to its databases

DataStax on Wednesday said it is joining the Houston startup The third AI bringing large-scale language models (LLMs) to its database offerings, such as DataStax Enterprise on-premises and NoSQL database-as-a-service AstraDB.

The deal, according to DataStax’s chief product officer, Ed Anuff, is part of the company’s plan to bring artificial intelligence to where data lives.

ThirdAI can be installed in the same cluster, on-premises or in the cloud, where DataStax is working because it comes with a small library and the installation can be configured with Python.

“The advantage is that the data does not have to go from DataStax to another location, it is only given to ThirdAI which is close to it. This ensures privacy and speed because no time is lost in transferring data over the network,” said DataStax.

“ThirdAI can be run as a Python package or accessed via the APIaccording to the customer’s preferences,” said the spokesperson.

Enterprises running DataStax Enterprise or AstraDB can use the data in the database with ThirdAI technology and LLMs to improve their AI applications. Basics from ThirdAI can be trained to understand data and answer questions, such as product recommendations that will lead to sales, based on customer profiles.

ThirdAI’s integration LLMs you will see DataStax display the initial Bolt technology, which can better achieve AI learning on CPUs compared to smaller GPUs. The advantage of this is that CPUs are often cheaper than GPUs, which are often used for AI and machine learning tasks.

“The Bolt engine, which is an algorithmic accelerator for training deep learning models, can quickly reduce computation. The algorithm achieves neural network training in 1% or less floating point operations per second (FLOPS), unlike traditional tricks such as multiplication, pruning, and regularization , which adds up a bit,” said ThirdAI in a blog post.

“Speed ​​is seen naturally on any CPU, whether it’s Intel, AMD, or ARM. Even old CPUs can be reengineered to train billions of models faster than A100 GPUs,” he added.

Bolt can also be invoked with “minor” changes to Python’s machine learning pipeline, according to ThirdAI.

The announcement by ThirdAI is the first in a new partnership program that DataStax is establishing to bring additional technology from AI startups that will help businesses with data residing on Datastax databases to develop artificial intelligence applications.