IBM Launches New WatsonX Foundation Models for Enterprise
IBM recently unveiled an AI platform called watsonx, which allows customers of generative AI to remain within IBM's ecosystem. The foundation model, now accessible to all after a beta period of two months, is specifically designed for businesses to construct, fine-tune, implement, and oversee foundation models for various purposes such as talent recruitment, customer support, IT tasks, and updating applications.
Furthermore, it grants the organization a competitive edge in contrast to Amazon SageMaker Studio, Google Vertex AI, Microsoft Azure AI, and Anthropic’s Claude extensive language model.
In May 2023, IBM initially gave a sneak peek and initiated a list of people waiting for the release of watsonx. This unique model, based on generative AI, was trained on massive amounts of unstructured data. What makes watsonx special is its ability to adapt to various functions and tasks with minimal adjustments, eliminating the need for repetitive training on new data sets. The advancements in ChatGPT further demonstrate the potential of foundation models in constructing conversational large language models.
Take a look at this handy guide about GPT-4 on TechRepublic.
According to IBM, watsonx has been influenced by over 150 individuals from various sectors who have taken part in the beta and tech preview programs. Out of these participants, more than 30 have kindly provided early testimonials.
Watsonx: Three Core AI Products
IBM announced that watsonx consists of three different configurations of generative AI models.
According to Tarun Chopra, the VP of Product Management, Data and AI at IBM, the company's recent launch on July 11 primarily emphasized on two platforms, namely watsonx.ai and watsonx.data. Furthermore, he stated that IBM will introduce watsonx.governance later in the year.
"On July 11th, we introduced the initial two SaaS services on IBM cloud, with watsonx.data also accessible on AWS, on-site. These elements function independently, however, we stand alone in amalgamating them as a unified platform," he expressed.
Build A Data Pipeline For AI
Chopra stated that watsonx.data is specifically created to assist customers in addressing issues related to the amount of data, intricacy, expenses, and management surrounding data utilized in AI tasks. This platform enables users to effortlessly access both cloud and on-premises environments via a solitary entry point.
He mentioned that watsonx.data serves as a lakehouse hub, similar to Databricks or Snowflake, which operates independently as an open-source repository. Additionally, it functions as a data source with features resembling a plugin for enhancing AI models.
VIEW: Open-source or commercial AI for enterprises? Former Siri developer Aaron Kalb shares his perspective (TechRepublic)
Chopra mentioned that it is possible to link the AI model with an S3 bucket or any other cloud object storage where your data resides. Alternatively, you can transfer the data to a repository. In the latter scenario, if the user possesses data connected to the AI model, they can effortlessly transfer that data to the watsonx.data repository. This repository offers an array of enhanced capabilities and features compared to a regular cloud object storage.
The company stated that watsonx.data utilizes suitable query engines such as Ahana Presto and Apache Spark. These engines are capable of handling various workloads, including data exploration, data transformation, analytics, and training and fine-tuning of AI models.
Chopra stated that if you have Excel files, jpegs, different tables, web pages, and other similar data to bring into the training set, you can store it in a watsonx.data instance. This way, you can establish the complete lineage for all of these sources. This is important because consumers may inquire about the origins of the data and it is necessary to provide them with this information.
Choose From Three Model Sources With Watsonx
Chopra stated that watsonx stands out in the field of artificial intelligence due to its distinctiveness in combining both hybrid and multicloud deployment. Moreover, it possesses the capability to leverage open source technologies, specifically running on Red Hat OpenShift and incorporating Hugging Face's libraries, with the advantage of already having thousands of them accessible within watsonx.
Chopra stated that watsonx.ai offers great versatility as there isn't a single solution to address all issues. It serves as a workspace where users can utilize three deployment options and three libraries. These libraries include an IBM-provided model, models from open source communities, and models created by customers themselves.
The organization stated that the models facilitate tasks related to processing natural language, such as answering questions, creating content and summaries, classifying texts, and extracting information.
IBM WatsonX: More Releases Coming Soon
IBM plans to make available choices for graphic processing units (GPUs) on their cloud platform. These GPU selections are intended to cater to the needs of major business operations, as stated by the company. Moreover, they are set on creating a comprehensive and powerful infrastructure that is both adaptable and optimized for artificial intelligence (AI) models on IBM Cloud by the end of this year.
Additionally, the organization made a statement regarding watsonx.data, expressing that it will employ the fundamental models of watsonx.ai to offer users the capability of employing human-like language to interpret and manipulate data efficiently.
According to IBM, the company plans to broaden the application of enterprise foundation models beyond natural language processing in the coming year. Moreover, they intend to develop models with over 100 billion parameters for specific purposes. IBM also mentioned that their governance capabilities will focus on assisting organizations in implementing lifecycle governance to minimize risk and enhance compliance.
