Sign up now to get your free virtual ticket to the Code-Less/No-Code Summit on November 9. Hear from executives from Service Now, Credit Karma, Stitch Fix , Appian, etc looking for more information.
A decade ago, IBM Watson was the name of a mainframe system that caused a stir around the world for its competitiveness in the game show Jeopardy.
IBM Watson in 2022 is a very different thing and is no longer a single system or even a single service, but a product brand name for a growing array of artificial intelligence (AI) capabilities from IBM. the more developed. Those abilities include natural language processing (NLP), speech-to-text and text-to-speech functions.
Over the past few years, IBM has continuously expanded catalog of AI services runs as a service on both the IBM cloud as well as on the public cloud platforms of Google, Microsoft, and Amazon. AI services provided by IBM Watson are also accessible through application programming interfaces (APIs) allowing developers to remotely use AI capabilities in applications.
What’s missing from IBM’s Watson portfolio, however, is the ability to directly integrate the code of core AI services into applications that run locally on-premises – or in the organization’s own cloud deployment. – without triggering an API call to IBM.
“There is an entire army of independent software vendors (ISVs) and partners looking to embed AI in their software products and services,” said Sriram Raghavan, vice president of AI research at IBM, told VentureBeat. “We have the opportunity to be the provider of enterprise AI core competencies that they can use to focus on innovation.”
To support ISV and its partners, IBM today announced the availability of three new AI software libraries that can be embedded directly into applications. Services include NLP, text to speech, and speech to text.
What IBM Watson’s Direct Embedded AI Software Library Enables
Raghavan explains that the software library form factor adds an extra level of flexibility for software developers looking to use AI.
As a software library embedded directly into the application, the application does not need to connect to external IBM resources. For applications that need to reside locally or can handle sensitive information for which external calls are not allowed, having the code as part of the application is essential for operations.
Directly embedding AI libraries into software applications is something IBM has been doing for its software for years.
“What ISV is benefiting from is that IBM has to solve the problems when we try to embed it in our own software, so a lot of innovation has gone into this form factor,” says Raghavan.
For example, NLP can be a very large codebase with the complexity and many use cases it can serve. With the embeddable software library, IBM optimized the offering to help ISVs use the code needed for the functionality they wanted to deliver.
How ISVs Can Embed IBM AI Libraries Directly in Containers
IBM has developed a series of reference architectures for implementation using its AI software libraries. It’s a model that’s not radically different from any other modern enterprise application implementation.
The entire NLP provisioning can be deployed with three to six containers, running on top of a Kubernetes container orchestration system, such as IBM Red Hat’s OpenShift platform. IBM also does not mandate the use of any specific AI inference or acceleration technology.
Raghavan explains that, for example, if an ISV wanted to use an NLP library to perform sentiment analysis on a piece of text, what they were calling at the time was a real pre-trained model. making inferences.
“Now, obviously as part of that, we’ve done a lot of optimizations for some of the smaller models and use cases that will run fine even on CPUs,” says Raghavan. “There are also reference workflows that when you scale them with the GPU, you can accelerate performance.”
More coming from IBM Watson in the future
Taking it a step further, IBM Watson is aiming to make it easier for ISVs to customize pre-trained models without having to retrain the entire models, which can be a time consuming process and complex. Raghavan emphasized that the first three AI software libraries IBM is releasing are just the beginning, with the potential for more libraries in the coming months and years.
“Internally we already have a richer set of libraries that we use to embed AI in our products,” says Raghavan. “We see this as an opportunity to bring continuous innovation from IBM directly to ISVs, the way we bring it to IBM products.”
VentureBeat’s mission is a digital city square for technical decision-makers to gain knowledge of transformative enterprise technology and transactions. Explore our summary report.