IBM launches generative model services and more

IBM, like most tech giants, is betting big on AI.

IBM Watsonx, a new platform for building AI models and accessing pretrained models for generating code, text, and more, was announced at its annual Think conference.

It’s a slap in the face to IBM’s back-office managers, who were recently told the company will stop hiring for roles it thinks AI will replace in the future.

IBM says the launch was inspired by the challenges many businesses face in deploying AI in the workplace. In an IBM survey, 30% of business leaders cite trust and transparency issues as barriers to AI adoption, while 42% cite privacy concerns, particularly around generative AI.

In a roundtable with reporters, IBM’s chief commercial officer, Rob Thomas, said, “AI may not replace managers, but the managers that use AI will replace the managers that don’t.” It transforms work.

IBM claims that Watsonx gives customers the toolset, infrastructure, and consulting resources they need to create their own AI models or fine-tune and adapt existing AI models to their data. Watsonx.ai, IBM’s “enterprise studio for AI builders,” allows users to validate, deploy, and monitor models, consolidating their workflows.

But don’t Google, Amazon, and Microsoft offer this or something similar? Yes, briefly. Amazon has SageMaker Studio, and Google has Vertex AI. Azure AI Platform.

IBM claims that Watsonx is the only AI tooling platform with a variety of pre-trained, enterprise-developed models and “cost-effective infrastructure.”

“You still need a very large organization and team to bring [AI] innovation in a way that enterprises can consume,” IBM SVP Dario Gil told reporters during the roundtable. IBM’s horizontal capability depends on that.

We’ll see. However, IBM Watsonx.ai offers seven pre-trained models, some of which are open source. It also includes thousands of hugging face models, datasets, and libraries. IBM will donate open-source AI development software to Hugging Face and make several of its in-house models available on its AI development platform.

At Think, the company is highlighting fm.model.code, which generates code for fm.model. NLP, a collection of large language models, and fm.model.geospatial, a NASA climate and remote sensing model. (Strange naming? You bet.

fm.model generates code like GitHub. Copilot.code builds coding workflows from natural language commands. Fm.model. NLP includes organic chemistry text-generating models. And fm.model. Geospatial predicts changes in natural disaster patterns, biodiversity, land use, and geophysical processes.

These may seem familiar. IBM claims that a training dataset with “multiple types of business data, including code, time-series data, tabular data, geospatial data, and IT event data,” distinguishes the models. Take it at its word.

“We allow an enterprise to use their own code to adapt [these] models to how they want to run their playbooks and their code,” IBM CEO Arvind Krishna said in the roundtable. “It’s for use cases where people want their own private instance, whether on a public cloud or their own premises.”

IBM says its software products and services use the models. Like fm.model. Watson Code Assistant, IBM’s Copilot, generates code using plain English prompts across programs like Red Hat’s Ansible. IBM claims that integrating fm.model.NLP with AIOps Insights, Watson Assistant, and Watson Orchestrate, its AIOps toolkit, smart assistant, and workflow automation tech, will improve IT performance visibility, IT incident resolution, and customer service.

FM.model.geospatial supports IBM’s EIS Builder Edition, which helps organizations address environmental risks.

IBM introduced Watsonx.data, a “fit-for-purpose” data store for governed data and AI workloads, along with Watsonx.ai. IBM says Watsonx.data provides a single point of entry, query engines, governance, automation, and integrations with an organization’s databases and tools.

Watsonx.governance protects customer privacy, detects model bias and drift, and helps organizations meet ethics standards, according to IBM.

New equipment
IBM announced a new GPU offering in the IBM cloud for compute-intensive workloads like training and serving AI models in conjunction with Watsonx.

The “AI-informed” IBM Cloud Carbon Calculator lets customers measure, track, manage, and report cloud-generated carbon emissions. IBM claims it was developed with Intel and based on IBM research to visualize greenhouse gas emissions across workloads and cloud services.

IBM is doubling down on AI with both products and the new Watsonx suite. Vela, an AI-optimized cloud supercomputer, was built by the company. It also collaborates with Moderna and SAP Hana to scale generative AI.

The company expects AI to add $16 trillion to the global economy by 2030 and automate 30% of back-office tasks within five years.

“When I think of classic back-office processes, not just customer care—whether it’s doing procurement, whether it’s elements of supply chain [management], whether it’s elements of IT operations, or elements of cybersecurity—we see AI easily taking anywhere from 30% to 50% of that volume of tasks and being able to do them with much better proficiency than even people can do them,” Gil said.

Wall Street has historically rewarded optimistic (or pessimistic, if you’re a humanist) predictions. In Q4 2022, IBM’s software segment’s automation solutions grew 9%. Data and AI solutions, which emphasize analytics, customer service, and supply chain management, increased sales by 8%.

Seeking Alpha suggests lowering expectations. IBM sold Watson Health at a loss after technical issues ruined high-profile customer partnerships. IBM faces competition in AI from tech giants like Microsoft and Google as well as well-funded startups like Cohere and Anthropic.

IBM’s new apps, tools, and services: impactful? IBM hopes. We’ll see.