IBM has made it public that it believes AI will replace workers, and at its annual Think event this week, the tech giant plans to help business leaders do the same. clarified.
Central to this strategy is IBM’s newly announced Watsonx product suite. It is essentially a collection of ML tools, hardware, models, data storage and consulting services, with Big Blue’s claim to make it easier for customers to integrate machine learning. existing product stack.
IBM touts the platform as a “complete technology stack” for training, tuning and deploying AI models, including underlying models and large language models, while ensuring strict data governance controls. .
IBM: We Can Be the Foundation for Your AI Banishment
In its current form, Watsonx is actually a collection of three product offerings: Watsonx.ai, Watsonx.data, and Watsonx.governance. As we all know, Watson enjoys a high level of brand recognition these days.
Watsonx.ai is perhaps the most important of the three, as it combines IBM’s curated set of pre-trained underlying models with access to Hugging Face’s open source model library.
Depending on the size of your model, training can be an incredibly time-consuming and expensive task. These foundational models are essentially starting points for customizing and tuning the models, IBM explained.
IBM CEO Arvind Krishna said, “The underlying model makes AI deployments much more scalable, affordable and efficient.” “Clients can quickly train custom AI capabilities and deploy them across their business.”
When Watsonx.ai is released in July, the underlying models available will include: fm.code, an AI model designed to generate code snippets and automate IT tasks. fm.NLP, a natural language model that can be tailored to industry-specific domains. fm.geospatial uses climate and remote sensing data to help businesses understand and plan for weather-related disruptions.
And for things that don’t quite fit into those boxes, as I mentioned, IBM has partnered with Hugging Face to provide a library of open source models.
As for the hardware on which these services are built and delivered, IBM says it will make “new” GPU instances available to customers. But we don’t know much about them other than the fact that they use his Nvidia accelerator.
register We reached out to IBM and Nvidia for comment on the architecture behind Watsonx. I will let you know if I get any reply.
Given what we know about IBM’s recently revealed Vela AI supercomputer, the H100 is also out of the question, but we can speculate that Big Blue is using Nvidia’s three-year-old A100 for training. increase.our sister site next platform I recently took a closer look at IBM’s Vela system. If you want to know more about how the system works, please see here.
That said, even if IBM deliberately obscures and uses software abstractions to migrate workloads to more efficient architectures as they become available, I we won’t be surprised. Given the number of AI accelerators on the market from the likes of Intel, Nvidia and AMD, this could help IBM avoid being trapped in a walled garden.
In addition to AI model training and refinement, Watsonx includes a data store service built on Open Lakehouse, a type of architecture designed to support analysis of both structured and unstructured data. is also included. IBM says it is tuned with artificial intelligence in mind. However, the details were particularly thin in IBM’s announcement, as the service would take months to roll out.
In addition to providing data storage, Watsonx.governance also provides tools to reduce the risk of applying AI models to sensitive customer data. IBM says the product can “proactively” detect bias and drift in models and avoid ethical clashes with the use of AI. But like IBM’s Watsonx.data service, we’ll have to wait until later this year for the pieces of the puzzle to become publicly available.
IBM eats its own dog food
To no one’s surprise, Big Blue is working to incorporate many of these tools into its existing product portfolio.
In its announcement, IBM revealed four software products and services it plans to integrate with Watsonx. The first included Watson Code Assistant, a tool that automatically generated code snippets based on user input. The service appears to have attacked Microsoft’s GitHub Copilot X service, which is built on top of Open AI’s GPT family of large-scale language models.
IBM also plans to use NLP models to accelerate its AIOps Insights platform and use machine learning to pinpoint anomalies in IT processes to facilitate quick and effective mitigation by support teams. increase.
Watson Assistant and Watson Orchestrate will also be AI tuned at some point in the near future. The former is an AI chatbot designed to interact with customers and employees, while the latter is an automation toolkit.
Finally, IBM plans to integrate the fm.geospatial underlying model into its Environmental Intelligence Suite to identify environmental risks before they disrupt operations.
IBM describes many of these tools as helping employees work more efficiently, but wonders how these tools can also be used to “streamline” employees’ work. I can’t help it. Note that IBM is rather active on this idea.
Earlier this month, Krishna told Bloomberg that up to 30% of Big Blue’s back-office operations (about 7,800 roles) could be replaced by AI, and the IT giant will delay hiring of those roles over the next five years. said it was likely. ®
