IBM expands its goals for AI and Quantum Computing

While no one ever doubts the legacy of technological progress that NYSE:IBM has done over the past few decades, there are surely some who have recently wondered if the company can sustain that kind of effort for the future. At a recent analyst day at their historic Thomas J. Watson Research Center, IBM made a compelling case that it’s up to the task, particularly in the areas of AI — generative AI, in particular — as well as quantum computing.

What was particularly notable was the fact that the company demonstrated a much closer connection between the work its research group does with advanced technology and the rapid “productization” of this work into its commercial product organizations. In both prepared remarks and in response to questions, it was clear that there is a renewed focus on ensuring that two groups are locked in terms of future prospects and development priorities.

As with many other organizations, this has not always been the case with IBM. The result has been that some potentially interesting research efforts have not always reached the market. Thanks to a very clear directive from CEO Arvind Krishna (who previously ran IBM Research) about the company’s need to focus on a few specific areas – hybrid cloud, AI and quantum – current head of research Dario Gil said the alignment between research and commercial product groups has never been stronger . The net result should be – and is starting to show – important new opportunities that make it to commercial products at a much faster rate.

A real effect of this new strategic initiative is the company’s very rapid development of its suite of AI tools which they call watsonx. First revealed at the company’s Think conference earlier this year (see “IBM Unleashes Generative AI Strategy With watsonx” for more), watsonx continues to evolve, largely driven by new features first developed by IBM’s research group. What was particularly impressive at the recent analyst event was the number of real-world applications and customer examples using watsonx that IBM was able to talk about. While acknowledging that many organizations are still in the exploratory and proof-of-concept phase when it comes to GenAI, there was still a solid set of company logos from real-world implementations that they shared. Additionally, IBM had an impressively thorough classification of applications for which companies are starting to use watsonx and genAI.

On the application front, IBM noted that the top applications they are starting to see companies leverage GenAI for fall into three main categories: Digital Labor or HR-related activities, customer service or customer support, and app modernization or code creation. Within these categories, the company discussed content creation, summarization, classification and coding applications. Given the long history of legacy mainframe-related software running on IBM mainframes, IBM noted particular interest from companies looking to move from legacy COBOL code to modern programming languages ​​using GenAI-powered tools.

Beyond applications, IBM talked about a number of technologies it is working on within its research group to improve its watsonx offerings. Specifically, IBM discussed its efforts in performance and scale, model adaptation, governance and application enablement. For Performance, IBM said it is working on a number of new ways to improve efficiency in how large base models. It’s done through various combinations of technologies that do things like shrink model size via quantization, improve the ability to share limited computational resources with GPU fractionation, and more.

Given its focus on open source, IBM also provided more information about all the work it’s doing with the AI ​​application framework tool Pytorch, which Meta (META) open sourced back in 2017. Leveraging the open source community as well as its own efforts, the company talked about how it makes significant improvements both in terms of optimizing model performance and opening up the ability to run Pytorch-built models across a wide range of different multi-vendor computer chip architectures. Adding a hardware abstraction layer like Pytorch opens up the potential for a much wider range of programmers to build or adapt GenAI models. The reason is that models can be created with these tools using languages ​​like JavaScript that are much more familiar than the chip-specific tools and their lower-level language requirements. At the same time, these hardware abstraction layers often end up adding quite significant performance penalties due to their high level (a problem that Nvidia’s ( NVDA ) Cuda software tools don’t suffer from). With the new Pytorch 2.0, however, IBM said it and others are making joint efforts to reduce that impact by better organizing where different types of optimization layers need to be and, as a result, getting closer to “on metal” performance.

On the model customization front, it’s clear that IBM is putting in a lot of work because they’ve realized that very few companies actually build their own models – most simply adapt or fine-tune existing ones. (To read more about that development and some of its potential industry implications, check out my recent column “The Rapidly Evolving State Of Generative AI.”) To that end, they discussed basic model tuning techniques such as Low Rank Adaptation (LoRA), parameter-efficient tuning, multi- task prompt tuning and more, all of which are expected to be commercialized within watsonx in the not too distant future. They also described the need to provide pedagogical guidance in the model building process to help developers select the right size model and datasets for a given task. While this may sound simplistic, it is an imperative, as even basic knowledge of how GenAI models are built and work is much more limited than people realize (or are willing to admit!).

IBM’s governance efforts – that is, tracking and reporting details about how a model is built and developed, what data is used to create it, etc. – appear to be an extremely important and important differentiating capability for the company. This is especially true in regulated industries and environments where the company has a large customer base. While more details on IBM’s specific governance capabilities are expected shortly, they shared some of the work they are doing to provide safeguards to prevent the inclusion of bias, social stigmas, obscene content, and personally identifiable information (PII) in datasets intended for model input. In addition, they talked about some of the work with risk assessment and prevention that they have done. IBM recently announced that it will offer indemnification to customers using its basic models to avoid possible lawsuits related to IP protection. Along with this governance effort, these two efforts clearly demonstrate that IBM is in a market-leading position to address critical concerns that some companies have about the trust and reliability of GenAI technology in general.

In the area of ​​Application Enablement, IBM talked a lot about the work it is doing around Retrieval Augmented Generation (RAG). RAG is a relatively new technology that supercharges the inference process, makes it significantly easier and more cost-effective for companies to leverage their own data, and eases the process of fine-tuning existing base models so that organizations don’t have to worry about creating their own models. IBM says it has already seen a number of its customers begin experimenting with and/or using RAG techniques, so it is working to refine its capabilities there to make the creation of more useful GenAI applications much easier for its customers.

In the world of quantum computing, IBM is already seen as a leader, largely because of the time they’ve already spent discussing the innovations they’ve made there. What was particularly impressive at the analyst event, however, was that the company showed off a detailed technology roadmap that stretches all the way to 2030. While some tech companies are willing to share their plans a few years out, it’s practically unheard of for a company to provide so much information so far in advance. In part, IBM realizes it has to because quantum computing is such a dramatic and forward-thinking technology that many potential customers feel the need to know how to plan for it. Simply put, they want to understand what’s coming in order to invest in the roadmap.

Full details of IBM’s specific quantum computer development will be unveiled at an event the company will host in early December. Suffice it to say, though, that the company continues to be at the forefront of this technology and is increasingly confident of its ability to eventually make its way into mainstream enterprise computing.

Given the long, sad history of early tech companies that no longer exist, it’s certainly understandable why some have doubts about 112-year-old IBM’s ability to continue to innovate. As it recently showed, however, not only is that spirit of invention still alive, it appears to be gaining serious steam.

Warning: Some of the author’s clients are suppliers in the technology industry.

Disclosure: No.

Source: Author

Editor’s Note: The summary points for this article were selected by the Seeking Alpha editors.