Murli Thirumale is VP-GM of the Cloud Native BU Pure Storage, which acquired Portworx Inc., where he was cofounder and CEO.
One of the most overhyped aspects of the discussion around ChatGPT and other generative AI applications is that they will be a replacement for people, but these powerful new tools are not going to replace humans—at least not for a long, long time. Goldman Sachs says about a quarter of all work tasks in the U.S. could be automated, but that most jobs are only partially exposed and that automation will likely complement more jobs than it replaces.
In my view, the wave of interest in generative AI should be seen as part of the AI evolution, which includes machine learning. We’ve already seen what machine learning can do for automation. Machine learning enables systems to improve themselves through experience and data. Its many use cases include everything from customer behavior analytics and recommendation engines to predictive maintenance. I like to think of it as AI without the mystery.
For data scientists and others working in machine learning, Gartner notes that data science machine learning is evolving as ML grows across industries—from a focus on predictive models to a more dynamic, data-centered discipline, particularly with the growing interest in generative AI. From an infrastructure perspective, containers and Kubernetes can play an important role in accelerating AI research, especially as data scientists work with larger and larger data sets in the age of generative AI and natural language processing.
Separating The Science From The Pipeline
Consider the fact that data scientists work on the AI /ML stack, but they are not traditional computer programmers. They are researchers and content specialists—those who are more conversant in a domain rather than infrastructure. They are likely not familiar with things like Kubeflow or Jupyter notebooks that are essential to machine learning developers.
The data sets that researchers work with might be in the cloud or on-prem. Whatever they’re building, they want to be able to run a project in different compute nodes depending on the model. But tasks like patching and upgrading to make sure they can port their app from one system to another can take a lot of time.
By separating the science from the AI pipelines, containerization can be hugely beneficial to data scientists and other researchers working in AI. Containers let them drop their logic into code. By containerizing, researchers can hand off to someone else to run.
Generative AI Is Like A Flavor Enhancer For Your Brain At Work
Letting that underlying IT help you as a human thinker is what appeals to me on a personal level about generative AI. It can be a very powerful aid. I like to think of ChatGPT as my data buddy. Growing up in Indonesia, we used a flavor enhancer called Ajinimoto (a brand of monosodium glutamate or MSG) on everything. It really does make food taste better. To me, generative AI is like MSG for your work. It can improve the quality of what you are seeking to achieve by digging out extra research.
But the food has to be decent to begin with. Gen AI can be a way to rationalize what you already are thinking. The key is to ask it to either verify or invalidate—so it can either grow your knowledge or tell you you’re wrong. And it is up to you to troubleshoot and decide if the results are shoddy or really great stuff.
In enterprise IT, companies will see Gen AI put to tremendous advantage with their proprietary data as human workers employ it for internal and customer-facing use cases. At Pure Storage and Portworx, we use it in very detailed vertical slices where we have the utmost confidence in the data because it’s ours.
Under those conditions—proprietary data and a very specific set of tightly controlled, tightly defined use cases—you can use generative AI to provide very useful things. To create test cases for a new capability, for example, you can point it to all of your existing tests. In the data management world, that might be confidence testing or system testing, for instance. With its APIs, the generative AI can generate very specific new test cases that you can rely on because they’re all based on very accurate data.
As we look out to a future that will hold even more AI innovation, infrastructure that provides the elasticity and agility needed to support the evolution will become even more critical.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Read the full article here
