SAN FRANCISCO – AI adoption has skyrocketed in 2018, but this would never have been possible without the innovations of the hardware industry. From semiconductor companies to chipmakers and flash storage, revolutionary advances in hardware have provided the backbone to the big data and AI revolution.
Pure Storage are one of the key innovators within this space, working closely with firms like NVIDIA to develop dedicated and robust AI hardware solutions such as FlashBlade, a flash storage-based array designed specifically to power AI applications. As VP of Product Management, Brian Schwarz has played a vital role since the beginning as part of the team that created FlashBlade. Brian brings a wealth of experience in data center infrastructure and an expert on the intersection of AI and storage, with an emphasis on data-intensive and emerging workloads.
Why should large enterprises consider AI?
Large enterprises should consider AI for two reasons. One, AI can have a profound impact on business value creation, and two, the technology has matured in such a way that it is possible for any Global 2000 company to acquire and run a successful AI project.
The largest tech companies like Google, Amazon, Facebook and Apple recognized the technology inflection point early on and also found compelling use cases, but AI isn’t just for hyper-scale tech companies. AI will impact every vertical industry in the years to come. Like many inflection points, successful companies will navigate the change and come out stronger, and those that can’t will end up being less successful.
How can businesses start thinking about their own everyday problems in relation to AI solutions?
Enterprises need to start by identifying the sources of data they have and what insights they might want from the data. In many cases, AI can bring structure and recognize patterns in data better than humans can. At Pure Storage, we use AI to provide a predictive service to our customers based on a large set of data from deployed systems. This Internet of Things example uses the collective wisdom from our entire population to better understand outliers and “fingerprint” failure before it occurs.
We also believe that simplifying and unifying your data into a data pipeline is an important foundation to build. You can’t predict all the questions you are going to want to ask your data, so the best first step is to collect and organize it.
We expect the software tools a capabilities to evolve rapidly over the coming quarters and years. You will be agile and gain value from these innovations as long as you have a solid foundation and a large data set to use them with.
What will the next generation of AI infrastructure mean in practice for enterprises?
There has been a tremendous increase in investment and innovation in AI-enabled technology, and that is evident in the number of semiconductor startups building AI chips for various segments and use cases. You may also see a tremendous set of open source software contributions changing the landscape over the coming years. On the storage front, all-flash will be the foundation. Legacy storage isn’t fast enough to keep up with modern GPUs — it wastes power, and is too expensive to setup and run efficiently.
In this new computing model, developers will collaborate with data scientists who build models instead of traditional linear algorithms. Data is the new oil and the key element that powers all these systems.
What are the key obstacles to making AI work for global enterprises?
AI requires vast amounts of data, and many enterprises lack the infrastructure to deploy it. In order to unlock all the capabilities AI has to offer, you must have access to the right kind of data and the infrastructure to support it. It is distinctly different from how traditional infrastructure is used today.
AI can improve businesses and cut down on costs, but only if data strategies are locked down. Enterprises should start their AI journeys by pinpointing the business problem for which they’re trying to solve. When deciding where and when to implement AI, it is necessary to measure the ROI for AI initiates. Like many early innovations it pays to start small, run fast, and iterate.
Once the business problem is identified, you need to collect, cleanse, and transform the data. Begin by identifying what kind of data is needed, collect enough of it, review it for accuracy, label it, and then unify the data on a single platform. After the data has been cleansed and unified all on one platform, AI algorithms may begin implementation.
The business problem/opportunity we saw at Pure Storage was, “how can we revolutionize our customer service experience by predicting failures with our equipment before they occur?” Wouldn’t it be nice if your car, computer, manufacturing equipment, power generation or whatever your business relied on failed less? In our case, we started collecting and curating data to help solve these issues and it has made a profound impact on our customer satisfaction.
What does competitive advantage look like in the context of AI?
The most known competitive advantage of AI is probably automation. Businesses use AI and ML technology to automate processes that would be too expensive or people to solve via traditional brute force. AI/ML can find patterns in data better than humans can. Success in AI is innovation and adaptability, and as Charles Darwin said, it isn’t the fastest or strongest that survive, it is the most adaptable.
AI also has the power to help businesses extract valuable insight from structured and unstructured data. The more usable data an enterprise has, the more AI can provide the business with reliable insight through real-time and accurate data analysis. Data becomes the foundation for building ever-improving AI products. This is part of the reason Google, Amazon, Apple and others realized AI could be so valuable — they have tons of data to use as raw material. Every business has data, and frankly, data that is unique to them. The key is harnessing the value locked within it.
What have you learned about the future of AI through working with both vendors and enterprises?
While there has been tremendous innovation in the technology of AI, what amazes me most is what customers want to build with these new tools. AI is not limited to the largest SaaS and cloud service providers — We work with regional banks, chip companies, healthcare providers, local manufacturing plants and more, all eager to use AI to improve their businesses and even the world. The world of AI isn’t just drones and self driving cars, we have yet to see a Global 2000 industry that can’t be improved with AI.
There have been tremendous strides taken in the technology and improvement of AI, and the limits of what it’s capable of will always be tested. AI is not limited to improved operational efficiency and cut costs through automation. It has the potential to provide enterprises with important and valuable insights to improve their businesses. The possibilities of AI go beyond anything we can really comprehend right now since a full new form of computing has materialized. It is as impactful as the creation of the original personal computer or the advent of the networking and software that led to the Internet. AI is going to make business and quite likely the world, will look very different in 20 years.
You can join Brian and the Pure Storage team at The AI Summit San Francisco, September 19-20