The term “AI Washing” has become a standard part of our vocabulary, courtesy of vendors who want to tag anything that smacks of analytics with the AI label. Startups are especially prone to this, as AI is how they try to differentiate their products. Typically, their AI works great in demos, but when it’s actually installed and tested, the results fall far short of any promise.
Why is this the case? What is the difference between AI that is reliable and solves real problems and demoware?
- Domain Expertise. AI works best when there is a thorough understanding of the domain in which it is used. In our domain, this understanding and knowledge comes from being a leader in wireless, wired, and SD-WAN. Being named a leader for fifteen straight years in the Gartner Magic Quadrant for Wired and Wireless LAN Access demonstrates that fact.
- Data Science. Our data scientists work in tandem with our network engineers so that problems are tackled from the right approach and with the right tools. We use data, not naming gimmicks. It also helps to have years of experience working with big and small deployments to understand the inherent differences from one environment to another. Over 100 AI patents to draw on doesn’t hurt either.
- Data. This is one of those “Laws of AI Physics” that startups try to wriggle out of. Because startups have very few customers, they have very little data to draw on to build and train AI models. To compensate, they either try to bulk up their limited collected data with comments like “we collect the right data” or “we have access to billions of packets,” or they pretend large volumes of relevant data across a wide variety of customer environments doesn’t matter. Luckily for our customers, we collect highly curated and carefully organized data from 1.3 million network devices, 10 million clients, and 90,000 customers—every day. Every. Single. Day.
- Scale. Playing in a sandbox that is mostly trials and demos means that the analytics are not proven to work across enterprise environments. Algorithms designed for 10 APs and 100 users will deliver suspect results in larger environments, if they even work at all. When you’re responsible for a network that’s running thousands of APs, you need to know that the AI insights you’re receiving are coming from peer or like sites.
- Maturity. We’ve learned a lot over the last five years. While not everything worked the first time in all environments, our AI is now proven in small and large installations – across all industries. Add in customer validation that Aruba ESP (Edge Services Platform) delivers the AI-powered insights and automation that IT teams need to solve their problems and optimize network performance, and you are well on your way.
No matter how polished a startup’s demo looks, it’s just that—a suggestion of what might be accomplished if they could overcome significant hurdles. But they can’t. So, if you have a hybrid workplace, the need to improve IT efficiencies, or new business models that are accelerating your digital transformation, then proven, reliable AI will help you automate your network to achieve your goals.
If you want to see the difference between demoware and real AI, (and have some fun along the way), check out our “Would you rather?” video series:
- Would you rather: AI Search
- Would you rather: AIOps Peer Comparisons
- Would you rather: AIOps Dynamic Baselines
For a more in-depth review of how Aruba ESP leads the competition in more than just AI, check out our connectivity and security solutions.
We’ll be talking about Aruba ESP as well as Aruba’s AI and security solutions at Atmosphere ’21, our free, virtual conference that starts on April 13. Hope to see you there!
