By: Matt Fornito, Head of Artificial Intelligence, Trace3
Artificial Intelligence is present in our day-to-day lives and it’s expanding rapidly across enterprises. From Siri to self-driving cars to Netflix’s recommender system, AI is everywhere (heck it’s probably a prize in your cereal box). However, that’s the AI we see and interact with as consumers, known by most citizens for their AI capabilities. But what about the thousands of use cases that are less renowned but integral to organizations? How can organizations — how can your organization transform into leveraging AI? More importantly, what is necessary to not only have AI but to identify and successfully make a large-scale impact?
Most people outside of AI consider it this magical black box. I give it all my data, a magician inside whispers, “ALAKAZAM” and you now have AI and hundreds of millions in revenue gains. As Head of AI at TRACE3 and with several years of data science under my belt, I can assuredly tell you that’s not the case. Perhaps Edison’s quote sums it up best: “Genius is one percent inspiration and ninety-nine percent perspiration.” And by that I mean, everyone wants AI, everyone hears of or sees the value of AI, and AI is seen as this genius algorithm that can make sense of all of your data and automagically generate revenue. The reality is the desire for AI and the capability of these algorithms to make better sense of patterns and data than we can as human are that 1% inspiration. However, the 99%, the perspiration, that is everything that occurs underneath the hood. From storage, to networking, to compute, to data quality, to processes, to user access, to software stack, to talent, to education and evangelization – there are a plethora of components that need to operate correctly and efficiently to drive full enterprise scale solutions – to build these AI solutions that change how organizations make decisions.
However, while AI is an executive goal at over 80% of companies, and while it can be a complex undertaking to implement at scale to operate as an AI-transformed and innovative company, it does not start out as such. Artificial Intelligence is a garden that needs to be nurtured. And a garden starts with a seed. The seed is a use case, a business problem, a pain point, or gap in understanding. What many people don’t realize is that seed doesn’t need to be a chatbot that can comprehend and understand the entirety of human knowledge in every language and produce meaningful results. It can be much simpler.
When I was a data scientist at a major sporting goods retailer, the Director of User Experience noticed a higher-than-expected cart abandonment when shopping on the company’s website. When most people think of AI, they think of systems that require computer vision, natural language processing, or deep learning. The higher cart abandonment theory did not require such complexity. We built a separate webpage and conducted a simple A/B test sending 50% of customers through the old page and 50% through a new checkout page that was easier to understand. The process end-to-end took all of two weeks. We discovered the hypothesis was correct and that resulted in an additional $20 million in revenue per year by simply changing a webpage. Find a use case and business impact (return-on-investment). Plant the seed and nurture it and watch it grow.
As success begins to flourish that seed gains roots and flourishes, other seeds are planted, and before you know it, the organization is seeing a transformation – a flowering garden. And as a garden grows and grows, the problem moves from ‘how do I keep this plant alive’ to ‘how can I help this garden to thrive’ and additionally, ‘how can I maintain and manage the garden to sustain its beauty?’
The right infrastructure, software, people, processes, and data enable that garden to thrive.
Imagine that your organization now has 2-20 data scientists (or even more). We want them to be able to build things bigger, better, and faster. While management consultants can help innovate culture change and software platforms can optimize algorithms and processes, the most critical foundational element is the right infrastructure. Data scientists need data – and lots of it. The faster access to data on storage, the faster the data can move in transit, and the faster the models can be run on compute, provide accelerated AI success.
Imagine if instead of 1GB switches, your team had 100GB switches to move data – I don’t need to show you the math on the speed increase here. Now, imagine that you had an HPE Apollo server with 8 NVIDIA A100 GPUs to handle your multiple data scientists and their workloads. Compared to CPU-based workloads, these GPUs average 20x-200x greater performance. If the seeds are the use cases, the Apollo server is the fertilizer. Unprecedented growth begets more seeds begets new projects and new innovations and new models that continue to enable your organization to thrive.
Plant the seed. Water. Fertilize. You won’t regret what emerges.