By Tony Paikeday, Senior Director of AI Systems, NVIDIA; and Holland Barry, Senior Vice President and Field CTO, Cyxtera
In a session with members of New Tech Northwest on August 25, 2020, we will focus on the unique situation many companies have realized in their pursuit of AI.
Many organizations find that the majority of their AI models never make it into production. “Model debt” is becoming a bigger problem for most teams that want to innovate with AI, with many struggling to achieve success at scale. In many organizations, data scientists have led the charge to develop model prototypes for specific use cases. Their objective is fixed on innovating through experimentation, algorithms and creativity rooted in data. However, far too many organizations have tried to deploy solutions without considering the IT infrastructure and platform needed to support their AI strategy. The need for short term success has resulted in the sprawl of “Shadow AI” – innovation silos operating without an IT platform. But are those successes scalable? Will an organization be able to cost-effectively expand their rollout of AI over time? Or, is the solution at hand just a mere shadow of what could be achieved?
There is good news, however. Companies can re-assess their IT infrastructure for AI now and make adjustments that will enable future growth and expansion without breaking the bank. Our discussion will cover why AI models are unlike conventional software, requiring a unique approach to infrastructure. We’ll also address the impact of Shadow AI in your organization and methods to work across teams and departments to create IT strategies that effectively address your AI directives.
In advance of the session, ask yourself and your colleagues these types of questions:
Have you considered the rising data gravity and OpEx impact associated with cloud for AI model development?
Have you observed too few AI models get deployed in production, and has it been a challenge scaling AI across your business?
Are the needs of data scientists, DevOps and IT fully aligned in your organization, or is there room for improvement?
Are you confident your IT infrastructure is constructed properly to support your AI strategy?