According to IDC 31% of organisations are currently in some sort of discovery or evaluation stage of an embryonic AI project. A further 22% plan to implement AI in the next 12 to 24 months and another 22% are currently in an AI trial. Clearly AI is a topic on the majority of lips.
There’s little surprise here. AI projects are already providing significant returns for those early adopter who are implementing trial projects. A number of farmer’s co operatives in the US have given trust over to calculus. Their algorithm analyses historic , projected forecasts and competitive data to provide accurate crop buying strategies, a job that was once the domain of highly experienced (and expensive) agronomists and traveling fortune tellers. Result are already providing greater predictive accuracy with better yields and quality from crops using the data led approach.
As these machine learning projects continue to succeed and fail, many industries see AI as a key objective for the next 2 years. Driven not by IT but by the understanding that AI has the ability to provide significant competitive advantage to anyone looking to leverage their data and all the insight it promises to yield, beyond predicting chess moves.
As the demand for AI continues to loose perspective a number of challenges arise. Key to these challenges is the sheer amount of computing power and data storage the AI era needs to survive. According to the IDC report an amazing 77% of AI wannabees fear their biggest barrier to entering into the league of AI early adopters, is the limitations of their current infrastructure. AI puts considerable pressure on a networks ability to perform. Intense, no really intense workloads, are just part of the average AI framework – clearly this puts strain on other aspects of a network’s ability to perform, like an important ERP application for example. Then there’s storage. AI can only work once you have significant data. Real time and historic. Your’s and the stuff you buy in. And it grows, everyday it grows – that’s just the nature of machine learning, data becomes an expanding commodity that adds to the accuracy of your AI project and the fuel to your AI framework.
The net result of this is a significant increase in your compute and storage requirements /costs. That’s not to mention the additional admin resource, space (add in cooler and power consumption costs et al) and all round general integration costs, plus the usual scalability headaches that go with such murky waters.
There is some good news. The advent of ultra high powered computing servers such as IBM Power9 and the aggregation of traditional infrastructure services for processing, storage and back up through hyperconverged software layers can significantly future proof any network environment. Add to this the ability to reduce storage costs through software defined storage and there is hope for a scalable, cost efficient Infrastructure solution on which your AI initiative could sit for the long term.
AI promises the world, and where it’s been adopted in earnest, it’s provided some of it’s beneficiaries with compounding competitive advantage. So, as the rest of the world and its commentators reflect on how AI will replace our current economic construct, let’s consider this; things are about to change, prepare and lay down the foundations for data growth and compute power, else prepare to be replaced.
If you’re considering embarking on an AI framework and would like to understand your current It infrastructures compatibility please call us to speak to a qualified AI infrastructure consultant.