AI feels a lot like the internet in 1997
Remember the hype around the internet in the 1997 era? A Company’s stock could explode on the simple announcement that they deployed a website. Companies wanted to appear cutting edge and join the internet race even if the value proposition for many of their customers was unclear, and many executives lacked an understanding of what the internet even was or why they needed it.
To a degree, artificial Intelligence, in 2020, has taken the place of the web in 1997. Companies are rushing to claim that they have an AI product or that their products are now enhanced with AI, and that AI will bring loads of benefits to their customers and investors. However, if we're honest, most of the AI implementations today are logos and announcements with little actual working product. Why? AI/ML/NLP is very hard to do, and it is still early. Yes, AI is undoubtedly the future; of that, there’s little doubt, but implementing AI and its variants today requires careful planning, complex engineering, robust computing, and networking infrastructure, and, most importantly, a clear idea of what you’re trying to accomplish and why.
What’s more important now, rather than to throw another logo on a marketing doc, is to get an understanding of where your company fits into the world of AI, and to develop a strategy to move towards a viable AI iteration, one that ultimately benefits your company and customers. Creating a strategic foundation that will enable AI to be integrated into your products will save both time and money and, in the long run, will result in increased profitability and a competitive future.
Today, few people are capable of building a true AI product, and for the few that can, there is a dearth of tools for them to use to do so. Functions such as labeling and data analysis can be and are very complex and labor-intensive exercises and failing to do these correctly will result in a product that could perform poorly. Much of this work requires people with Ph.D. level mathematics or electrical engineering degrees and that are well versed in the complex world of data science.
Thus, it is more important to develop a long-term plan and vision of how AI will impact your production and products and architect accordingly than it is to rush something out the door that will have to be rewritten in a year or two. Just as in 1997, companies want to be able to show that they are leading-edge, but ultimately, by carefully setting the stage and building accordingly, the result will be a meaningful, scalable AI product, one that will create larger rewards for your shareholders, employees, and customers.