AI Will Not Yet Solve All The World's Problems–At Least Not Alone
We get a lot of interest in leveraging the capabilities of Artificial Intelligence for our work. However, finding specific opportunities to create measurable value is often more tricky.
A survey from BCG and MIT Sloan Management Review recently found that of 57% of businesses with AI strategies piloted or deployed only 10% are getting substantial value.
Part of the issue is that "AI" encompasses a broad swath of algorithms and disciplines ranging from natural language processing, to machine learning, neural networks, and computer vision. There is no single definition of AI, nor one generalizable model that can be applied as a "silver bullet." One of the most impressive recent showings is Open AI's GPT-3 model which provides an almost uncanny apparent "understanding" of written text. However, thus far, the applications are relatively limited, and the trained model size is too large to be cost effective. Learning from advances like this has allowed us to implement a zero-shot model to solve a challenging content tagging issue.
Amazon Go has also gotten a lot of press for its usage of AI technologies to create their "Just walk out" concept of the checkout process. While impressive, experts breaking down the cost of setup and ongoing usage of computers to power their computer vision expect that the break-even point for this investment will be something like 2040.
As the article concludes, humans best know their subject matter, and AI can provide cost-effective improvements when implemented in a way which creates feedback loops. Many of the best solutions partners we work with for SaaS augmentation of our solutions leverage AI to varying degrees in their specialty. We've been leveraging AI technologies for some time in our solutions and to improve the way we work.
HEAD OF ENGINEERING