Unpredictions – what won’t happen with artificial intelligence in 2018

Artificial intelligence and machine learning are two of the key tools for the digital transformation of many businesses. From Amazon Alexa to autonomous vehicles, artificial intelligence is progressing at a very fast rate. However, there remain many technological limitations in terms of what machine intelligence technology can deliver in the short-term. Earlier in the year, I shared with Digital Journal four “unpredictions” - that are unlikely to happen - with artificial intelligence during 2018.

●     We Won’t Be Riding in Self-Driving Cars – While many are predicting a driverless future, we’re a long “road” away from autonomous vehicles that will take us to and fro while we read the news or otherwise spend our time behind the wheel. For a number of years ahead, human operators and oversight will still rule the roads, because the discrete human judgments that are essential while driving will still require a person with all of his or her faculties--and the attendant liability for when mistakes happen. Besides technical challenges, humans tend to be more forgiving about mistakes made by human intelligence as opposed to those made by artificial intelligence.

●     We Won’t Be Replaced by AI Bots This Year – While it is possible that AI agents might replace (but more likely supplement) certain administrative tasks, the reality is that worker displacement by AI is over-hyped and unlikely. Even in an environment where “Auto ML” (Automated Machine Learning) is helping machines to build machines through deep learning, the really complex aspects of jobs will not be replaced. Thus, while AI will help automate various tasks that mostly we don’t want to do anyway, we’ll still need the human knowledge workers for thinking, judgment and creativity. But, routine tasks beware: AI is coming for you!

●     We Won’t Get AI-Powered Medical Diagnoses – Due to a lack of training data and continued challenges around learning diagnosis and prognosis decision-making through identifying patterns, AI algorithms are not very good at medical decision automation and will only be used on a limited basis to support but not replace diagnosis and treatment recommendations by humans. AI will be increasingly deployed against sporadic research needs in the medical arena, but, as with fraud detection, pattern recognition by machines only goes so far, and human insight, ingenuity and judgment come into play. People are still better than machines at learning patterns and developing intuition about new approaches.

●     We Will Still Struggle with Determining Where AI Should be Deployed – Despite what you might be hearing from AI solution vendors, businesses that want to adopt AI must first conduct a careful needs assessment. As part of this process, companies also must gain a realistic view of what benefits are being sought and how AI can be strategically deployed for maximum benefit. IT management, business users and developers should avoid being overly ambitious and carefully assess the infrastructure and data required to drive value from AI. Best practices and “buy versus build” analysis also should be part of the conversations about implementing AI applications.