The last few months have been quite a ride. I took off a few months to help my wife take care of our son back in October and what I assumed would be time spent mostly changing diapers turned out to be much more (I may write about this in a later post).
In addition to having lots of family and friends visit and a bit of traveling back to the east coast, I managed to cram in an Intro to Machine Learning class at Udacity. The only way I can describe it is INCREDIBLE! After spending the last few years leading large teams in digital strategy, project management, and consulting, it was refreshing to get back to coding and learning something new.
I can hear someone out there asking: “Why would you spend your free time in python, learning how to analyze data with machine learning algorithms?”
The short answer is that this is where the world is headed and we all should learn at least one new skill every year.
So what exactly did I learn? Here are the top 5 things that stuck with me from the Udacity class and my own self learning:
Machine Learning (ML) and Artificial Intelligence (AI) are here now, so if the machines will be our overlords, why not jump teams and join them…All kidding aside, AI and ML are already in many products that we use each day and will continue to use in greater numbers and frequency (Amazon Alexa / Echo, Google Now, Google Home, Microsoft Cortana, Self-driving Cars, Parking Assistance, etc.). Many of our interactions online are with Chat Bots that learning from human interactions to get increasingly more conversational each day. I wanted to have a deep understanding of the underlying technology to not only have relevant discussions on the topic with clients and business partners, but also look for new opportunities in this space.
Machines will not make (all) humans irrelevant. This was something that I found interesting as I delved into the accuracy rates of the various algorithms and methods that make ML and AI possible. With most generalized ML algorithms, the best we can currently hope for is between 75% — 90% accuracy. Not bad, but not that great either. There aren’t many situations where 75% accuracy is celebrated. This is where a concept of Hybrid Intelligence fills in the missing pieces to help machines with particularly difficult decisions that humans have evolved to answer very effectively.
The next wave of AI will have us texting with our AI companions in much the same way that we text each other now. The first wave of mass-market AI will be products like Alexa and Google Now that require you to voice your commands out loud. This is a great start and novel at first, but eventually speaking out loud at home when you are alone will seem weird and unnatural (I am there already with Alexa and Google Now). I believe Mark Zuckerberg in the right path with potential integration with FB Messenger and his own custom smart home set up. The coming AI interfaces will leverage the gains made in natural language processing to give us the the ability either voice commands such as lights off or text a picture of a friend to our AI assistant to unlock the front door when they arrive (this involves natural language processing, home security, facial recognition, chatbot, etc…fun stuff).
Many jobs and tasks that are thought to be safe will be disrupted.However, this only means that we need to acquire new skills to move up the ladder into positions or jobs that require human judgment and intuition. Jobs that are repetitive, sequential, etc. are ripe for AI and machine learning right now. However, tasks that require lots of input but have a predictable outcome are being taken on by machines now. Legal documents that are just templates to be filled in by paralegals, legal contracts that don’t/won’t change, technical documentation such as data schemas, network diagrams, devops processes; these are all jobs that machines are doing every day in increasing frequency.
The cost of prediction continues to fall as machine learning algorithms get more accurate and we accumulate more data. Many companies hire consulting firms to analyze multiple scenarios and lots of data to give their best options for a critical decision. These very highly paid consultants are not cheap, but do serve their clients well. However, I believe that AI and machine learning will not necessarily replace the need for consultants, but it will decrease the cost of consulting, data analysis and prediction across every single industry. As we see better use of the data that is accumulated each day and we all get better as asking the right questions of our AI companions, we will see its use in common use from preschool to nursing homes.
Just as millennials were the first generation to grow up with the Internet, we are about to see the first generation (my son included) that will grow up with AI as the norm. The next few years will be really exciting and I look forward to being a part of it.
Daniel is a digital consultant specializing in IT advisory on technology strategy, investment, and implementation. He helps companies solve complex and strategic problems across multiple industries and domains. His drive to find solutions for clients and attain personal growth for himself are what keeps him at the forefront of innovation and helps him guide teams and organizations to cultivate amazing products and services. He can be found on Twitter at @dewilliams.