I produce conferences, help run a professional association and offer workshops to those who are interested in digital analytics, machine learning, and data fluency.
Artificial intelligence is an umbrella term that covers new types of data manipulation including natural language processing, computer vision, robotics, and machine learning.
I'd say the invention of the wheel, the scientific method, and long-distance communications are probably more profound...but time will tell.
I got into computing in 1978, into online marketing in 1993, and into digital analytics in 2000. AI is a natural extension of learning about how to use technology for marketing.
Google, Apple, Amazon, Facebook, your phone, and eventually, your shoelaces. It's everywhere!
We're only just starting to scratch the surface. We have a vague idea about how AI can solve some our problems, but we haven't yet seen how it can solve new problems we have not yet addressed due to a lack of ability.
Bias in the data is one significant problem. Bad decisions can be made by simply accepting whatever the machine spits out, without worrying about whether the data encodes biases that we have an aversion to but have not expunged from our history. Race and gender are obvious issues, but the consideration of a wide variety of variables might subtlety downgrade individuals due to sentence structure, left-handedness, height, accent, or other insignificant and inappropriate features.
Perhaps more fearful is the use of this technology by bad actors. One need only look to Cambridge Analytica for a glimpse of a dystopia in the making.
What blows my mind the most is how complex it all is. It blows my mind that we are still trying to solve fundamental problems and tripping over seemingly insignificant problems that prove catastrophic.
The two biggest misconceptions about AI are that it is inherently evil or that it is the answer to all problems. It seems evil that the machine is making decisions or recommendations in ways that are "unexplainable." A lack of understanding leads to mistrust. Others think AI will solve all sorts of problems that we don't understand well enough to solve in other ways. Machine learning only works when the problem and the data are very well understood.
Humans will always be needed in marketing to perform three vital functions:
1. Determine which problem the machine should solve.
The machine may be able to recommend issues to address, but humans have to prioritize based on a holistic understanding of the problem set. There are so many trade-offs in choosing which problem should be tackled next, that it takes a human brain to make a 'gut feel' decision... and assume the responsibility and consequences for that decision.
2. Determine which data the machine should consider.
The machine can only crunch the numbers it is given. A human must decide which data sets might be the most informative or predictive. This requires imagination and ingenuity.
3. Determine whether the output makes sense.
The machine can make a statistically correct recommendation or take an action that accurately performs the required task, but has no ethical, moral, or even common-sense framework for identifying whether there are additional considerations. If you want to raise revenue, the machine will rightly tell you to sell $20 bills for $10. The answer is correct, but absurd.
The people who are faced with implementing it in the trenches. People trying to create systems to sell to the rest of us. And science fiction writers.
What blows my mind the most is how complex it all is. It blows my mind that we are still trying to solve fundamental problems and tripping over seemingly insignificant problems that prove catastrophic.
The human cognitive process, critical thinking, rhetoric, psychology, and business management.
I would develop courses in tool applicability. When should you use linear regression? When should you use a random forest? When should you use a neural net? When is an Excel spreadsheet enough?
Find a very narrow problem to solve that has an easy-to-measure outcome. Incoming email routing, multivariate testing on a landing page, email open rate. If you have enough transactions, then these simple problems are the best for learning about AI.
Data trustworthiness. Like the ingredients in a can of soup, marketers are going to need to understand their data flow supply chain well enough to know where the incoming data didn't live up to the promise of cleanliness, timeliness, consistency, etc. That, and suffering from over-blown expectations.
First, ask enough questions to determine if the vendor actually understands what "AI" means. If they're just using it as brochureware, you have to wonder what else they are misguided about. But mostly, test it! Run a trial! Do a proof of concept! If it's better than what you're using now and better than the others you are considering it doesn't matter if it's artificial intelligence or an artificial hip...it's better.
Marketing Data Scientist.
Have an opinion and state it clearly and often. If your ethics are aligned with the company, they will appreciate that somebody is representing the customers and keeping a watch on the brand. If your ethics are not aligned with the company, you shouldn't be working there.
Find a very narrow problem to solve that has an easy-to-measure outcome.
OPT IN.
OPT IN.
OPT IN.
Fully informed and on a drip-irrigation basis.
You can easily automate tasks as the tools improve. When you start automating roles you, by definition, stop being human.
Artificial Intelligence for Marketing: Practical Applications ;-)
Plus an infinite variety of resources depending on what role is of interest to the marketer. First, the individual must decide if they want to be:
Bicentennial Man
Ancillary Justice
X.AI
Learn which tools are best for which problems.