I help executives digitally transform their businesses using AI.
AI layers judgment on top of prediction to take actions without human intervention.
I disagree. Humanity is the most profound thing humans have ever worked on. It is incredibly important that we don’t lose sight of the human in our pursuit of AI. In my view, AI can help us to fully appreciate and understand the profundity of our humanity, if we use it the right way.
In 1998 I executed my first program in a language called BASIC. It was at that very moment—having a machine take data from the world, predict something based on it, and take action, without the need for my intervention—that I realized AI was going to rock our world!
"Humanity is the most profound thing humans have ever worked on. It is incredibly important that we don’t lose sight of the human in our pursuit of AI."
If it weren’t for the AI in your car, you’d be getting terrible, terrible gas mileage.
Fuel injection systems before the 1980s used to pump a steady stream of gasoline into the engine to mix with air and ignite, generating drive. Fuel injection systems today use sensors to collect data about the conditions of your engine—such as temperature, pressure, and airflow—and pump that data into a model that optimizes the amount of fuel to pump for the desired drive.
More sophisticated models actually use data from other sensors, including the angle and momentum of the car, to time the ignition cycle and update desired drive.
Cars have been using AI to optimize fuel injection since the early 1980s—which is incredible to think about, considering the amount of fuel that has been saved and pollution that has been reduced because of “early age” AI!
AI frees humans up to work on the things that are important to humanity—things that only humans can do.
AI may be able to do your job at lower cost. And if it does, the transition will take time and you will see it coming. New capabilities send a signal—an opportunity. They provide the context for you to have a discussion with your professional network about how you can better serve them. They prompt a reconnection with educational opportunities that you might not have considered before. They enable you to have the conversation about how you can again be of service to your community.
That we will all become “oversmoothed” to vanilla.
I was speaking with some members of the next generation to come out of high school, the “Zennials,” and they told me that they prefer the algorithm to know everything about them because it enables it to show them what to do (and what to watch, eat, read, or how to interact).
Models that support AI work by making generalizations from data and then making judgments based on those generalizations. If we all become the generalization, then we will no longer have our own unique perspectives—or the need for experimentation!
That it’s a new technology. AI has its roots in mathematics that have been around since before the American Revolution.
In fact, the only thing new about AI is that we now have a massive amount of mobilized data, and the cheap, scalable, and powerful computing architecture necessary to make its use cost effective.
The ability to govern, set policy, cultivate institutions, and grow our children.
The individual consumer. Your spend is your vote. What we as AI leaders need to do is inform the consumer what these technologies mean, how their data are being used, and what AI implies for the future…so they can make informed choices.
When self-driving cars started driving their passengers into roadblocks, my mind was blown, but not in a good way.
It underscored the level of quality we are going to need to hold these algorithms to in order for them to be deployed at scale. I do not believe we have yet considered the standards of performance and expectations we should set for AI that is to be deployed at scale.
"What we as AI leaders need to do is inform the consumer what these technologies mean, how their data are being used, and what AI implies for the future…so they can make informed choices."
My recommendation is to become deeply immersed in a topic of substantive importance—something that is domain-specific, like biology, environmental studies, or entertainment management—and major in it.
Then, minor in a methodological science such as computer science, applied mathematics, or econometrics. It is only in the combination of the two that we will see game-changing domain-specific applications of AI.
People will bring new perspectives on old problems that pure methodologists or programmers just haven’t been exposed to. These use cases are, in my opinion, the next frontier.
We need more programs that blend theoretical understanding of AI with practical experience. As a member of the advisory council for Emory Goizueta Business School’s STEM-credentialed Master of Science in Business Analytics, I believe they are setting exactly this standard.
I am always so impressed with the quality and drive of these students, and I believe programs like this are going to change the landscape of business education.
The MSBA program combines business, data, and technology to prepare students for careers as effective business data scientists. This 10-month, immersive program emphasizes hands-on learning in real-world partnerships with organizations like FedEx, InterContinental Hotels Group (IHG), and The Home Depot, so when they’re done, they hit the ground running in any industry.
Start small and look for a low-risk, solid win. Organizations that go for the big, slow, expensive initiative are taking on too much risk.
Low-hanging AI wins are everywhere if you know where to look. And once you have success there, then doing the next one is that much more exciting to your leadership team!
Focus on providing predictions and hypothesis tests with actionable, impactful storytelling. Big data and AI get a bad rap for being expensive and without reward, mainly because people forget that they’re supposed to be solving an important, common-sense issue.
How does your software provide us with actionable and impactful insights?
How does your software quantify and report uncertainty about its recommendations?
How does your software enable our organization to address bias, test hypotheses, manage organizational change, and deploy at scale?
Middlemen in media buys are getting their lunch eaten. Soon it will be breakfast and lunch, too.
Assemble your first "AI Review Board." AI cannot remove bias by itself. But AI does create the opportunity for a board of diverse perspectives to collectively evaluate an application by making the algorithm’s training and validation transparent.
It’s a great way to apply the jury theorem to reduce bias in AI, manage public expectations, and reward rising tech leaders.
We need to work towards a model of customer data ownership.
AI-powered customer advocates—owned and operated by the customers themselves via a mobile device—will enable differential privacy at the edge while still allowing incredibly insightful personalized recommendations. It will be like having a personal assistant instead of a big brother.
I see new channel strategies in text and chat as critical to brand evolution. Combined with AI, we can foster responsive, fully social (as opposed to para-social) relationships with customers. And we’re not even close to fully leveraging these strategies!
"Big data and AI get a bad rap for being expensive and without reward, mainly because people forget that they’re supposed to be solving an important, common-sense issue."
Minority Report. The movie is an incredibly prescient examination of where the world could go in the years to come. And we need to consider the issues the movie raises with respect to liberty, justice, inequity, and privacy.
The Master Algorithm by Pedro Domingos, a computer science professor out at U. of Washington, and someone I shared the stage with a few years ago at the National Association of Business Economists meeting. Pedro gives a really balanced and informed view on what AI is, where it’s been, and where it’s going—which stands in stark contrast to the much over-hyped perspectives we often hear about.
Our team at Search Discovery has developed a technology called Astrologer that plugs into any site with an analytics implementation and a clearly defined conversion goal.
Based on the digital behaviors of a site user in real-time, it leverages machine learning to predict conversion probability of the user in real time, and then exposes those scores for use by the company.
We activate to personalization engines on the website that optimize for the goal, or out to other platforms like Facebook where the user can immediately be retargeted based on their score and segment.
Here’s one of my favorite podcasts from Digital Analytics Power Hour. Michael Helbling, Tim Wilson, Moe Kiss and I have a great time talking through natural language processing and how it’s being used in marketing AI today.