April 3, 2020
Working in AI has changed how I view the world. In traditional software projects I think about solutions in terms of time, scope and resources: What do we want to build? How much time do we have? How big is the budget?
Now I think in terms of models, data, probability, and trends. Tripling the engineering budget won’t help us automate a business process if we don’t have sufficient data coverage. A great data scientist can develop a new mathematical approach which suddenly increases accuracy from 80% to 97% and improves profitability by 5x. It happens every day.
I’m also developing a keener intuition for identifying what I know—and don’t know. In traditional software projects “unknowns” are unacceptable risks to be mitigated through better planning. In AI “unknowns” are just part of the work.
Very few people are comfortable with this level of uncertainty. Fewer still can look at an exponentially-growing threat like the coronavirus and image a probabilistic range of outcomes. Most people are glued to their phones looking for reassurance or predictability. As usual, the scariest stories get the most eyeballs, advertising dollars, and political attention.
Here is the reality: nobody knows how this pandemic will evolve. The world’s smartest, most capable people are struggling to build models which even describe the current situation. We just don’t have the data. We can’t answer basic questions such as, “how many people died from the virus” because:
Other factors like R0, symptoms, or the impact of government policies are even more ambiguous. Experts like Michael Osterholm have spent decades modeling pandemics and can predict a return to normalcy this Summer or a total collapse of supply chains depending on assumptions. We. Just. Don’t. Know.
Fortunately the most unpredictable wildcard is a cause for optimism: the impact of human ingenuity. The world’s smartest, most capable people are collectively racing for a solution. Scientists are sharing data at an unprecedented scale.
In my own narrow circle of friends I’m watching:
We will begin adapting in a million unpredictable ways, all of which are impossible to model. Duke University found a way to clean N95 masks at scale. Our hospitals will begin optimizing for treating waves of infected people.
The virus exposed many fragilities with our globally-connected societies. In the coming months we will see the upside of this highly-connected, economically-optimized world as human innovation is unleashed at scale. It is unpredictable, but it is a cause for optimism.
May 2, 2021
Your goal as an AI leader is to get your teams to think like pros. You want them to strategically look for ways in which AI can lift the entire business instead of just solving a narrowly defined problem. Your team should constantly seek ways to advance the bigger vision of becoming an AI-driven company. In this issue of FeedForward, I’ll describe the difference between how pros and amateurs think about AI.
March 31, 2021
In this video Justin Pounders, Director of Machine Learning and AI Research at Prolego, breaks down natural language generation (NLG) into its most basic components and describes how you can begin building out these components in your business. (And, no, it doesn’t depend on GPT-3!) He describes how NLG depends critically on two questions (WHAT you want to say and HOW you say it), the types of data you can feed into NLG systems, and a development path for being able to summarize multiple sources of data in plain English.
March 30, 2021
Like most engineers, I hate tedious work. That’s why I love the idea of automatic machine learning (AutoML). As much as I want to love AutoML, it’s been incorrectly framed as a substitute for data scientists. This confusion arises from a misunderstanding of what actually happens in machine learning projects.