Remembering Mondays Before The AI Boom
Thank you for subscribing and being part of this growing community. I appreciate all the feedback on the technical strategy series! I will be building more content based on your requests and questions. I am adding answers to the course as well. Thank you for taking the time to share your thoughts with me.
I have new people joining my staff to work on marketing and the logistics of all the content I create. I now have a “formal” bio. It’s written in the 3rd person, which is strange to me. Vin Vashishta is…
The process of writing a 3rd person bio is unexpectedly complex. I had to list my experience and score accomplishments with several metrics. I spent an afternoon looking at myself in the 3rd person.
As odd as it was, it brought up memories of what being in the field was like before there was interest in the field. The first myth of data science is that the field is new. Data scientist was only coined in 2012, but the field existed in several forms before then.
My first data science class was in 1996. I did work with computer vision before AxelNet, and it was primitive. Coding edge detection in C is not my idea of a good time. I have nostalgia, but I am so glad that people have taken the field past the early days.
I graduated after that first wave of enthusiasm collapsed into an AI winter. I thought I was headed straight into Microsoft to do AI research and work towards a Ph.D. They wouldn’t even hire me for a QA role. I connect with people working to break into the field because I remember what that rejection felt like.
In 2007, BI brought data back into the business and me back into the field. My mind was still stuck in the AI paradigm. I wanted to take data to the bleeding edge, but no one in the business was interested. It wasn’t until 2010 that I saw the light.
I worked on a simple analytics project, creating a massive revenue stream for the business. The light bulb came on for me then. The AI winter was a disconnect between what was on the distant horizon and what was possible today.
My focus shifted from academically possible to feasible. I went after my first clients in 2011. My pitch was to marketing teams. They had enough data to build basic customer behavior models. Most businesses could do early experiments with advertising to understand what changes their customers’ “perception of value.”
That phase was my central theme. The causal root of buying behaviors for most products is perception. Why do people buy shirts from Gap vs. American Eagle? Perception is the root of preference when those preferences are subjective.
I didn’t hear the term ‘Data Scientist’ until 2013. My initial pitch to marketers had failed, and I was working on supply chain use cases. They understood data and continuous improvement. Lean and Six Sigma ran deep in the manufacturing space. Using data to improve operations fit with their view.
I was also publishing on business strategy. My first significant publication was Comparative SWOT For Competitive Analysis. SWOT was typically applied to understand the business’s Strengths, Weaknesses, Opportunities, and Threats. I added it to the competitive intelligence toolkit. I explained how to perform a SWOT analysis on competitors and evaluate business strategy based on both sides of the market.
Data and analytics were emerging strengths and threats. My sales pitch began to include the concept of competitive advantage and threats from competitors. I got more clients with that approach. I was selling hope and closing with fear. I had the desired impact on the perception of data’s value.
My second significant realization came from those early clients. Everything I did was an incremental improvement. Every sale I made came from a variation on the themes that the business already understood. Data was a new, better way to accomplish their current objectives.
Maturity was a slow process and what I built had to integrate into existing workflows. More than that, my work had to fit into existing paradigms. Selling advanced use cases was a dead end as well. No one was interested. I sold to their current use cases and explained the benefits vs. an alternative approach.
By 2014, I was getting attention from a larger community. I was one of a few people who had put machine learning models into production. I had a small track record of generating client returns, mostly cost savings. I was brought into a project working with hard sciences researchers.
At that point, I was a software engineering statistical analyst. I was a year away from confronting the challenges of model reliability. I was clueless about the more profound complexity of applied machine learning. My methods were casual, not causal.
The scientists I worked with beat me into a new way of modeling. Science and data were powerful. Machine learning models had their place in the scientific world, but they were supporting cast members. Research methods and validation were required. I was discovering why deep learning models failed across multiple use cases.
Deep learning challenges like double descent and overparameterization are known and managed in domains like Physics. Theoretical Physics had to confront modern modeling challenges in the 1970s. They understood how to learn a function before large datasets were feasible.
The scientists showed me that our field was reinventing aspects of the hard sciences and taking credit for them. Scientists were not fans of our field and still harbor deep resentment today.
On its rise to stardom, data science has made a lot of enemies. C-level leadership feels like technology is running their businesses. More people in the technology world call for CEOs to be replaced with technical strategists. I am one of those who believes it will be challenging to lead a business without technical capabilities in the next 5 years.
However, I don’t want to push them out. It is better to upskill them and create a pipeline of new C-level leaders from the data organization. It’s impossible to forklift the C-suite out and drop technical people in. Maturity is incremental, or it won’t happen.
People like me are also disrupting the way strategy gets done. I advocate for introducing experimental methods and models into decision support systems. Those should replace case studies and the legacy frameworks used by most strategy consultants. I firmly believe strategists must come from the data domains to be successful in the modern competitive landscape.
I take the same approach of upskilling current strategists and building a pipeline for data professionals to enter the field.
One of the exercises for creating my new bio was defining my impact on the field. I got businesses to spend money on data science, and I helped define the role of data scientist as a value generator.
Defining the data scientist role involved pointing out the most valuable use cases and explaining the capabilities required to build solutions. The role quickly became too big for an individual. Data Engineers, Machine Learning Engineers, Data Scientists, Applied Researchers, Data Analysts, Data Librarians, ML Product and Project Managers, ML Platform Engineers, and so many others are now required for a very mature data practice.
We pull from many other domains, but something new is happening. Data science is making novel contributions. The field is almost ready to take its place as a hard sciences discipline. Our research will add to the body of knowledge across domains and bring new capabilities to scientific research. It has been a very long road, and I am privileged to have seen this journey.
Now I am focused on defining a new role, technical strategist, and getting businesses to pay for causal methods. Technical strategy is my Trojan Horse for causal because it makes semi-dominant strategy possible. Companies that can deploy advanced decision support systems will have a significant strength. Those who do not face substantial threats. My sales pitch hasn’t changed much in 8 years.
I will avoid the mistakes I made last time and make new ones. My most significant oversight was not teaching data science in 2016. I should have, even though I was not perfectly qualified. We might have a better curriculum if I had done more than start conversations.
This time I will teach. I learned curriculum and course design in preparation. I’ve been working on honing my teaching skills for almost 15 years, and now I have formal education backing it up. I will also continue to work in the field. My courses will be disconnected from reality if I don’t.
The doors to data science are opening. We are connecting with other scientific disciplines and creating partnerships. We are making inroads into the business and building coalitions across the enterprise. We are moving forward to support new technologies, quantum computing and web3.
The field is maturing incrementally. We have managed to hold on to our academic wing that is moving at lightspeed. We are also building a solid connection to implementations and value creation. Most fields lose one or fail to achieve the other.
The next 10 years are going to be amazing.