As I’ve said, my first book for Business Expert Press was published in August of 2015 and was entitled Preventing Litigation: An Early Warning System to Get Big Value Out of Big Data. I didn’t see the hint of AI and deep learning until I was nearly finished with the editing process.
Previously, I had learned to “follow the money.” So, for example, I noted that, in 2014, Google had acquired DeepMind (for something like $650 million) and that MetaMind had been launched in December that same year with an $8 million investment by Marc Benioff, the founder, chairman and co-CEO of Salesforce and by Vinod Khosla, one of the founders of Sun Microsystems and a renowned Silicon Valley investor.
The “big picture” was that venture capital firms had invested $310 million in AI startups in 2014, a 20x increase from the $15 million that startups had attracted in 2010.
I could see an even bigger future for deep learning, and I said so in Preventing Litigation, on page 197. I wanted some of the sand on that beach, but I had no idea then that I would later create some of it.
After Preventing Litigation was published, I paid attention to MetaMind. I reached out to its founder, Richard Socher, who had received his Ph.D. from Stanford in 2014, and asked what I needed to address my litigation topic. He replied: “Labeled data.” And, as I have mentioned, that was the only hint I needed because, having been a litigator and as I will explain, I knew where I could go for the “labeled data.”
But besides the insights of where to go for the data and what I could do with it, I already knew from my research for Preventing Litigation that what I wanted to build was worth my time and trouble to build it. The business case involved a software system that would lead to less litigation. I was not interested in more efficient ways of managing or conducting litigation that had already been filed. I knew that litigation was expensive, and a drain on net profit. I believed that a technology to enable “less litigation” would be valuable.
The important point here, which doesn’t have anything to do with deep learning, is that, in the first nine chapters of Preventing Litigation, I had constructed the business case for an early warning system that was predictive and preventive in nature. I knew in advance that there was a strong business case for avoiding litigation.
In early January of 2016, I attended a Deep Learning Summit in San Francisco. I had seen that Richard Socher was going to be a speaker. I was eager to get his reaction to the size of the problem I was addressing and to an early version of a User Interface (UI). Our conversation was brief but instructive. Dr. Socher was amazed by the size of the problem ($160 billion per year based on data for 2001–2010) and his reaction to the UI I showed him was “You have to build this.”
Now let’s fast-forward by three years, to January of 2019. On January 31, 2019, Brian Peccarelli, the Chief Operating Officer of Thomson Reuters, noted in a blog post about Davos 2019 that the superstar of the conference—the topic that nearly everyone was talking about—was not a celebrity or a public figure. It was AI. Mr. Peccarelli’s take-away was “[a]t last, we’re seeing this transformative technology beginning to unlock its true potential, and we’re also seeing ways it can help address some of the biggest problems of society.”1
I want to say something about AI, the acronym for Artificial Intelligence. AI is very broad term. As the term is used, it’s often shorthand for one of the subcategories that’s being discussed. Machine Learning is a subset of AI; Neural Networks are a subset of Machine Learning; and Multi-Layer Neural Networks are a subset of Neural Networks.
The shorthand for a Multi-layer Neural Network is deep learning.
Unfortunately, deep learning is often then erroneously referred to as AI.
Is it true that AI will change your future? Yes and no. My view is that deep learning is already shaping the future, and that it will continue to do so.
That makes the history of deep learning worth noting, but I’m not going to recount the history for you here. That’s a look in the rear-view mirror, and not a way to look at (or invent) the many ways that deep learning will take us forward.
But you can search the Internet for “deep learning history” and find several links. One post I like is “A History of Deep Learning,” by Andrew Fogg, dated May 30, 2018.2 Here’s another example: Dataversity’s “A Brief History of Deep Learning,” by Keith D. Foote, dated February 7, 2017.3
Why is deep learning such a force for now and into the foreseeable future? Because it’s software that has enabled computers to “learn.”
Notes
1 Peccarelli (2019).
2 Fogg (2018).
3 Foote (2017).