CHAPTER
5

AI Is The Exponential Technology Of This Decade

As I discuss in Become an AI company in 90 days, AI is a broad, revolutionary technology like electricity, computers, and the internet. The internet provides a great recent case study on what to expect from an exponential technology.

Because discussions about AI can quickly devolve into academic debates about terminology, let’s sort out some definitions we established in our book.

  • AI is a general term for “intelligent software.”
  • Machine learning is a type of AI.
  • Deep learning is a type of machine learning and is state-of-the-art AI.
  • Natural-language processing (NLP) is an AI product pattern, a practical application of AI that solves common business problems. Most modern NLP applications are built by using deep learning.

I’ll use NLP as an example to show AI’s progress through the three phases of exponential technologies. NLP makes a good proxy for AI’s evolution for a few reasons:

  • People create natural language data for the benefit of people. As a product of AI, illustrates how computers can take over a distinctly human activity.
  • Like AI in general, NLP will have a massive impact on most companies. Many companies dedicate a significant portion of their workforce to processing unstructured data in contracts, emails, and customer conversations. Automating this work will shift the operational cost and speed curves of every industry.
  • Prolego has deployed a number of cutting-edge NLP applications, so I can speak specifically and authoritatively about how big companies are using it.

EVIDENCE THAT NLP IS AN EXPONENTIAL TECHNOLOGY

The most direct evidence of NLP adoption would be a dataset of the number of NLP applications deployed each year. Unfortunately, no such dataset exists. So instead I will use correlated data as evidence that NLP is an exponential technology.

Because research is a leading indicator of technology progress, any exponential AI technology should first demonstrate growing traction by research funding. Figures 7 and 8 show how research funding for NLP has grown over the past 10 years. Both scholarly references and papers are showing exponential growth.

A more direct measure of AI adoption is the rate at which deep learning NLP models are being used in applications. Hugging Face runs a popular GitHub open-source repository that NLP developers use to build applications fast. We analyzed the rate of model updates in the repository over the past 18 months. Although the survey period is short, the adoption rate shows signs of exponential growth.

Figure 7: Search results for “deep learning NLP” on Google Scholar. Because academic research is a leadingindicator of technology applications, we can assume that NLP applications will follow a similar exponentialadoption curve later. Source: Google Scholar
Figure 8: Number of NLP research papers published during the last decade. Publication of research papers isanother leading indicator that NLP is an exponential technology. Source: NLPExplorer papers

A more direct measure of AI adoption is the rate at which deep learning NLP models are being used in applications. Hugging Face runs a popular GitHub open-source repository that NLP developers use to build applications fast. We analyzed the rate of model updates in the repository over the past 18 months. Although the survey period is short, the adoption rate shows signs of exponential growth.

Figure 9: Trailing three-month average of monthly updates to NLP models in Hugging Face’s GitHubrepository. The rapid growth in updates is a function of more developers using the library, more NLPmodels, and more frequent software updates. Source: Hugging Face

Let’s Future Proof Your Business.