SRK Consulting

There are three types of CIOs: those who are convinced that AI is an urgent priority, those who will soon realize they need to make AI an urgent priority, and those who will be out of work in two years.

For those in the latter two categories, the world changed on November 30, 2022 with the release of chatGPT which immediately became the fastest growing platform in the history of the Internet. Its success is not due to enormous technological advancement but merely because the latest in natural language processing tools became something anyone could use. No subscription. No API. Just ask a question and get an answer. And then follow up with a clarification. And then get a better answer.

There are thousands of articles describing chatGPT so I’ll focus on the impact of large language models (LLMs) rather than on the technology. And by “impact”, I’m convinced that LLMs will introduce greater change, faster, than did the Internet itself.

To start with, a word about knowledge. Knowledge can be broadly categorized as:

● explicit (documented, such as in a textbook or manual),

● implicit (stored in someone’s head and not documented), or

● tacit (“we know more than we can tell” – a good example is “I know how to tie my shoes, but I don’t know how to explain it”).

An LLM trained on the Internet is a repository of all the explicit knowledge available publicly. ChatGPT and other pre-trained transformers with generative capability have been legitimately criticized as dangerous “stochastic parrots” that simply regurgitate without insight anything that has been published. It’s a valid criticism of the limitations of the tool; however, the impact of the tool remains disruptively, transformatively potent due to one simple fact: most business activity is deliberately not novel.

 ​I encourage you to contemplate how much of what you do is a creative effort. Consider how much of your work product is sui generis, unique, novel, developed by you and you alone. 

Whether we want to admit it or not, most of what every knowledge worker does involves regurgitation of someone else’s prior art. Very little of knowledge work is truly creative. Except for endeavours where creativity is the actual product, we deliberately avoid creative effort because “creative” solutions carry higher risk than proven, established solutions with a body of knowledge and human resource to support them.

Most knowledge work is intended to produce the least risky result which means using that solution which is most explicit. And explicit solutions are accessible to and reproducible by the pre-trained generative “stochastic parrot.”

“AI will not replace you. A person using AI will.”

I encourage you to contemplate how much of what you do is creative effort. Consider how much of your work product is sui generis, unique, novel, developed by you and you alone. And then consider that the work which is best supported by documentation, established practice, and a body of knowledge is the work which is most susceptible to being accelerated, simplified, and ultimately automated by the emerging stochastic parrots. And then, if you’re me, you go learn everything you can about parrots.