As I was writing a different newsletter, a report about an AI comedy special based on George Carlin came across my feed and it inspired me to pivot to a new topic. George Carlin passed away in 2008 and this week a comedy Youtube channel ran old media of him through AI to recreate his voice and create new jokes with an LLM built on his words.
I am not going to talk about the ins and outs ethics and legality of the special. I'll point you to tech journalist Tom Merritt if you want to explore those ideas a little more. What I want to talk about was the reaction of Carlin's daughter and the implications for organizational learning.
Kelly Carlin, George's daughter, did not appreciate this 'impression' and expressed that on Twitter (X). Among her comments about it was a thought that caught my attention because it brings up a concern I have about AI roll out in organizations.
At the end of her comments, Carlin suggested that instead of watching a retread of what her dad might have sounded like, that "how about we give some actual living human comedians a listen to?"
There are thousands of comedians getting on stage in night clubs, going to open mic nights and developing their craft every week. Some of them get a break and get on television. A few of those get a bigger break and get a spotlight show - them, alone on stage for an hour recorded and broadcast. If this AI comedy special becomes a common phenomenon, it will dry up the pipeline for comedy by clogging media with simulations of classically funny performers creating jokes that are kinda-sorta like their old ones, but maybe with a jab at Elon Musk.
AI destroying the pipeline of talent is a problem that I have been worried about since learning about Chat-GPT. In organizations, AI can eliminate a lot of the opportunity to learn how to perform the complex knowledge work that we expect of managers and executives. Most of the rollout of AI that I have seen so far has been focused on taking away the work no one wants to do; speeding up the drudge work that doesn't directly lead to generating revenue and replacing work that would be done be interns and early-career workers.
The benefit of using AI for this kind of work is clear: increased productivity, faster turn-around, potentially lower costs. The challenge is that this work serves a purpose other than just getting things done. While working at the Institute for Research of Learning, Jean Lave and Étienne Wenger researched how learning happened through work. They described the activities essential to developing expertise in a given practice. The earliest activities that someone does as they progress are fall into a category they call legitimate peripheral participation(LPP). LPP is work that is of low risk but important. It is essential and the worker is engaged in and around the product and more advanced workers. It is work that allows you to be engaged with a practice, but does not require you to know it well already. It is the grunt work, the small things.
In the coming years, I am concerned about companies that eliminate a majority of their legitimate peripheral activities by pushing them off to AI. Not just for what they do to the labor force, but for what they do to themselves. There is a chance that organizations will work to streamline and declutter workflow so much that new professionals won't have a chance to start their learning process on the ground floor. And if companies don't foster talent deliberately, they may well end up with a development pipeline that is near empty when it comes time find new managers and executives.
Sadly, I don't have an answer for how to prevent this. I wish I did, because answering the question of how companies develop talent in the age of AI is going to be a lucrative business in the next 5 years.
If you'd like to read more on legitimate peripheral participation, Lave and Wenger wrote a book titled Situated Learning: legitimate peripheral participation. It is more a philosophy and anthropology book than a guidebook, but if you like to geek out about learning like I do, it is an intriguing read.