The Luddites and AI

In his speech given at the National Cathedral, March 31, 1968, Dr. Martin Luther King Jr. said:

”the arc of the moral universe is long but it bends toward justice”

That was certainly true for slavery and racial injustice and hopefully it will be true also for the war in Ukraine. The question is if it also holds water for the development of AI and its ramifications for jobs. Continue to read to see what I think. But first a glance in the history book.?

The Luddites

One group that often is ridiculed for ending on the wrong side of history is the Luddites. The Luddites were a group of English textile workers in the early 19th century who protested against the mechanization and automation of the textile industry during the early industrial revolution. The Luddites took their name from the legendary weaver Ned Ludd and they became a strong and highly organized, sometimes violent, movement in the early childhood of industrialization England. On a shallow level, the luddites were afraid that the machines would take their jobs. They felt that the advancements in technology posed a threat to their traditional way of living. In reaction to this, they destroyed textile machines as a protest. Eventually, the upheaval was put down by a 12.000 men strong government army. The historic events of the Luddites created the modern term “Luddite”, today used as a description on someone (hopelessly) opposing technical development. Not seldom accompanied with a smug smile. In retrospect, the Luddites seem to be wrong. The machines did not create huge unemployment. For every new efficiency creating invention of the industrialization that made jobs redundant, new customer needs and business sectors emerged, leading to no lack of job opportunity ever since. Modern economists call this the Luddite Fallacy. At a first glance, the Luddites ended on the wrong side of the arc of history.

However, digging deeper in the arguments for the rage of the Luddites bring nuance to the picture. They were not only afraid to personally lose their jobs. No, they had concern about the ramification on general craftsmanship. Here is how they argue. When the industry owners introduced machines to weave the fabric, these machines could be operated by less skilled labor. Hence, the long tradition of apprenticeship and training from skilled artisans to younger novices, which built both relations and communities, were threatened. The Luddites argued that this would over time lead to loss of craftsmanship and eventually lower quality products i.e. more volume at poorer quality. Or to put it more blunt - more crap to the people. In this sense, the luddites foresaw the downward spiral of consumism manifested by the useless plastic toys that were handed out with a McDonalds Happy meal.

Are the Hollywood writers modern Luddites?

Today, we have a modern version of a Luddite conflict going on. It is the Hollywood writers going on strike in reaction to AI being used in script writing of movies and streaming shows. One argument is of course that they are afraid to lose their jobs, but the case is more made from the angle of reduced quality and miss out of training of the new junior writhers in the business. They fear is that creative story telling will eventually be a lost skill.

Degenerative AI

When ChatGPT4 hit the world in Marsh 2023, the reactions and debate skyrocketed. The large language models (LLM), based on neural networks trained on all the text on internet, both surprised and impressed users around the globe. So much that ChatGPT attracted more than 1 million users in 5 days and now has more than 100 million. The LLM’s are one type of generative AI alongside with similar tools for image and music creation. The debate in AI communities has always been when we will se general AI aka AGI (Artificial General Intelligence). In Marsh, the AI researchers disagreed on many things concerning ChatGPT however, they agreed that in an instance the world had taken one big leap closer to AGI. When you realize that LLM:s are only trained to pick the most likely word coming after the previously presented sequence of words, it is mindboggling to experience how thoughtful and reasonable a conversation with ChatGPT is. And many are we who have started to use LLMs’ in our daily work. One favorite use case for me is to reply cold and pushy sales mails to my work inbox. Instead of ghosting or replying impolitely short, I copy the mail into ChatGPT and ask for a polite decline reply to the invitation with reference to a busy time schedule. Out spits the LLM a 10-line polite answer that I can snip into my reply mail. 10 seconds of work. Easy-peasy not to behave like an AH. Not seldom do I get a surprised thank-you-mail in return.

But it is also here the trap with LLMs’ are set. Happy for my smart professional use of ChatGPT, and the happy thank-you mail, I realized that I could not be sure that not also the reply mail was auto generated. Most certainly it was, as well as the first pushy sales mail. Here we were, ping-ponging autogenerated mail replies. Right now, just like in the case with my mail reply, the internet is flooded with auto-generated text in a pace probably exceeding the human generated material. ?So far, Chat GPT is trained on internet data created up to 2021, therefore mainly by humans. However, next models will be trained on fresher and bigger data sets contaminated with autogenerated texts. With the pace of AI-development and chatbots, the human-generated material will gradually be depleted. The fear is that the quality of chatbot responses will degenerate as the training data becomes more and more auto generated. Instead of generative AI we got degenerative AI.

To demonstrate both the strength and weakness of LLM Chatbots I asked ChatGPT to tell me a joke about Luddites and AI. Here is what I got:

Why did the Luddite refuse to use AI in their kitchen?

Because they were afraid it would turn their toaster into a "toas-terminator" and their blender into a "blender-skynet"!

?

Not very funny.

So how is it goanna be? Will LLM chatbots be the time that the Luddites are right? Will depletion of training sets for our ever bigger more powerful Chatbots eventually lead to degrading quality and stupidity like the Hollywood writers claim? Maybe I have the answer in my own wedding.

What I wore for my wedding 30 years ago

This year is my 30st wedding anniversary. I was 23 and my wife was younger. We were both poor student but didn’t let that to stand in the way for the right decision. Hence, we decided to get married but really keep the cost of the weeding down. One cost that came lower on the priority list than wedding ring and cake was my outfit. At that time, tuxedo was the trend. Luckily, we had a tuxedo I could use in the family. It was my grandfather’s tuxedo, made by a tailor in Kristinehamn 1939. It is a fantastic piece of art where you can see every handmade stitch in the lining.

My grandfather and I were of about the same size, so I could use it without modification (although the size of the 30ies trousers can’t be described as slim fit). I still have that tuxedo hanging in my wardrobe. And here is the question:

What mass-produced clothing you can buy today would you think your grandchildren would wear on their wedding? I guess we don’t have that type of craftsmanship left today. So, in that sense, I recon the Luddites were right. In some sense, the introduction of machines in textile came to a cost of lost quality. I fear that the same will be true for AI.

PS. This text is not generated by an LLM. The only dataset used is what happened to stick in my brain. DS.

?