Tuesday, March 19, 2024

I’ve been thinking about the evolution of our work as it relates to the maturation of information technologies. Which doesn’t make me at all unique; much of the world is discussing the topic given the introduction of large language models (LLMs) of artificial intelligence (AI). I’ve been listening to tech reporter Kara Swisher’s new book, “Burn Book.” If you don’t know Kara, she was in Silicon Valley from pretty much the beginning. Her insights are illuminating, all these years later. Notably she states, “When you invent the ship, you also invent the shipwreck. When you invent the plane, you also invent the plane crash. When you invent the car, you also invent the car crash.”

Some of you already employ AI against your own data sets, meaning you “train” the AI on the data you own, give it decision parameters, teach it over time what is true and what is false, then eventually it learns to do some of the more menial tasks associated with your production work.

You are also a consumer of AI employed for your benefit. If you make an unusual purchase with a debit or credit card and immediately receive a notice that the purchase is flagged due to it falling outside of your normal purchasing patterns, you’ve likely been grateful that AI works quickly to inform you of a potential fraud hazard. Mere humans simply can’t sift through large data sets nearly as quickly or accurately as can a computer.

No doubt you’ve heard about how AI will change healthcare, when medical imagining will provide much earlier diagnoses with greater reliability than the human brain, which, of course, is prone to tiredness, mistakes and well, just being human.

Each of these examples you would probably file in a container labeled, “Positive”. As in, these are generally accepted as beneficial advancements. Yet just as the day you first held a smart phone in your hand you could never have imagined Uber, there are far greater implications that most of us aren’t yet thinking about. But someone is.

At October Research, we reside in the world of publishing. You may have noticed that many publishers, be they print media, literary publishers, music producers, individual authors, individual actors and others have filed suit against the largest of the LLM providers. They allege that the LLMs were trained on copyrighted material and by unauthorized use of the assets belonging to the plaintiffs. The plaintiffs allege their intellectual property is used as a discreet ingredient and added to the recipe in first making, then baking, the AI cake. The rub is, after a cake has been mixed and baked, how do you extract out a single egg from the finished cake? You probably can’t. So, it might all come down to a price paid retrospectively for the data skimmed from the internet, mixed into cake batter, and slid into the oven that produces an LLM AI tool.

How do these original producers of information remain in business (let alone remain profitable) if all their work is vacuumed up and assimilated into a cake served on a buffet belonging to a large LLM provider?

I don’t know. But it’s an important question and one I’ll ask you. When your transaction data is vacuumed up outside of your shop and used to train an AI product that consolidates what you do into a single algorithm, how will you remain in business (let alone remain profitable)?

If you’re in the title insurance business, your previous response might have incorporated the formerly valid point that at the end of the day, title policies still need issued. And human involvement, therefore, will remain a critical aspect of the work. But, what if there were an accepted push to delete the requirement for issuance of an insurance policy? Would that change your answer?

I don’t know where any of this lands at the end of any given day. I do know that in 2010 taxi drivers and taxicab company owners thought that Uber was only going to disrupt the black car segment of the market. They believed that most riders would continue to use traditional services. They certainly never imagined Uber Eats, Uber Rental Cars, Uber Courier, Uber Fleet or Uber Charter.

You likely didn’t realize that all those times you were “proving you weren’t a robot” by identifying cross walks, buses, house numbers, motorcycles and street signs, you were actually involuntarily providing AI training for autonomous driving cars, but you were.

By all means, employ AI to peruse and harness your own data for the benefit of efficiency. I’m all for it. And as you do, make sure you retain the ownership of your data. Or be ready to be the former taxi driver, whose only viable option is to hail an Uber. Don’t be a cake left out in the rain.

Until next time,

Mary Schuster
Chief Knowledge Officer
October Research, LLC