Short one this week as I am deeply, deeply disappointed in myself that I did not notice that yesterday’s post was the 666th new post on this little newsletter. I apologize for not recognizing such an important milestone and promise to do better in the future.
More and more media companies are making deals with imitative AI firms to allow them to use said media company’s material to train their models and to sometimes use the imitative AI tools in their offerings. I am genuinely confused by what these firms are doing.
Let us assume that imitative AI will get good enough at producing news and creative content that people will pay to have it create said content for them. Now, I do not think that will happen, honestly. The pace of model improvement has slowed, and it is physically impossible for them to not be bullshit machines — it is baked into how they operate. In addition, they cannot ever produce anything new or interesting because they can only regurgitate based on what their training data says come next. At best, you get a mid-level product based on things in the past. But let assume that’s fine for many people. Why, then, would media firms want to help them?
In the best-case scenario for imitative AI firms, they create a product that produces things that people want to see and read based on their own prompts. that last bit is key — personalization is the only way this even remotely works as a business. Given the outrageous costs of running these systems, they need to replace entire industries to be popular. If you cannot get individuals to pay for them, then they won’t work. If, say, Disney still needs to have people perfect the material imitative AI produces, then it is not likely to be cheaper than using skilled employees at the start given the break-even points of these systems. Cutting Disney or the New York Times out is the one path that seems most likely to lead to profits. But why, then would Disney agree to help their own destruction? I think there are three possible reasons.
One: These firms do not believe that imitative AI is here for the long term and are getting while the getting is good. I’d like to believe this, but given how much hype is surrounding these systems, and how little contact with hard reality many CEOs of the largest firms encounter, and how this is being sold as a people replacer, I don’t think this is likely to be true.
Two: These firms believe that they can use imitative AI to increase their own profits but that it will not ever get to a point where it replaces their firms completely. They believe in the “augment people” promises rather than the “replace people” promises. They figure that these firms will settle at a price point that allows the media firms to make money using their tools, due to efficiencies and firing people. Realistically, that is not likely, but it is a believable story in the C-suite.
Three: Capitalism just doesn’t work anymore. The people running these firms are forced to only care about the next quarter’s earnings call, and since Wall Street is incapable of seeing hype for hype, they have to be seen as participating in the imitative AI boom. The people in charge no longer have any incentive to think about the next quarter, much less the next year, and so they make moves that goose the bottom line even if they are sub-optimal for the long-term health of the company. And, hey, if they facilitate, say, Google AI summaries and thus destroy traffic to their media websites and platforms? Well, they will still have gotten their last quarterly bonus to live on.
Weekly Word Count
3900 — not great but at least some movement on the Encyclopedia Brick novel. Now I just need to decide if that title is terrible or awesome. Well, and finish it, but priorities, people. Priorities.
Have a great weekend, everyone.
Three.