5 - Some AI landscape updates
The rapid increase in AI is surprising to me
Today, I’m going to take a bit of a different approach. I spent this past week working on a model to calculate expected future returns to extend some concepts in last week’s article. This is, of course, taking longer than I thought it would. In the absence of a data driven analysis for today’s newsletter I’m going to chat through a few things I’ve found interesting over the past few weeks in about half the words I normally do.
You’ve probably already heard of this repackaging of ChatGPT, but there are a couple of interesting implications for this GitHub hobby project that has taken the internet by storm. It is a set of code that allows anyone to run multiple ChatGPT agents with a user defined intention. Basically, you no longer need a human to write the prompts. Another instance of ChatGPT will write the prompts for you. The All in Pod episode 124 has an excellent description and discussion on what this could mean for the future of software developers, and for white collar work in general. I finished that podcast feeling nervous for the future of sales, finance, software development, and content creation.
Granted, history shows us that new technology doesn’t destroy jobs without creating others. Below from McKinsey demonstrates how other sectors of the economy fill in segments that have shrank.
At the end of the day, my suspicion is that critical thinking skills and prudent judgment will separate the wheat from the chaff in the job market going forward, regardless of the industry those employees come from or go to. I wonder if this will increase the acceptance that a single person can apply their skills in multiple industries, and the premium paid on experience will shrink. In any case, right now the cost of creativity is going down. You can generate 1,000 different Twitter posts at a fraction of the cost of paying social media managers to do this work.
Military and police applications
In a scarier extension of the above, the US air force conducted its first flight using AI. On this flight the AI acted as a copilot, fulfilling jobs that a human previously did. This could be a beachhead where continued development of AI will spring forward into deeper and more important military systems. It is true that AI is largely as good as the data they are trained on, but unfortunately the military’s data collection processes are probably pretty expansive. Coupled with a report that the NYPD has adopted robotic K9’s, it’s hard to see how we aren’t inching toward a terminator style future ushered in quickly to increase “safety and quality of life”.
Coupled with the drone warfare we have seen in Ukraine, demonstrating how robotic and data driven warfare can be vastly more effective than alternatives. West Point’s Modern War Institute posted this detailing how drone warfare is changing the battlefield. The biggest point on this list to me is the ability for low budget actors to use this tool. It remains to be seen if these developments are more like the introduction of rifling which temporarily provided one side with an asymmetric advantage until the other side began using the technology, or if this will drastically change war strategy across all dimensions (sea and land) of the battlefield the way the introduction of air power in WWI did. It’s also worth noting that drone production and quality seems to be an area that China dominates the US in at the moment.
What is AI?
I heard a long time ago from an interview with a Google executive that occurred in the late 2000s (that I unfortunately cannot find a link to) which described AI as anything that seems like a new and emerging technology. Their example specifically was Google maps. End users called Google maps an AI when it launched, but under the hood, Google maps could be simplified to a recursive maze algorithm, a concept that computer scientists had been comfortable with for decades. For now I think it’s too early to say what the ratio of bark to bite will be for these technology advancements.