7 of my opinions about AI
I differ from many people.
1. AI will affect productivity, but maybe not in the statistics
The question of whether AI will show up in the productivity statistics has two components.
Will AI have a dramatic effect on how we live?
Will this effect show up in the productivity statistics?
I think that the answer to (1) is probably yes, and the answer to (2) is perhaps no. That is because of the way that the national economic statistics are force-fit into the mold of the GDP factory, which does not capture many of the ways that AI could affect our lives. I doubt that existing economic statistics capture the benefit that people get from having cameras and GPS in their pockets, much less how we will benefit from AI.
In The Measure of Progress, a book that I received while working on this post, Diane Coyle writes,
the all-important figure for GDP, dates from the 1940s when physical capital was the binding constraint on growth in the postwar era, natural resources seemed free, and the pressing economic policy challenge was seen as effective demand management so the Great Depression could never recur. (p. 4)
In 1947 (when modern economic measurement through the SNA was brand new) about half the economy was measurable, by 1990 less than a third, and by 2019 less than a quarter, or more likely only about a fifth. (p. 14)
The central argument of this book is therefore that this 1940s measurement framework, the SNA and other standard economic statistics extending it, is no longer adequate for understanding the economy, and in fact in some ways actively hinders understanding. (p.15)
Mainstream economics is framed around the idea of a production function, in which output is a produced by units of labor and capital. I think that the production function is a plausible approach for describing the manufacturing sector of the late 19th century. But mainstream economics has yet to adapt to 21st-century reality. I have made this case repeatedly.1
2. A Large Language Model’s main value is as an interface, not an encyclopedia
Many people treat AI as an “answer-bot,” along the lines of Google search. Many of the metrics used to evaluate models are based on this thinking. Experts tell us how well the models perform on exams in math or science. Or they tell us how useful (or not) the models are in doing research.
But I do not think that answering queries is the compelling use case for AI. If we start from where Google was in 2023, before ChatGPT, the improvement provided by Large Language models will be less than earth-shaking.
In my opinion, the key characteristic of the new AI is not encyclopedic knowledge. It is its personality. For the first time in the history of computing, we can relate to a computer using human language. We don’t have to learn how to write code or navigate a menu. I think that the economic effect of AI will depend on how far this can be taken.
3. We are starting to see the value of AI in software development.
It seems that if you want to create an app or build the software infrastructure for a simple business, you can do that with AI tools. Right now, only people with traditional software development experience are using AI to write code. But eventually ordinary civilians will do so.
4. We will see AI start to do work that we routinely do using computers.
Recently, I had a conversation with someone about a “Klingbot.” This would scan all of the information sources that I regularly look at, find the most notable news and analysis, and save me the time of sifting through them. Perhaps it would even write blog posts in the vein that I would write them. It could do my routine work much faster, freeing me to spend more time doing “pure thinking.”
I believe that everyone in the laptop class could imagine a bot that does their routine work. If an AI could adapt to any individual and become your personal bot, then that would be very powerful.
I suspect that most people could use AI to filter information for them. Think of a spam filter for everything, not just email. Filter your social media feed, your text messages, and your news sources.
5. AI will be a mentor/coach
I think that the term “AI companion” turns many people off. The press likes to focus on lurid examples. But I think that there is a good chance that for most people the pluses of regular conversations with an AI will outweigh the minuses.
AI companions are gaining consumer traction, even though the current implementations are primitive and despite a lot of cultural pushback against the concept. Going forward, the implementations will get better and the pushback will fade away.
6. AI “clones” will be useful.
People will want to interact with AI clones of celebrities and teachers. They probably will want to interact with AI clones of friends and relatives who are not easily accessible in person, including those who have died.
7. AI will help us to train robots.
I continue to see a lot of potential in the ability to train robots by talking to them using natural language.
The best way for AI to raise GDP would not necessarily improve human welfare.
In my industry AI is used for up-coding, that is finding ways to increase the number of Hierarchical Condition Category codes that a person has. It does this precisely by prompting the doctor during the visit to write up the codes. This increases revenue from MAPD (thus, increases government spending) which "increases" GDP.
A similar dynamic seems to play out in sports gambling (AI attuned to figure out how to manipulate people into betting).
Like a lot of technology I can imagine good or ill uses. But GDP only measures whether a purchase was made or not.
Coincidentally I've been thinking about how GDP (and related measures) is failing to capture perceived wealth. A lot of things seems to fail to be properly captured, but technological improvements seem to be particularly poorly caught.
If I can use an AI to write half a dozen emails that would normally take me 10minutes each that's an hour of my life I've got back (less time taken to read the emails before hitting send). That number totally fails to show up in GDP
Worse, if your case 7 applies and we can use AI to program our lawnmowers, roombas etc. we'll see a diminution in the amount of human provided services in those areas and thus, for those people who pay for gardeners etc. rather than do it themselves, a reduction in calculated GDP because they no longer need to pay Jose, Maria and their friends to do it for them. (Although some of those services are already missed by GDP since they are paid in cash and undeclared by the recipient). But that dimunution is in fact false because the people who used to hire gardeners and cleaners and now have a robot to do it are saving time and money thanks to the new technology. That ought to make them, and the nation, richer not poorer