So far I have fenced my own playground around a few simple cartoons, graphs and inquiries into simple facts. Without doubt, the products have on occasion been superior in several ways to what some of my former students and clients could have created for themselves.
Simple cartoons are already equal to what the average commercial art program's students produce, though they are limited to the styles already extant on the Web and exhibit none of the individuality of great cartoonists. The best use file formats allowing them to be scaled for multiple output media.
Graphs based on mathematical equations are for the most part accurate. Rendering is consistent, allowing the visual presentation of complex relationships in two or three dimensions based on tables of data that can themselves be called forth by textual inquiries.
So far these results imply no threat, except to the least competent of artists and graphics technicians. That leaves the formulation of inquiries and statements of problems as the probable core of any issues we might have with the beast(s) in the machine.
I asked Perplexity.ai "what is the variability in price of three-quarter inch ball bearings in the USA from 1950 to the present?" The system answered that it did not have enough information on the specific values required to construct an answer. So it gets an A for humility.
It went on to list some specific prices at which ball bearings had been sold within the overall time frame. And it provided a bibliography and a set of related inquiries.
I next requested as follows: "Produce a graph of ball bearing prices in the past seven decades." Perplexity replied as follows.
"The (following) graph illustrates the Producer Price Index (PPI) for ball bearings in the United States from 1983 to 2024, with June 1983 as the base year (index = 100). It shows a steady increase in prices over the period: (Abridging…) 1983: Index value starts at 100… selected years between, then …2024: Peaks at 292.218. (Graph adjacent.)
This trend reflects nearly a threefold increase in ball bearing prices over four decades, driven by factors like inflation, raw material costs, manufacturing advancements, and market demand. While specific data for the period before 1983 is unavailable, similar long-term upward trends are likely due to economic growth and industrial expansion."
Clearly, given the lag time of less than ten seconds, the result is better than any college professor would expect from sending his star pupil to the library to spend the weekend perusing the card catalog and old issues of Ball Bearings Weekly for the last 50 years, write the report and return on Monday. One notes that the time frame was not quite as specified, but what the heck?
AI has already earned an A at most colleges. I’ve known many teachers who could not IMHO have recognized the source as other than a random sophomore.
My real question is not about AI’s ability to access vast amounts of information rapidly and usefully. I’m now wondering what use the AI system(s) might make of my inquiry.
If it is reported to me and I just embed it — with documentation of course — into my term paper I find it to be no more threatening than the aforementioned cards and magazines. However, what other uses might it find for my question and its reply?
Imagine now that next semester another student asks the same question. Will AI have stored the reply and find that "others" are interested in this subject and expand its radius of search? Will a slight variation in the formulation of the query change the result, factually or syntactically, creating a new line of analysis? Will future inquiries into related subject matter be biased?
We can be sure of one thing: You now know more about the price of ball bearings than you ever wished to know.
The other thing you know is that you can ascertain the price (value?) of any random commodity in seconds, making you the equivalent of the head purchasing agent for your machine production company and seriously threatening the job of said agent, whose years of experience just flew out the window.
I’ve been creating "content boxes" on line for over 25 years, display spaces to be filled in by clients’ specialists who supposedly know more about their bosses business than the average person. I’m fairly sure they don’t know more than all those specialists working for all the bosses at all the competitors’ businesses, nor can they write as cogently and all in seconds.
The few short lines I promised have grown past two pages, so I’ll stop here, but the simple questions suggested about the future of AI pale beside those being asked by those who would use it, or abuse it.
—RC
