I Have Spread a Lie

You can find my amplification of misinformation in my What I Am Reading March 9^th, when I share an article by Tyler Cowen, a very informative economist at George Mason University. I would share the post with you, but it has been taken down, but here is the false quote:

“But these three [inventions], perhaps, have fallen out by a certain fatality or providence of such a kind, that though they have added much to human power, they have not much increased human goodness; nay, rather, the first and last have furnished men with the means of doing more mischief, and the please say more second has made them more vain and arrogant.”

I found this particular quote interesting because it speaks to a problem with today’s information technology, social media. It purportedly describes an older technology that allowed words to scale, the printing press. As it turns out, Matthew Ingram, who obviously vets his material a bit better than I, had posted a link to the same article by Cowen, but subsequently discovered the fraudulent nature of the quotation and wrote about it here. That led to additional concerns addressed by Maria Bastillos in The Nation.

Dr. Cowen and many others, including us at ACSH, have been very interested in the rise of large language models, ChatGPT, perhaps the most well-known, although other versions are not out and about. These are, in essence, programs that have been fed nearly all of our digitized words and have been able to build very successful statistical models of which words most commonly follow others. They can perform so well that they can create very readable text. As it turns out, they also make things up. Whether you want to call that a lie, or as the researchers refer to it, a “hallucination,” is up to you. I prefer lie. Most people believe that Dr. Cowen used ChatGPT to write that piece.

The whole affair has left me chagrined. I suppose I shouldn’t be; after all, didn’t I write just last week about how those most concerned with misinformation and who firmly believe they can tell a lie from the truth are the most significant sources of misinformation? My bad. But what should I think of Dr. Cowen? He took the article down but nary a word explaining how that might have occurred. Obviously, I must take his future thoughts with a much larger grain of salt. Trust is hard to gain and easy to be lost. I would think a mea culpa is in order; it would undoubtedly go a long way in restoring a trusting relationship.