Text is a representation of human thought. This may look like text phenomenologically, but it has no meaning except when it’s filtered by a human. There is a deep ontological flaw with these applications.
The AI does not have to understand words and objects in the same way as you do to have real world use cases.
You can also see faces on clouds, but is that you or is it the cloud representing a face?
People used the clouds to tell the future in the past, so there certainly was a value to reading them. Maybe GPT-3 is the next best thing to do the same, maybe it's more, time will tell.
The fact that a combination of letters generated by a Bayesian filter makes sense is just a coincidence. It just passes our “makes sense” filter which allow our brains to differentiate signal from noise. But it doesn’t represent anything.