Since OpenAI’s ChatGPT was released five weeks ago, it’s caused a bit of a stir, not least among education faculties. It’s hard to see how having the ability to generate a moderately convincing and unique essay in under two seconds won’t disrupt education as we know it. Sure, it’s got its limitations, but those limitations are disappearing with each generation of this type of software. For the many blogs, news and information services which are simply platforms for advertising and generating search engine results and click-throughs, this stands to effectively eliminate the cost of generating the content in the first place.
My colleague Mike Hartley last year pointed me at Roald Dahl’s The Great Automatic Grammatizator, in which the industrious engineer Adolph Knipe builds a story writing engine and, given this is the 1950s and there’s no internet, proceeds to corner the market on fiction writing – much to the further detriment of writers still penning their content the old fashioned way. So much so that the story ends with the line:
Give us strength, Oh Lord, to let our children starve.
The Great Automatic Grammatizator – Roald Dahl
While I don’t blog to generate advertising revenue, instead being more interested in the practice of kicking out a short essay on occasion without too much planning or editing, I was certainly interested to see how this tech is to play with. So I duly headed over to ChatGPT, set up an account, and hit it up with some questions. Disappointingly (though maybe not) it wasn’t too interested in generating outright controversies:
It was however more than ready to prove the existence of Santa Claus:
It was also more than ready to assert the opposite when asked as well, showing that for at least some topics you can have it prove up is down and down is up. For topics that are less cultural however it seems disinclined to go this direction. It didn’t stop me trying to push the point, however:
After a bit of a wait while it seemed to have cardiac arrest, it got a bit further…
…and then went out to lunch for a while before cracking on towards its exciting conclusion…
So after its dire warnings about keeping things based on reliable scientific evidence, I thought I’d finish off with asking it to get a little creative. I thought the result was actually very impressive, given that it needed to know what Fermi’s Paradox was in order to get started, and that it structured this to present two perspectives and then draw a conclusion using them:
