A post has been floating around various wargaming related groups. I’d be surprised if you find me give you the first to link to it, but here it is anyway.
In the article, a wargaming blogger “interviews” the popular and ever-growing-more-so ChatGPT about wargames. Read it directly to get the idea. If you’ve not been reading about ChatGPT, well, then read on to the bottom of the interview where the author asks ChatGPT to explain ChatGPT. Or you might want to read many of the thousands of articles written by humans.
Comments responding to the above mostly discuss the failure of this AI technology to do what it is apparently doing – write intelligent-sounding prose conforming to fairly specific requests. One comment (not on the original article, but elsewhere) talks about how a ChatGPT essay may sound insightful on an unfamiliar subject, but if it is discussing a topic you know well, its clear lack of understanding will be obvious.
Now, I’ve noticed the same thing about newspaper event coverage, but that’s probably an entirely different post.
I’ve also seen commenters talking about how they are using ChatGPT in gaming. Role-playing games with extensive exposition is one obvious use but there is also a comment about using ChatGPT to help apply (board) game mechanics to a new game design. Let us, however, stick to the abilities of ChatGPT as a replacement for the wargaming blogger. As a replacement for yours truly.
Reading its answers above, I have to admit that it is not that far off in theme, style, and substance from some of my own posts. Veteran wargamers (I don’t think it’s just me) have a tendency to reminisce about vague “reminders of a past decade.” As these come to mind, and flow forth through my keyboard, I often use a quick on-line search to get publication dates, development arcs, and the like. More often than not, I’m using a Wikipedia or other expedient source. At best, I pull in one or two other genuinely-referenceable sources but, in no case, do I hold my self to the level of, say, an academic or a reporter. I don’t know how informative this method is to the casual reader, but it helps me to organize my thinking – I can take these memories from impressions closer toward genuine insight.
Many a time I’ve googled-up “top games of the Korean war era” or the like. If that search serves up Mafia II, rest assured that by the time it makes my top five list, I’ll have struck (what I believe to be) the irrelevant game or games from the list. ChatGPT doesn’t have that domain-specific (as we say in the AI biz) understanding and, consequently, can appear foolish – clearly opining upon things it doesn’t understand.
But how different are we? How different am I? I may know a little bit more about wargames but ChatGPT’s ability to comb through thousands-upon-thousands of references surely beats my one-or-two search engine queries. Add to that, its organization and its prose might actually be superior to mine. I have no doubt that a hybrid approach, having an AI generate a first draft which then edit/expand, would produce better-written material.
But is that worthwhile*?
No matter how intelligent a chat bot may appear, it is but regurgitating variations upon what has come before. I’d like to think that when I write, I’m saying something that adds value to the world, even if that value is perceived only by myself. There is additional a value realized through the act of creating. I write something every day because it makes me a better writer, even if what I have written that day is forgettable and thus quickly to be forgotten.
As I, today, write this post ChatGPT is barely more than two months old (and a mere 10 days if you count from the January 30th “stable release”). I think the next few months and, especially, years will teach us a lot about what we are seeing. Already, I’ve seen an article in the Wall Street Journal about the end of essay-based university instruction, perhaps even university grading as we know it. One blog post suggested this technology is about ready to supplant the entire porn industry. I’ve seen Facebook heavily pushing on me young and beautiful looking chatbots, suggesting they can better fulfill all my social needs. For some time now, AI techniques have generated financial news and the like and even the current quality of ChatGPT output is sufficient to handle most commercial and political press releases**.
It will be interesting to look back on this moment a few years from now to see if we really are seeing something significant or just another stepping stone in the progress of information technology.
Game pairing: ELIZA
*On top of everything else, I’m sure ChatGTP could avoid the tidal wave of typos and misspellings that seem to plague my writing, no matter how carefully a proof my work before hitting “Publish.”
**Since I can’t help but complain about today’s media, there is an increasing trend – especially among smaller news organizations – to print press releases as news. Are we at the point, currently, where most of “news site” content could be autogenerated? Might that actually improve upon it?