Further Chatting about Chat

So, some may wonder what I was clarifying with my earlier post today, just as some no doubt wondered what was the point of this earlier post regarding "lore."

I feel that much of the discussion surrounding what chatGPT is, or what it does, reflects sentiments against the idea itself.  Much like the way that television media discusses the ills and evils of social media or the addictive internet (its most devastating competitor for eyeballs), writers of magazines, both on and off line, have been waging a war for a year now to sow fear about and resistance against Chat.  I've seen Chat accused of making people stupider, destroy privacy, destroy facts, take jobs from people, wreck education, cause real-world harm and no doubt initiate the rapture.  These articles are being written by people increasingly working in an industry that's progressively embracing Chat (an industry I'm part of) because in truth, it's fantastically useful in the hands of someone who knows what he or she is doing.  And because, yes, it's going to destroy thousands of jobs held by people who have been writing magazine-level dreck all their lives (and are unable to write anything better).

Once upon a time, I used to write for a real estate magazine.  I wrote three articles a month, about 650 words an article, which paid me $450.  Sometimes, I was asked to write a fourth or a fifth article for an associate magazine of the one I usually worked for, because the umbrella company ran about fifty different publication "fronts" all over western Canada.

These magazines were free, being available at stands located on downtown corners and LRT stations, in strip malls and sports centres and outside convenience stores.  They were essentially rags designed to sell advertising ... and for the most part, the writing was precisely the level that I see pushed at me by articles arising out of google's algorithm.  I had this side job for 7 years, during which time I produced some of the worst writing I'd ever produced.

The three articles were as follows: (a) an article on some element of real estate, which I was to explain for the uninitiated; (b) an article on the real estate market, which had to look like economic reporting but in fact could only come to the conclusion that the market was great and was on the rise; and (c) an article highlighting some neighbourhood in Calgary where people might want to live.  And obviously, every neighbourhood was wonderful, just wonderful.

My deadline would be the Tuesday before publication, which was the Friday after.  I succeeded in that about 19 times out of 20.  Sometimes, I would write an email to ask for another day to finish one of the articles.  The only reason I was ever late was because I hated writing these articles; ugh.  And now and then, I couldn't force myself to do it when I had to.  In all honesty, I both researched and wrote the articles on the same day they were due.   On average, research and writing together, for one article, took between 25 and 40 minutes.  I'd get up early Tuesday, sit down in a state of misery and slam out three articles in under two hours.

After a year, much to my surprise, I found that the editors loved me.  And I mean LOVE.  Because I was on time (other writers were turning in their stuff late Thursday night or Friday morning, though their deadline and mine were the same) and because I would throw in little bits of dry humour or an odd framing of a point that had paying clients congratulating the company on my being such a "great" writer.  But I've been honest with all of you here; I was not a "great" writer.  I was a terrible writer, basically cribbing details off the internet and then rewording them to fit the requirements.

If I had chatGPT then, like I have it now, I could have written all three articles in about 25 minutes.  Not because chatGPT would have written the articles for me, but because and I'd have had to do was to take Chat's awful writing and edit it out at my normal writing speed, about 65 words a minute.  No research would have been necessary, because it would all have been there in the program.

I promise all of you here, if you're reading an article written by someone for the internet, he or she is copying that article from a slightly worse effort by chatGPT.

Many people are scanning down through this post because they think I've chosen to write nothing of consequence, but the truth is that Chat is as an important technological advance as the internet itself.  This won't be understood by most people, partly because it's being touted that Chat is "writing" the content.  It isn't.  The content is a distilled mass of foregoing content, often in the hands of someone like me that can write and isn't concerned with whether or not it's "morally right" to use this tool.  For the last 20 years, every writer in the world has been stuck either with google or wikipedia when seeking to research anything; and while the usefulness of either vastly dwarfs the library where I used to sit and research freelance articles I wrote in the 1990s, these tools have been made obsolete — not only because the answers one gets are more to the point and reflective of what's actually wanted by a question, but also because each question asked is kept in Chat's memory, so that future questions can be asked IN CONTEXT.

Once this is grasped, that I can say to chatGPT, "Earlier you said that cream had to be separated in a machine -- what sort of machine is that," without having to go back to the beginning of every question, it's easy to understand what a leap forward this is.  I can maintain a single dialogue about oil fields in Indonesia that runs 50 or 60 thousand words, including text that I've punched in from other sites that Chat also remembers, all in context, so that I can question and pick out and have chat evaluate constantly.  Whereas the company I work for had some bad experiences with Chat last year, they've realised that yes, all their writers can produce about four times as much work each month in relatively less time.  I'm not discouraged from using Chat; I'm encouraged.  And everyone I know whose working in any field associated with writing right now will same the same — privately.  Publicly, this isn't supposed to be common knowledge.  Publicly, the narrative is being run by those companies who are directly threatened by Chat.  Publicly, the general population, who know nothing about writing, or how to do research, or anything about where words actually come from, are being told, "There's nothing to see here, move on, move on."  Which they are doing gratefully.

When I had that conversation about "lore," I wasn't looking for an answer to the question, "what is lore," but rather to expose that the whole matter and subject of lore is really just propaganda being used to sell company products, especially those products that have become stale and old as the years have passed. But if we repackage those products as "lore," this gives them a specific specialised flavour that encourages "true believers" to invest themselves towards the whole cannon, not just the works they've happened to buy. It's the same strategy we're seeing with Marvel and many other products, whose "shout out" mechanism for past glories pushes the stay-invested model.

Chat's answers, as the short back-and-forth I published earlier today explained, were dredged out of the industry's garbage-writing "lore" of a different kind ... the endless propagandistic dredge of reasons why we should all keep buying D&D products and thinking about the game in such and such terms, feeding the fanboy game player's need to "keep on top of what's happening" all the time.  Asking Chat these questions is much like those conversations I used to have on boards like RPGNet many, many years ago, where I'd spend time trying to corner some true believer into admitting that character backgrounds or game balance was really just a lot of bullshit.  The difference is that Chat won't move the goal posts; it won't derail the argument into some other discussion; it won't devolve into name calling or mockery or attempts to get others to dogpile as a means of winning the argument.  It'll try, as best it can, to argue the point on its merits ... which can be fun, as with something like "lore," there are no merits.

Some will say that you can make Chat say anything you want.  That you can argue it in a circle until it spits out whatever answer you think it ought to spit.  I don't think I asked leading questions in my example.  It's fairly easy to see when Chat got stumped; it helplessly repeated itself because there was no other answer except to repeat the propaganda.  This is how propaganda always is.  The only reason it "works" in human context is by those strategies I just named.  Proponents of sexism, racism, gun violence, abortion, whatever we might name, succeed in their rhetoric by moving goal posts, gish galloping, derailing, appealing to authorities, half-truths, name calling or any number of other deceptions.  This is what intelligent people do to maintain their biases.

Chat isn't intelligent.  It doesn't think at all.  It's bound by what it knows.  And what it knows is tons and tons of both good stuff and crap that's been dumped on the internet these last 30 years.  Digging into that is like putting on gloves to go dumpster diving.  Messy, full of possible sharp objects, not nice smelling ... but at the same time, people throw out a phenomenal number of interesting things.

I think it's interesting to pluck them out and turn them this way and that.