Garfield And Morrissey Don't Know What They're Talking About

I quite like Mondays. Not a popular opinion, I know. 

It's mainly because Monday hasn't been a regular working day for me for well over a decade, not since I went to see my manager and said I wanted to drop my hours from full time to four days a week and she got out the calendar for the rest of the year and put a line through every Monday. From then on, whatever my working pattern has been - and it's changed more times than I can remember - I've never worked Mondays.

Monday isn't my favorite day, though. I don't have one any more. It used to be Sunday (Suck on it, Morrissey!), a preference dating all the way back to adolescence, when I had classes six days a week (and chapel on Sunday mornings), making Sunday the only day of the week I didn't have school all day. 

Now Sunday is my main work day, so in a weird way Sunday is my Monday, although when you don't have to go to work the day after, it hardly has the same dread impact. Sunday's also my Friday, I guess, seeing it's both the first and last day of my working week. Or it is on the weeks I don't also work Saturday...

Anyone know why I started this? If you do, please let me know in the comments because I certainly don't.

Oh, wait a moment, yes I do. It's coming back to me now. I was going to say here we go with another Friday Grab-Bag on Monday. That was it.

 

Mark This Comment As NOT SPAM. 

Since we've mentioned the comments (We? Are we the Queen now? Or, I guess, the King. Still getting used to that...) this might be a moment to mention that I'm very, very bad at checking Awaiting Moderation. And when I say bad I mean I pretty much never do it at all.

This morning, I just happend to look in there (For reasons.) in what must be the first time for at least a year. I was very suprised to find about half a dozen genuine comments among the endless stream of sexual harrassment and unwarranted praise that prefaces links to websites of extremely dubious provenance. I authorised all of those valid comments, most of which were of fairly recent origin, so if anyone's commented lately and not seen their comment appear, now might be the time to check.

I haven't replied to any of them, for which I apologise. I wasn't smart enough to do it as I was signing them off and now I can't remember which posts they were for. I may go back and try to find them but I probably won't, so feel free to drop another comment and remind me if anything needed an answer.

 

Address Unknown

Whenever Wilhelm does one of those posts about all the peculiar offers and invitations he gets via email thanks to having a blog, I wonder why no-one ever gets in touch with me to promote their product or give them free content for their their platform. I mean, I'm glad they don't but no-one wants to feel like they're invisible. 

Okay, everyone wants to be invisible, at least sometimes but that's an entirely different fantasy.

It so happens that this morning, in the course of fact-checking for the item above, I discovered that contrary to my belief, this blog has never in all the dozen years of its existence had an email address. Well, not one that both worked and that anyone could find. Believe it or not  - and if you're a regular reader I'm sure you will have no difficulty believing it whatsoever - I did not know this until  about a quarter of an hour ago.

When I set up the blog I was, like many people, somewhat anxious about the potential for self-exposure. I anticipated many bad things happening, as new bloggers often do, not realizing that by far the more likely outcome wouldn't be torrents of abuse but complete and utter silence. 

In the interest of anonymity, I put the least-possible information in my public profile but I did include an email address. That email address was invfull@gmail.com. It's been the contact address for the blog ever since but even though my Blogger profile has been accessed over eleven thousand times no-one has ever used it. 

There are two good reasons for that. Firstly, it appears I neglected to tick the box that makes the email address visible when you look at my profile and secondly I don't appear ever to have gotten around to creating an email account of that name at Gmail. I imagine I made it up back in 2010 with the intention of registering it at some future time, if and when I felt comfortable about it, but then I forgot all about it, just like I forget most everything else.

In my head, though, I've been thinking all along that there was an email address on the blog, somewhere. I've occasionally wondered why I never get any mail from it but I figured it must be one I never look at. I have plenty of those. I never bothered to check until now.

Sorry about that.

I can now confirm that this blog does have a valid email address attached to it and it's visible from my profile. Confusingly, it's my Twitter email because I'm signed in to the maximum permitted number of Gmail accounts and I'd have to sign out of all of them to add another (Why? I don't know! Ask Google. It's annoying.) 

The email address for the blog also redirects to one I look at every day so I will see whatever anyone sends there. For a while, anyway. If it turns out to be a nuisance, I reserve the right to sever that link, after which it'll be down to how often I remember to check. Either way, it will now at least be technically possible to get in touch with me without resorting to hijacking the comment thread on the latest post (Which I'm also totally fine with, by the way).

 

From The Comments

Wow! It's almost as though there's a theme developing here. It wasn't planned but I guess I'll go with it.

Angry Onions was kind enough to drop by with a link to a very informative video explaining exactly how ChatGPT works. It's a shade under half an hour long. Less if you skip the commercials. I watched it right after breakfast this morning and I'm embedding it here because it's well worth the time it takes to watch.

In general terms, there wasn't much in the video I didn't already know. The specific mathematics involved, even dumbed down to the nth degree as they are, still go way over my head. For instance, I remember matrices from school and not being able to understand them then. I don't understand them any better now, so referencing a matrix as a way to simplify a much more complex process really doesn't do much for me.

That in itself highlights one of the most significant aspects of the whole AI phenomenon, namely that, for the end user, an understanding of how LLMs and Generative Pre-Trained Transformers work isn't really necessary or possibly even advisable. As has been observed countless times, no-one needs a degree in electronics to switch on a TV set. 

What's important is what shows up on the screen, not how it got there. The underlying technology is interesting, of course, if you have an enquring mind, but what's essential for the end-user to understand are the controls, not the mechanics. 

The current suite of applications we're choosing, highly inaccurately, to call "Artificial Intelligences" are either tools or toys, depending how we use them. When we use them as toys, it doesn't much matter how we play with them, so long as it amuses us but if we're going to use them as tools, we do need to know what they're good for, what their limitations are and how to handle them safely. It's not clear everyone's reading the manual.

There were two statistics in the video I found surprising, even shocking, and neither of them had anything to do with how ChatGPT functions. The first was the sheer number of people currently using the app. Kyle states that as of 2023 "ChatGPT has one hundred million monthly users". I knew the whole AI phenomenon was getting a lot of media attention. I did not know so many people were actually using it.

The second, which relates directly to the first, is that running ChatGPT to service the demands of those hundred million users "might cost $500m a day." Seriously? For an application that's free to use? How the hell are they funding that? And how long is it going to be before the pay walls come up and the AI gravy train runs off the rails?

I asked Bard because I thought it would be rude to ask ChatGPT itself.

 

I have two things to say about that. 

First, I haven't fact-checked any of it against original sources and as we all know - or should - all these AIs are very happy to make something up if they don't know the answer. (Also, AIs don't "know" anything. I know that. No need to remind me. It's a figure of speech.) I'm just quoting what Bard said as a starting point for further investigation, should anyone care to look into it further. 

That's what I find Bard useful for. I'd no more accept anything it told me at face value than I'd take a child at their word if they told me they'd been playing with fairies at the bottom of the garden. No-one would, would they?

Second, saying the same thing twice does not provide "additional details". I suspect there may have been a lot of high school essays in that original data corpus.

Caveats aside, I think it's not unreasonable to assume that billionaires and megacorps are going to be driving the AI revolution forward but that we, the general public, will end up paying for it in the longer run, one way or another. I can absolutely see why that might be a concern. There's not much of a history of forces like that working in anything other than their own financial interest, after all.

In the video Kyle alludes to the frequently expressed concern that unrestricted public use of these models risks creating a feedback effect that will eventually render all data on the internet at best suspect and at worst corrupt. He also (And not, in my reading, entirely facetiously.) nods towards the longstanding paranoid scenario in which machines attain sentience and decide to dispense with humanity.

I'd put the danger of the latter somewhere just above a zombie apocalypse. The former, though, does seem to have some traction. ChatGPT itself is currently insulated from any such corruption due to its siloing within a data set that cuts off sometime around September 2021 but that protective restriction also means the engine becomes less useful for practical purposes day by day. If the AIs are going to have a future they're going to have to live in the present, not the past.

If a hundred million people are pumping AI output into the data-sea every single day now, how many are going to be doing it next year and the year after? It's a fair bet most of those users won't be fact-checking and cross-referencing before they dump their results into the pool. Gigo still applies only now it's going gigagigo.

Which sounds like a suitably gibberish moment to end this segment. Of course we'll be coming back to it soon enough. Probably for the rest of our natural lives.

 

I Nearly Bought A Game But In The End I Didn't.

Story of my life, really. I keep getting emails from Steam telling me something on my Wishlist's up for sale. The list itself has ballooned to more than two dozen titles so something's almost always on some kind of offer. I always check the deal but it's rarely tempting enough to get me to bite.

Not everything on my wishlist is there because I really want to play it, of course. Some things are just bookmarked in case I want to write about them some day. There are a few Point & Click Adventures on there that I would like to get, most of which I played in demo form at one time or another. 

One of them is Crowns and Pawns, about which I was very lukewarm back in February of last year but towards which I have since warmed a little. I did eventually add it to my wishlist and when it came up recently at 45% off I almost pressed "Buy". I'm really in the mood for a good point & click about now.

Luckily, before I spent the money it occurred to me I must have quite a few games in my Amazon Prime Gaming library by now. I checked. I have one hundred and thirty four. I've finished fewer than a dozen. There are more than a hundred I've never even looked at. 

Among them are at least fifteen Point & Click adventures, although most of them are versions of bloody Monkey Island. Still, I ought to be able to find something. Then I remembered one of the games I got free from Amazon, Lake is also on my Steam Wishlist. If I take that one off my list, I thought, and start playing it through Amazon, it'll be just like I bought it!

So I did and it is! That ought to keep me busy for a while.

And finally...

Close With A Song!

Anyone like Suicide?

  Moon Duo obviously do!