Saturday, April 29, 2023

Oh, Come On! There Must Be Something You Like...

After last month, when I added Prime Gaming to the blog roll, I thought I'd written my last Amazon Prime roundup. Then I saw the line-up for May and found I had a few things to say about it, so here we are again. I'm not going to go through the whole list, just the highlights, but if you want the full skinny it's on the official blog or you can see a rundown in brief in the video below.

I found the comment thread following that video on YouTube revealing, although obviously it's in no way representative. Still, you can always read the temperature of the room from these things. I scrolled to the end of the thread and apart from a couple of brief nods towards a couple of the titles, all I saw were complaints about the quality of the games on offer:

"Prime give us some good old AA games"

"What I don't like about Prime Gaming is not that much about "how big the game is". I mean, it's ok for me if I don't get any AAA game. But almost every game included in Prime is really bad, like games that no one would buy them."

"what games are these Prime oh my god too weak games where are those breathtaking games because if it continues like this it will no longer be an advantage to renew the annual plan I expected more I see it's bad"

"Awful , just awful. What happened to you prime gaming ? Last year You're gibing away AAA games and this year its either old games or indie games. Atleast give some old AAA games that ppl enjoy. Another bad month with not a single good game."

Nice to see the level of literacy on YT as high as usual but we get the general idea.

There were also a couple of direct comparisons with Epic Games:

"...nothing worth it. Meanwhile Epic Games on a roll giving better and better games every week"

"Epic games gives better games than these and that too for free"
And there were the usual threats to quit:
"Time to leave Prime"

"Will not renew this month. Hope next month."

" ...will end my Prime subscription right now"

Some of this is misguided or plain ignorant. The trailer for Lake on YouTube, for example, has 2.6m views, suggesting a certain pre-existing popularity. Much of the discontent, however, is founded in reality - of a sort. 

There is an argument to be made that the games we've been getting from Prime this year are less prestigious and of lower authority than in previous years. I highly doubt, however, that it's because Amazon is "having a hard time" as one commenter, almost certainly ironically, suggests. 

I suspect it probably has more to do with a change of emphasis somewhere in the vast, labyrinthine structure of the megacorp although benign neglect is always a possibility. If Amazon has any genuine interest in gaming, it's probably directed at their own games and publishing deals.  

It's long been difficult to align "Amazon Games" with "Prime Gaming". They seem to have little or nothing to do with each other. Maybe the company's current gaming focus is simply elsewhere.

And why wouldn't it be? To compare Prime Gaming with Epic is misleading at best. Epic is mostly a games company. Amazon is mostly not. Moreover, even within Amazon's remit, Prime Gaming is at best an Add-On to an Add-On.

The comments made me wonder about the insularity and tunnel-vision of the commenters and by implication of the wider gaming community. Are there really people who pay their Prime subscription solely - or even mostly - for the free games? It's hard to believe but I guess it's possible.

I took out my Prime subscription years ago for the specific reason I imagine most people did at that time: free, fast shipping. I looked at how much it was costing me to have things sent to my house and how long it was taking for them to get there and  calculated that I'd save money and get stuff faster by paying the sub. I didn't at the time even realize there were other benefits attached and back then I'm not sure there were many.

Over the years since I began subbing Amazon Prime, the service has grown to include a raft of features, some of which, like Prime Video, I use literally every day, while others, like Amazon Music or Reading, I use only very occasionally.  

Gaming falls somewhere in the middle. Every month I claim a few games and some months I even play one of them. The main reason I even know about Prime Gaming it is because I used to claim the free games from Twitch before Amazon bought them and rebadged the service. That pretty much sums up how important Prime Gaming has to be to the overall Amazon Prime offer - it's a thing that came bundled with a thing they bought for other reasons and this is where they put it so they wouldn't keep falling over it in the hallway.

Or that's how it's always felt. Amazon doesn't care about Prime Gaming per se; it's just a bullet point on a list. Look how far you have to scroll down that list of reasons to subscribe. It comes almost at the end, just before Amazon Photos (A thing I didn't know even existed.)

As a Prime subscriber who also games, I appreciate Prime Gaming. It adds value to the offer for me. But if Amazon announced tomorrow they were closing it down it wouldn't for a moment make me consider cancelling my subscription. If I wanted to pay a sub specifically for access to new games every month, Prime is not where I'd be spending my money.

So, even if I was disappointed with the current tranche of free games that come with my Prime sub, I wouldn't start mouthing off about it on social media. And anyway I'm not disappointed. There's something every month that gets me a least a little excited and May's no exception. 

The games in this month's offer that I'll be claiming are

  • Lake
  • Planescape:Torment (Enhanced Edition)
  • Agatha Knife.

I might also take The Also Gone, although it looks a bit... depressing.

Lake is a game I reviewed in demo form a couple of years ago. It's been on my wishlist ever since but even with discounts I've been too mean to buy it. Well, now I don't have to! That's about a month's sub right there.

Planescape:Torment is a game I already own, at least technically. I have it on DVD - or possibly CD, it was that long ago - in a box on a shelf somewhere. I'm sure I don't need to describe it. It has a ferociously high reputation as an all-time classic but I could never get into it at all. I tried but not only did I dislike the premise, I found it... boring. Really dull. Too much text and none of it engaging.  Still, I've always felt I ought to at least give it another chance. Now, maybe, I will.

Agatha Knife is new to me but it looks right up my dingy, dark, alley. The Steam page describes it as "a whimsically twisted adventure" in which the eponymous Agatha is "torn between her love for eating meat and her friendships with animals". Well, aren't we all? Okay, not me. I haven't eaten meat since the 1980s and I didn't much like it then but the same applies to plants. You know they scream when you cut them, right?

Well, that ended on a cheerful note. Oh, well. Time I was mowing the lawn anyway.

Friday, April 28, 2023

All Aboard The Skytrain! - Honkai: Star Rail: Very First Impressions

Yesterday, I wrote about how I woke up to an email from a company I'd never heard of  - Cognosphere - inviting me to try a new game called Honkai: Star Rail. I wanted to take a look but after reading the Terms of Service I had some reservations, which I fed back to the developers by way of the email address they'd kindly provided.

I wasn't really expecting an answer, least of all something drafted specifically to address my concerns. At most, I thought I might get a canned response of some kind, most likely restating the general principles of the issue but clarifying nothing.

Imagine then, how surprised and impressed I was to wake up less than twenty-four hours later to find this in my inbox:

"Hello Trailblazer,
Thank you for contacting us!

We do not prohibit non-commercial personal use and you may, subject to all of the following terms and conditions, produce and publish derivative content based on Honkai: Star Rail materials for non-commercial personal use.

You can use Honkai: Star Rail's in-game footage to create videos, upload videos, or livestream. For more details, we invite you to read the Star Rail Fan Creations Guide and our Terms of service in the articles below:

If you have any further questions, please don't hesitate to contact us again.
Thank you for your support to Honkai: Star Rail.

Best regards,
Honkai: Star Rail CS Team"
Now, this clearly wasn't composed just to answer my email. It's obviously a standard reply to queries of this nature which, by inference, I'm guessing means they must see a lot. Nonetheless, it provides an accurate, helpful and very satisfactory resolution to the concerns I had. 


I took the trouble to (skim) read the linked article 17883171 ( Aka Honkai: Star Rail Fan Creations Guide v1.0) and just like the ToS, it's clear, comprehensive and comprehensible, covering numerous means and methods by which players and fans of the game can express their creativity without running up against copyright issues. 

All of this creates a very positive image of the company, in my mind at least. With the legal decks cleared, the question now remaining is whether the game itself can build on that strong foundation.

After an hour's play, my initial impression is that yes, it most certainly can. After I got back from walking Beryl the dog, who immediately fell fast asleep in the armchair behind me, where she remains, immobile, four hours later, I got straight to installing the game, a very swift and straightforward process that took just a few minutes.

Rather than make a new account I thought I'd try my Genshin Impact login. It worked. I ummed and aahed a bit about whether to play on the European or American server, plumped for the US, then caved when the game asked me if I was really sure, what with my IP address being in Europe and all. 

I'm not entirely sure if Honkai: Star Rail is a quasi-mmo or a straight-up rpg and even an hour's play hasn't cleared that up for me. I never really worked out what Genshin Impact was, either. For a long time I thought other characters I was seeing in the game were players but in the end I realised they were all NPCs. Or I think they were. Frankly, I'm not certain, even now.

Anyway, if the game's not a full-on mmorpg, I don't care where the servers are located. I'm not going to meet or talk to anyone, so it makes sense to go as close to home as possible. All the cultural and practical advantages of playing with first-language English speakers outside primetime cease to apply.

One reason I'm still unclear about the exact nature of the game is that, after an hour, I think I'm still in the tutorial. Again, it's hard to tell. I have reached the control deck on the space station and a disembodied voice has told me that from now on I'll need to make my own story, but the tutorial pop-ups are still coming. Then again, in some games (Yes, I am looking at you, Chimeraland.) those kinds of pop-ups never seem to stop.

There was a time when an hour would have seemed like quite a stretch for a tutorial but those days are long gone. Now, tutorials regularly seem to last for days. Maybe even weeks. I guess it's a good thing. I mean, we're always learning, right? 

Yeah, me neither. I'd rather just get on with it, make my own mistakes and hopefully learn from them, even if it isn't the efficient way of doing things. Luckily for me, no fricken' game ever tells me the things I want to know in the tutorial, so I get to figure it all out for myself anyway!

By that, I mean they very rarely tell you how to remove screen clutter for a nice, clean, minimalist look and they never tell you how to take a screenshot. I can't fault H:SR on the clutter issue; it's as neat and unfussy as I'd like it to be. Didn't need to touch any controls there at all. Screenshots, though, are a different matter.

Of the hour I spent playing H:SR at least ten or fifteen minutes must have been fiddling around in the settings and keybinds, looking for a screenshot hotkey, then googling to find out there isn't one. What there is instead is one of those in-game cameras that lets you pause the action and take posed photos and selfies. They're great and I love them. I use them all the time in any game I play that has them. They still aren't a replacement for an actual screenshot key.

Belghast put up a post yesterday detailing how he's set up his system to take screenshots of whatever he's playing automatically, one every sixty seconds. I don't want to go that far but I do want to be able to record things so I can use them here. 

I tried FRAPS but it didn't work first time so rather than fiddling around with it I thought I'd try Windows own baked-in screenshot function. I never use it normally, mainly because I never remember it's there, but it's actually pretty good. The only problem is the keys you have to press - Win+PrtScr - are on opposite sides of the keyboard. You'd need to be Robert Pershing Wadlow to make that combo with one hand.

Enough about the mechanics and my ineptitude with them. How about the gameplay? Well, you can't - or shouldn't - read too much into a tutorial but the first session, the first hour, of any game is crucial. If you're going to use it to tell people how to click on things, you're going to get judged on that. It's when the hooks, if any, are going to be set. 

Did Star Rail set them and if so, how deep did they go?

Here's the thing. I would totally watch Honkai: Star Rail the anime or TV show or even Visual Novel. The writing is good, the characters are strong and likeable (Other opinions are available.), the setting is intriguing, the graphics are stylish, the plot is involving, the voice acting is excellent... The problem, as usual, is with the game part, specifically all that fighting. 

Not that I disliked it. Combat is turn-based rather than real-time action, which makes it far more accessible for me than Genshin Impact. I didn't have too much trouble getting to grips with it, although the ironic thing about tutorials is that by using pop-ups to tell you where to click, they can end up leaving you not knowing much more than when you started about what to do when the "Click Here" boxes vanish.

No, my problem is that the combat's there at all. When you're enjoying the banter between the characters and trying to follow the metafictional plot, having to stop every few minutes for a battle in a corridor is just annoying. 

It may be, of course, that the plot drops into the background more when the tutorial comes to an end but that never really happened in Genshin Impact so I doubt it'll happen here. The question is whether I'm going to find myself sufficiently invested in finding out What Happens Next to put up with all that repetitive shooting each other in the face. 

The jury's still out on that - more evidence definitely required - but Honkai: Star Rail makes a strong first impression. I'm going to need to find out where this train goes next before I make any decisions about whether to stay for the ride.

Thursday, April 27, 2023

This Train Terminates Here

I woke up this morning to an interesting, if unexpected, email. A company by the name of Cognosphere had sent me an invitation to download a new "space fantasy RPG" called Honkai: Star Rail. It sounded very intriguing, being set, by some manner or means, on a kind of space-train, known as The Astral Express. I was keen to find out more.

The email came with all the necessary buttons to press for registering an account and downloading the game, so I began pressing them. At that point I wasn't quite sure who the developers were or what games they might already have made but it seemed like a safe assumption it  must have been something I'd at least tried, since they'd gotten hold of one of my email addresses from somewhere.

There was one big clue. Honkai: Star Rail is part of something called the HoYoverse. That didn't ring much of a bell with me, although I thought it might possibly have something to do with Genshin Impact.( It does. See Note #1 below.)

It didn't seem like something that mattered enough to go look into it before I downloaded the game, so I just got on with it. I'd gotten to the point where the tiny zipfile I'd downloaded had unlocked an installer and I'd picked a drive location from the advanced options, when the inevitable request to accept the Terms of Service popped.

Terms of Service and EULAs are really boring to read. They're long and full of legal jargon and most people just tick the box to say they've read and understood them without actually bothering to do either. Sometimes I do that, too, especially with new games from developers whose games I already play.

With new games from sources unfamiliar to me I tend to be cagier. I wouldn't claim to read every line of every paragraph but I always at least skim the whole thing and close-read, often several times, the specific sections relating to what I can and can't publish about the game, here on the blog. These days, if there's a strict NDA or draconian rules about screenshots or video, I decline to accept the terms and cross the game off my list.  

The TOS for Honkai: Star Rail is exceptionally clear and well-written. While it uses a lot of legal language, it does so in very good, plain English. None of these things are fun to read but this one makes getting through it about as painless as it could be. 

Several key prohibitions are highlighted in bold and prefaced with straightforward explanations and warnings about why users should pay particular attention to them. Where appropriate, long lists of examples are included. All in all, it's a very impressive example of the species.

As is altogether too common these days, it includes a section relating to use or reproduction of "materials" from the game, the list of examples of which include, among many other things, stories, storylines and visual images of various kinds. It's forbidden to reproduce, republish, display, transmit or perform several other communicative processes on any of these aspects of the game without first receiving express written consent.

That, I figured, would really put a crimp in my writing about the game here on the blog so I declined to sign the TOS or to register an account or download the game, to do any of which would have implied acceptance of the terms. It may seem a little over-cautious but you'll note I'm not even quoting verbatim from the TOS here, let alone reproducing images of it. Can't be too careful.

Still, I would like to play the game. It's not every day - or every decade - you get the opportunity to ride a space train. I could, of course, just download it, play it and keep quiet about it, but where's the sense in that? I don't play these games for fun, you know! Well, I do, but much of that fun comes from writing about them. 

As I said, the TOS is exemplary in its accessibility so it shouldn't be surprising that it concludes with an email address to which you can refer any questions you may have before accepting it. It did surprise me, all the same. It's something I'm not sure I've seen before.

Since they'd been so kind as to include a way to contact the legal department, I thought I might as well use it. After all, you don't get if you don't ask. Here's the email I sent. I'm pretty sure it's safe to reproduce, seeing as how it's all my own work:


Could you please confirm whether the prohibitions on copyrighted materials in this section extends to screenshots from the user's gameplay when published online in a personal blog for the purposes of illustrating either an account of the user's activities in game or in a review or commentary on the game? Similarly, does the prohibition apply to video of that nature posted in a user's social media channel for those purposes?

If these actions are so prohibited without express prior written consent, can you issue such consent to me for those purposes or advise me how and where to apply for such consent?

I'd like to try the game but I'm not interested in doing so if I can't also write about it and show screenshots on my blog.

Thanks in advance for your advice."

We'll see what that gets me, if anything. Until then, I guess I won't be playing the game. And even if I do, you won't be reading about it here.

By the most extreme of contrasts, yesterday I also signed up to test an in-development mmorpg, Monsters and Memories. Currently at a very early stage of development and being produced by a pretty small team, M&M is a rather cosy-looking entry into the somewhat crowded field of retro-repro Classic MMORPGs. 

I've had my eye on it for a while but it's been too early to mention anything much about the project beyond that it exists. The days of getting all excited over some concept art and a few promises are long gone for all of us, I imagine. 

This weekend, though, Niche Worlds Cult, the amusingly self-deprecatorily named company behind the game, is running an open stress test. All you have to do if you want to try it is sign up. So I did.

The process is very straightforward. Just give them an email address, confirm it and download the game. Even that's part of the test so you're helping out whether you actually get to log in or not. There are two fixed sessions, which could be extended if things go well, but the proposed schedule is

Friday, April 28th, 9pm-1am EDT -- (3am-7am CEST)

Saturday, April 29th, 1pm-5pm EDT -- (7pm-11pm CEST)

I'll be sound asleep for the first of those but the second sits nicely across Saturday early evening for me so there's a good chance I'll be able to give it a go. If I do, I'll be able to record anything that happens here, with screenshots, because when NWC say open testing they really mean open:

The test is purely a technical one with any gameplay possibilities being merely incidental:

PLEASE NOTE: By registering for, downloading, or logging into Monsters & Memories, you acknowledge your understanding that this is a technical "stress test" of an early development build and not a pre-alpha, alpha, beta, or any other form of test indicative of the game being in a ready-for-release state. 

Even so, should there be anything to report, I'll be able to do so freely and without reservation. In a helpful clarification that appears during the registration and download process but which I unfortunately didn't screenshot and now can't find, it's explained in very clear terms that testers are free to record and discuss their experiences on whatever social media suits them. 

NWCs position seems to be that all publicity is good publicity, an old saw that's very much lost its teeth in recent times. In the case of games seeking to draw attention in an extremely competitive marketplace, though, I can't help but agree. I wish EULAs and TOSs would include some provision for "personal use", allowing players to share their in-game activities through social media. It seems counterproductive to make it harder than it need be for word of mouth to do your marketing department's job for them.

Note #1 - Cognosphere is the new (As of a year ago.) publisher of Genshin Impact. I remembered the GI publisher as miHoYo, which is how I came to connect it with the HoYoverse in the first place. As the linked article explains, it's not a substantive change of ownership. Cognosphere is "a company owned by miHoYo, registered in Singapore as opposed to the Shanghai-based miHoYo." This seems to be a response to the Chinese government's increasingly strict regulation of China's video game industry. 

Note #2 - Whether the Cognosphere TOS is significantly more restrictive than the miHoYO version that would have been in place when I was posting about the game and using dozens of screenshots, I have no idea. I guess either it is or I didn't read the old one as attentively as I did the new. I can confirm, though, that the current Cognosphere TOS for Genshin Impact looks pretty much the same as the one for Honkai: Star Rail, so I guess I won't be blogging about Genshin Impact any more, either. Not that I was planning on it.

Note #3 - All images in this post are taken from Monsters & Memories. For obvious reasons.

Tuesday, April 25, 2023

"I'm Not Sure What To Say About That"

I got quite excited the other day, when Nimgimli pointed me to something Square Enix were trying out on Steam. They're calling it a "Tech Preview" and the description describes it as "a demonstration of Natural Language Processing (NLP), an AI technology".

As far as I can tell, Square have grubbed up a very old adventure game called The Portopia Serial Murder Case, originally released in 1983 (!) for a system called the NEC PC-6000 and force-fed it their proprietary AI tech to see what might happen. According to Wikipedia, the original game has suffered similar indignities in the past, having been ported to a few other long-forgotten platforms before eventually making its way into the mobile market and finally arriving on Windows and Steam as a shill for Square's "natural language processing technology."

Again according to Wikipedia, TPSMC ,as I'm going to call it to save all that typing, went down pretty well in Japan on its original release and was seen as being influential in the development of the visual novel genre. It sounds like the kind of game I'd enjoy, with its "point-and-click, murder mystery plot, open world, suspect interrogations, nonlinear gameplay, dialogue choices, and alternate endings" which, as a review in the Official XBox Magazine a few years back pointed out, were "standard for 2015, but way ahead of its time in 1983".

The Tech Preview is free to play on Steam so I thought I'd give it a try. I was somewhat surprised at the size of the download which, at more than 8GB, seemed like it would have required a government supercomputer to run in 1983, not a rinky-dink home computer. The System Requirements actually ask for 15GB though, so I guess I got off lightly.

Those requirements also contain a curious note about Speech Recognition: "Voice Input (Speech To Text) requires CUDA cores. (*1) VRAM 8GB and above required when using “Medium” in Voice Input quality settings. (*2) VRAM 16GB and above required when using “High” in Voice Input quality settings. Keyboard and/or Microphone support only." It seems to suggest you can play the game by talking to the characters rather than typing your conversations and reading their replies. 

That's something that would very much have been seen as the expected future of computing back in the 1980s, when Japanese tech giants were throwing vast amounts of money and effort behind the so-called "5th Generation", one of whose fundemental principles was to replace keyboard input with the spoken word. Much like videophones, which were a solved problem by the early '90s but which failed to attract any kind of interest from the public, speech recognition, when it was perfected, turned out to be something most people didn't want - and arguably still don't.

Still, I'd like to try it in an adventure game whose gameplay revolves around interrogation. It seems like a natural fit, especially when you throw in AI. I'm already excited for the inevitable and probably imminent integration of language processing AIs like Bard and ChatGPT with Smart Speakers and Virtual Assistants like Alexa and Siri. To see something like that in a game would be amazing.

There's only one drawback. I don't have a microphone.

Okay, technically I do have one. I actually have several, including two integrated webcam/mics that I acquired when I foolishly bought into John Smedley's lucicrous SOEmote fantasy ,of which the less said, the better. 

What? You don't remember SOEmote? Oh, what the hell...

I guess if you have a naturally gurning, cartoonish face like Dave Georgeson, maybe it would seem like a good idea, although evet then it would actually have to work...

Anyway, while I might theoretically own some mics, I don't have one to hand and even if I did, I definitely don't want to start fiddling around trying to get this game to recognize it. I mean, I might, if I thought there was the least chance of it actually working, but I've tried the text input. "Fool me once..." as they say.

About that text input. Y'know, I've played a lot of adventure games. I was already playing them when TPSMC first came out, back in the very early '80s. I used to sit there, with my first wife, in our first appartment, typing keywords and phrases into the ZXSpectrum's rubbery keyboard, hoping to see some kind of acknowledgement or recognition on the portable TV in front of us.

A decade and a half later I'd do the same thing, with Mrs Bhagpuss leaning over my shoulder and making suggestions, as we plowed our way through Broken Swords #1 and #2. Even EverQuest used a text parser before someone at SOE finally got around to adding clickable keywords.

Any - all - of those text parsers were orders of magnitude better than whatever the hell "Tech" it is that Square Enix thinks they're previewing here. And I'm not the only one making that assessment. As I write, the game or demo or whatever it's supposed to be has Steam's lowest possible rating - Very Negative - and in my opinion it's lucky to have a score as high as that.

Try some choice comments from angry reviewers:

"...the "AI" in this game is a joke. It is perhaps even worse than the text parsing of the original game, from 40 (!) years ago."

"GEEZ It's like they didn't even try...."

 "Zork (1977) had a better understanding of what your commands meant"

"Early text-based RPGs are way smarter at responses than this."

"basically it understands less inputs than the original King's Quest"

"This is a fascinating idea, but the execution is... I'm not sure it even rises to the level of "weak." It's not in evidence at all."

Aha! I think we've hit on the problem!

"Barely works. They say they removed the AI component due to safety concerns... so this doesn't even have the tech that it was supposed to be a preview of in the first place."

Well, that explains a lot. Okay, it doesn't explain how the game as it stands is literally worse than the basic, acceptable standard of games of its type from the era in which it was made, but it does answer the question I was asking myslef throughout the whole of the five (Count 'em - five!) minutes I could stick playing the blasted thing, namely "AI? What AI?"

I thought I should check if the allegation was true, so I asked Bard. Bard, being an actual generative AI, not a pretend one, was quick to give me the full story:

Hi Bard, Can you tell me if Square Enix has disabled the AI component of their tech demo for The Portopia Serial Murder Case?

Yes, Square Enix has disabled the AI component of their tech demo for The Portopia Serial Murder Case. The demo was released on April 24, 2023, and it allowed players to interact with the game world using natural language processing (NLP). However, the AI component was disabled on April 27, 2023, after players began to exploit it to generate offensive and inappropriate content.

In a statement, Square Enix said that they were "disappointed" that the AI component had been used in this way. They said that they were "committed to providing a safe and enjoyable experience for all players" and that they were "working to improve the AI component so that it can be used in a more positive way."

The disabled AI component was a major disappointment for many players who were looking forward to trying out this new technology. However, Square Enix has said that they are working to improve the AI component so that it can be used in a more positive way. It is possible that the AI component will be re-enabled in the future, but it is not clear when this will happen.

Other than Bard's curious ability to predict the future (Check the dates it's quoting - I almost missed it.) and fondness for repeating itself, that seems to be a reasonably accurate account of what happened. It's corroborated by this report at and many others as well. 

So much for that "preview". It does invite the question "Why leave the bloody thing up with the claim the AI's still in there, then?", but since this is Square Enix we're talking about, I wouldn't expect a coherent answer.

Someone is going to get this tech hooked up and working effectively but clearly that someone is not going to be Square Enix, who seem to be about as realistic in their expectations and as technologically capable of achieving them as Smed and Smokejumper's SOE. 

And we al know what happened to them...

Monday, April 24, 2023

Born On A Ray Of Sound: RIP Mark Stewart

 "To inspire somebody would be one of the best things we could possibly do."

That's Mark Stewart of The Pop Group, speaking to Paul Rambali of the New Musical Express back in 1978. Mark died last week. I was surprised to learn he was 62, a couple of years younger than me. I always just thought he'd be older.

As The Pop Group's wikipedia page puts it "Despite their lack of commercial success, their music inspired and influenced many bands and artists that followed them. They are cited as an influence by artists and bands like Minutemen, Primal Scream, Sonic Youth, Steve Albini of Big Black, Trent Reznor of Nine Inch Nails and Massive Attack." And the rest, eh?

In retrospect, I think it's possible to overstate The Pop Group's outsider status. I recently had cause to go through a stack of music papers from the late '70s and I was surprised at the sheer number of times the band's name appears, often prominently. They were cover-featured on the 30 September 1978 issue of NME, behind a gorgeous photo of Mark, looking like a male model or a matinee idol from the age of the Silver Screen. At that point they'd never made a record.

The uncanny ability to draw attention was theirs from the start and Mark's charismatic force was a major part of it. I was in the audience for The Pop Group's first gig. It was at a nightclub on Clifton Downs in Bristol called The Locarno, a sizeable venue for any band who'd never cut a record or played live, let alone one all of whose members were still at school. The place was packed. The Pop Group had a rep in their home town even before they showed up to earn it. Even their first gig was An Event.

As I remember it - and it has to be stressed my memory is fallible and I wasn't, sadly, in the habit of keeping a diary back then - there was no support band. Instead we were treated to a couple of hours of the band's own favorites and influences, an eclectic selection of unfamiliar and frequently uncomfortable slabs of jazz, funk, dub and if I'm not imagining it, musique concrete. I'm pretty sure there was some Pierre Henry in there, somewhere. I think that's how I came to own one of his albums.

It felt like a long two hours, although I had reasons not to want it shortened that had nothing to do with the music and upon which I don't care to elaborate. Eventually the band appeared. They looked uncanny and sounded... less alien than they would in a few months. Among a storming set of powerful originals they included rousing covers of Jonathan Richman's Roadrunner and - unbelievable as it sounds now - Marc Bolan's Solid Gold Easy Action.

A few weeks later my best friend Chris and I published a fanzine, which we called Mrs O'Reilly's Dog, because even then I was determined to make sure anything that had my name on was as obscure, inaccessible and uncommercial as possible. In that opening and, as it turned out, only issue, there was a review of the Pop Group's first gig written, I think, by Chris, in which the band were described as something along the lines.of "four midgets and a giant".

I can't recall the exact phrasing. I do have a copy I could refer to but it's in the loft and I'm too old to be climbing through that tiny hatch any more. Anyway, chances are the review wouldn't be there to check. Someone in The Pop Group or their management took offense, either to the way we'd characterized their physical appearance or the way we honed in on those cover versions, I can't remember which. 

Mark Stewart's height, sometimes given as six-foot seven, comes up frequently in pieces about the band. I doubt that was a problem but no-one likes to be called a short-ass - or a covers band. 

Chris agreed to remove the offending review from all the copies we had. He literally cut them out with scissors. We would have sold them with half a page missing, had anyone wanted to buy them, which fortunately for our nascent journalistic integrity, they did not.

How we dressed in 1978. If we were The Pop Group.
Possibly despite that incident or possibly because of it, relations between our band and The Pop Group flourished. We supported them twice at local gigs and I learned, too late, that we'd been invited to support them at a more prestigious London show.  Supposedly a couple of the group really liked us, while the rest very much did not. I was told which was which at the time but it's long lost knowledge now. It'd be nice to think Mark was one of the pair who thought we had something but I fear it wasn't him.

I retail this anecdote second-hand because I, personally, don't recall ever doing more than passing the time of day with any member of The Pop Group. All communication went through either Chris or his brother, Phil, our drummer and guitarist respectively. The first I ever heard of this supposed invite was when one of them told me they'd turned it down because London was too far to travel. If there's one thing I regret more about my time in the band than that we never made a record it's that I didn't involve myself more in the administration but what nineteen-year old wants to do that? It just seemed so much easier to let someone else do it.

All of this I told, years later in the mid-1990s, to an eager American guy who'd just started a website dedicated to The Pop Group. I'd found the site while trawling the web, something that seemed like an adventure in itself back then. Following my already-established pattern I left a long comment, which led to first to a correspondence, then to my sending him a copy of the controversial review, and finally to a personal visit from the website's creator, who came to the city where I live and stayed with us for a night.

Double page spread in Sounds, three months even before the NME cover.


All of these events framed and shaped my attitude to what information I was willing to reveal about myself across the internet for many years. I mean, it was exciting and all but when people come to your house it gets weird.

Backtracking, almost the last time I saw The Pop Group was around 1980 at the Corn Exchange in Cambridge, where I was at university, studying English. They were co-headling a bill with Linton Kwesi Johnson, the dub poet. Mark Perry's band, The Good Missionaries were supporting.

I dragged a bunch of college friends along to see them. Like most of the audience, they were mainly interested in LKJ. When it was announced, after the gig had begun, that he wouldn't be appearing after all, half the audience left. The Good Missionaries played their fairly uncompromising set, which shifted a few more people out into the street. Then The Pop Group came on to see if they couldn't get rid of the rest.

By then, they'd released what is still one of my favorite singles of all time, the incendiary She Is Beyond Good And Evil, as well as their now-legendary first album, the enigmatically-titled "Y" but most people left in the room hadn't come to hear any of that. The Pop Group didn't care. They played it all anyway, giving one of the most inacessible, challenging performances I've ever heard. There certainly weren't any crowd-pleasing covers this time. 

She Is Beyond Good And Evil - The Pop Group

By the end of the gig, the hall was more than three-quarters empty but most of the people who'd stayed were onstage with the band, having a great time. It was a rioutous, joyous, celebratory affirmation of acceptance of the other, a philosophy The Pop Group lived as well as preached. Until a year later, following the release of their second, even more uncompromising album, "For How Much Longer Do We Tolerate Mass Murder?", when the band split up. 

As Rolling Stone described it "The Group broke up in 1980, as Stewart’s interests drifted deeper into reggae and his bandmates wanted to explore free jazz.", stretching the old "musical differences" cliche to its absurdist limits. The various members dispersed, forming a number of more commercially successful outfits that included Pigbag, Maximum Joy, Rip, Rig and Panic as well as my own favorites, the before-their-time and decidedly not commercially successful Head

Mark Stewart wasn't in any of them. He moved smoothly into a successful solo career, rich in collaboration. He transferred his very considerable talents to Adrian Sherwood's On-U Sound collective, where he recorded and toured through the eighties alongside the likes of Tackhead, African Headcharge and the New Age Steppers

Hysteria - Mark Stewart and The Maffia

I saw him on stage again, for the final time, sometime in the mid-eighties, when he made a triumphal return to Bristol, playing in a kind of On-U Sound package tour with Tackhead and Gary Clail, among others. The entire event was top-class musical entertainment but even among that stellar line-up, Mark effortlessly prevailed. His overwhelming physical presence and the floor-shaking resonance of his voice made him a standout in any company.

According to all of the personal tributes from those who knew him, he was an affable, friendly kind of giant, immensely willing to help guide and mentor other musicians. He offered a wellspring of creativity and imagination and acted as an indefatigable campaigner for social justice as he conceived it - and he was a right good laugh, to boot. 

I know nothing of that. I just know he was a blindingly good performer who fronted a blindingly good band at a time when few either wanted or understood what he and they had to offer. It meant those of us who thought in the know were able to feel all clever about it.

Fortunately for Mark, there were some people out there who really did get it. Time proved him right and everyone else wrong. He found an audience who truly understood his message and in his wish to be an inspiration to others he absolutely succeeded.

He will be missed. He already is.

Friday, April 21, 2023

What's In A Name?

Man! I've written a lot of long-ass posts lately. I really ought to lay off for a bit. Well, I guess I'm working all weekend, so there's two days I won't be posting. I wouldn't want to miss today, though, so what shall I do?

Oh, I know! One of those Friday Grab-Bags! Except not one where I start out meaning to post half a dozen short paragraphs on six different topics then end up writing two thousand words on one of them and taking all day about it. No, these are going to stay short. Really...

That's Not My Name

Let's start with DCUO, Haven't mentioned that one in a while. I saw there was a new Episode, Shock to the System, and I thought it might come with some free stuff for my Lair, so I logged in to see. 

It didn't. Boo! Hiss!

The first thing I saw after that disappointment was a message saying the server was about to shut down in twenty minutes for maintenance so I didn't even have a chance to investigate the new content. I did manage to find my way to the new area, Dakota City, and take a few screenshots. It looks gorgeous, like all cityscapes in DCUO. Very retro-futuristic, all 90s glass and deco stylings.

More importantly, I spotted an announcement on the launcher warning - or I probably should say promising - a Name Reclaim event next month. This is, apparently, only the second there's ever been and it's been much demanded by disgruntled players, frustrated by the "millions and millions" of abandoned, forgotten and unplayed characters sitting on prime naming real estate.

There's a fascinating six-page thread in response on the forums, from which I learned that last time this happened there was an absolute feeding frenzy, involving people selling names for hundreds of dollars and going nuts over the availabilty of "OG" names like Sentence, Paragraph, Shopping List, Goat Milk, Dog Food or Credit Card. (Actual examples quoted in the thread.) People are weird. 


Here's how it works: 

On or around May 11, 2023, we will reclaim names from characters on inactive accounts, freeing up those names so that you or anyone can again use them when creating new characters or renaming existing ones.

  • Characters on accounts that have not logged into the game since April 11, 2022 (or, in the last 12 months) will be renamed.
  • Renamed characters will have an underscore and Roman Numeral added to their existing names.
    • Accounts and characters will remain otherwise intact. No characters will be deleted.
  • Those names will become automatically available to current players through the regular character creation or rename processes.

If you want to save that name you gave some character you made back in 2017, played once for ten minutes then forgot about for five years, just log the dormant account in before 11 May and it's safe - for as long as it takes for the tide of name-deprived anger to rise again and they have to do this a third time. I'd guess sometime around 2030 if precedent stands.

You only need to log one character all the way in and it flags everything on the account as protected. I'll be logging both my accounts in although the chance of anyone wanting to steal my names is about as small as Ray Palmer when he goes into action. 

Ray Palmer is The Atom, or used to be, by the way. I knew that without having to look it up although I have no idea who The Atom is now, except I bet it's not Ray Palmer any more.

So much for the PSA. On to the next freebie.

Sixteen Candles

Similar story. I read that Lord of the Rings Online was giving stuff away for the Sixteenth Anniversary, which reminded me - wasn't I playing LotRO a while back? I bought an expansion or something, didn't I? How quickly these things slip away.

I'd tell you what I got when I logged in except that I've been into DCUO, taken my shots and written all of this post so far and LotRO is still patching. And I've played it this year, I swear. There can't be that much to update. And I started patching it before DCUO! I'd make a joke out of it if it was funny but the LotRO launcher and its staggering inefficiency stopped being funny a decade ago, if it ever was.

It's still not done so let's get on to the next thing. I'll come back to LotRO later, assuming I can be bothered.

The Stanford Experiment

With all the talk about AI going around, I'm sure no-one needs another post about it from me just yet. Which is a shame, because I have plenty more to say. Oh, boy, do I. 

I'll restrain myself for the moment. Instead, I'll just link this opinion piece from Rob Fahey at  As he says, there's a lot of crazy talk right now about where all this is going but I'm willing to bet that one of the first practical gaming-related applications we'll see is this: smarter NPCs that don't just repeat the same stock phrases over and over. It seems like a gimme.

In the piece, which is well worth reading all the way through, Rob mentions a "proof-of-concept" by researchers at Stanford that involved "a bunch of LLM-driven AI agents running around in a Sims-style simulation". He didn't link to it but here's a write-up on the experiment. And also the Stanford/Google team's PDF on it, if you're really interested. 

You can watch a replay of the little guys doing their thing here. It's weirdly hypnotic.

Sixteen Again

Oh, LotRO finally finished patching. Let's go see what I got!

Bleh. I forgot I didn't start playing LotRO until a couple of years after it launched. I got a present but it was the 14th Anniversary package. Old news!

It was the usual, anyway: some xp potions, a few fireworks, a title and a hat. More junk to take up inventory space. Although the hat's not bad. I'd wear that. 

There are better and more up-to-date rewards at the Anniversary event itself but for that I'd actually have to play the game. Maybe next week. I do fancy running around with "a twee celebration goat pet" as Syp puts it

 You Ought To See Me In A Gown

And finally, because I am, for once, going to keep my promise and make this quick, I suppose we should have some music, just because we always do. Let's see, I have a whole raft of stuff stashed but most of it is earmarked for a couple of posts I have in mind...

Ah, how about this?

Nostalgia - Suki Waterhouse

When I picked Suki Waterhouse as a standout from the cast of Daisy Jones and the Six, I had no clue she was an actual musician and performer, not just an actor playing one. I knew I'd heard her name but I didn't realise it was from seeing it in the sidebar on YouTube. 

It still doesn't explain what her cut-glass accent is doing in the show (Yes, I know why. I'm being arch.) but it certainly explains why she's so convincing in the role.

I could go on but I won't. Things to do. Places to be. 

Back after the weekend.

Thursday, April 20, 2023

A Hopeless Enterprise

Janelle Shane
at AIWeirdness (It's over there in the Blog Roll. I hope you're all reading it.) posted a couple of pieces recently in which she asked various AIs to create some ASCII art and then rate how well they'd done. They thought they'd done brilliantly. They really hadn't.

It was an interesting experiment, demonstrating not just the fallibility of these so-called "intelligences" but their peculiarly hubristic self-confidence. It was also somewhat odd, in that there's a lot of ASCII on the web and we're repeatedly being told that the AIs trawl every corner of it for data points, so how come none of them seemed able to find a picture of a giraffe, something Google has no problem doing.

It was a mildly interesting diversion and a timely reminder that we can all too easily get swept away by the hype when a new technology come bounding over the horizon but I'd probably have left it at that, had it not been for something Wilhelm posted over at TAGN. He was asking the AIs if they could tell the difference between MMOs and MMORPGs, with varying degrees of success.

I popped into the comment thread with a note about Janelle Shane's ASCII experiments and that led Wilhelm to run one of his own. That in turn led to some email correspondence between us and eventually to my own investigation, the results of which you're about to see.

ASCII art isn't generally something that interests me much. I think you'd need to be more of a coder than I ever was to find it really fascinating. It does, however, seem to be something with which an AI trained on data from the worldwide web should be both familiar and comfortable, so I found it increasingly strange to see just how very, very badly all the generative AIs seemed to react to any mention of it.

The blob ChatGPT produced in response to Wilhelm's request for an ASCII picture of the Starship Enterprise was the last straw. It seemed so much less likely a result than pulling up the correct image would have been. I decided to go ask them what they thought they were playing at. Well, one of them, at least.

First of all, I wanted to make certain ChatGPT-4 knew what the Starship Enterprise was and what it looked like. I logged in to OpenAI and asked:

"Please tell me what you understand by the name "Starship Enterprise". If possible, append an illustration."

ChatGPT was happy to comply:

"The name "Starship Enterprise" is most commonly associated with the fictional spacecraft featured in the popular science fiction franchise, "Star Trek." The Starship Enterprise is a highly advanced starship operated by the United Federation of Planets, and serves as the primary setting for the various "Star Trek" TV shows, movies, and other media.

The Starship Enterprise is a massive, state-of-the-art vessel that is capable of interstellar travel, exploration, and diplomacy. It is typically depicted as a sleek, saucer-shaped primary hull with a large secondary hull and nacelles for faster-than-light propulsion. The ship is equipped with advanced technology, including warp drive for faster-than-light travel, phasers and photon torpedoes for defense, a transporter for beaming personnel and objects, and a holodeck for virtual reality simulations."

Okay, so we know for sure that ChatGPT knows what the Enterprise is and what it looks like. But could it show me a picture? Why, yes it could:

"Here is an illustration of the Starship Enterprise as depicted in the popular "Star Trek" franchise:"


Okay... Now that's weird, for starters. I didn't ask for an ASCII image. I just asked for "an illustration". There are a billion images of the USS Enterprise on the web. Why go for that one?

[Edit - As I was writing this post, Wilhelm emailed me about his own experiments, including this admission, which he managed to wring out of ChatGPT: "Unfortunately, as an AI text-based model, I am unable to create visual illustrations. However, I would recommend searching for "Starship Enterprise" using an image search engine to find various illustrations and renditions of the iconic spacecraft." I guess that explains why it defaulted to ASCII although not why it made its excuses when Wilhelm asked for an illustration but for me just kept quiet and hoped I wouldn't notice.]

And you can, just about, tell it's the Enterprise. It may not be a great picture of the ship but it has the basic outline. It's not a blob. If ChatGPT can pull an ASCII image of the classic version of the Star Trek ship out of its hat when it isn't being asked for one, how come it can't get anywhere even close when it is?

Thinking about the implications, I decided to ask ChatGPT for some more ASCII pictures of the same ship:

"Please show me three images of the Starship Enterprise made up from symbols from the ASCII code set."

This is what I got:

"Sure! Here are three ASCII art images of the Starship Enterprise made up of symbols from the ASCII code set:"

A couple of things spring immediately to mind here: firstly, and most obviously, none of these looks even remotely like the Enterprise. The first two could conceivably be diagrammatic views of spacecraft but the third looks like a jellyfish or a broken chandelier. The Enterprise, though, they are not.

Since we have established that ChatGPT both knows what the Enterprise is and what it looks like when represented in ASCII, why is it incapable of reproducing that information accurately when asked? 

The other, less immediately noticeable peculiarity is that, having been asked for three examples, ChatGPT provides them but numbers each of them "1". Again, why?

But maybe we're making some false assumptions here. ChatGPT might know everything about the Enterprise but does it know what ASCII art is? I thought we ought to make sure:

"Please explain what ASCII art is, with examples."

"ASCII art is a form of digital art that uses characters from the ASCII (American Standard Code for Information Interchange) character set to create visual representations or images. ASCII is a standardized character encoding scheme that assigns a unique numerical value to each character, including letters, numbers, punctuation marks, and special characters, which can be represented as text in computers and other devices.

ASCII art is created by arranging these ASCII characters in a pattern or grid to form an image or design. ASCII art can be simple or complex, depending on the amount of detail and the arrangement of characters used. It can be used to create various types of artwork, such as portraits, landscapes, objects, and even abstract designs, using only text characters.

Here are a few examples of ASCII art:

Fine. Those are certainly examples of ASCII art and they even look like the things they're meant to be. Again with the 1 numbering convention but we'll let that slide.

ChatGPT even had a postscript:

"ASCII art can be created using various tools, such as text editors, specialized ASCII art generators, or by manually arranging ASCII characters to form the desired image. It has been used in various digital media, such as emails, web pages, social media posts, and online forums, as a creative and expressive form of digital art."

Probably more than I needed to know but, sure. It knows exactly what ASCII art is and what it looks like. Next question - has it been taught how to do it?

"Have you been trained to recognize ASCII images of this kind and can you produce new ones yourself?"

Chat GPT was happy to confirm that it had and it could:

"As an AI language model, I have been trained on a large corpus of text data, including ASCII art, which allows me to recognize and understand ASCII images to some extent. I can also generate new ASCII art based on the patterns and structures I've learned from the training data. However, please note that my ASCII art generation capabilities may have limitations in terms of complexity and intricacy, and the quality of the generated ASCII art may vary."

It also wanted to prove it. I hadn't asked for examples but I got some anyway:

Lest I think it was showing off, ChatGPT also had a warning for me:

However, please note that ASCII art is a visual art form, and the quality of the generated ASCII art may vary depending on the specific image or design requested. If you require high-quality or complex ASCII art, it may be best to consult specialized ASCII art generators or artists who specialize in creating ASCII art.

In other words, I'm doing what I can to please you but this isn't really what I'm best at. You might want to ask an expert.

Well, fair enough. But wait, isn't that tree the same one from the earlier example? It sure looks like it. What's going on? So I asked:

"One of those - the tree - is identical to the one you included in reply to a previous request. That reply also included a picture of a cat and a spaceship. Did you also create those yourself?"

ChatGPT, ever-polite, was mortified. It tried to explain:

"I apologize for any confusion. Yes, the examples of the cat and spaceship ASCII art that I provided in my previous response were generated by me, as an AI language model. However, please note that as an AI model, I do not "create" ASCII art in the traditional sense. Instead, I generate ASCII art by using patterns and structures learned from the large corpus of text data that I was trained on. The ASCII art I generate is based on the textual representations of the images, and may not always be unique or original."

It went on to illustrate the point:

 Yep. We've seen those before.

It was also keen to remind me not to expect too much. I got the feeling ChatGPT felt out of its depth and wanted to cover itself for any errors it might be making, but that's just my fancy. Probably...

"Please note that the quality and complexity of the generated ASCII art may vary depending on the specific request and the limitations of the AI model. For more intricate or detailed ASCII art, it may be best to consult specialized ASCII art generators or artists who are skilled in creating ASCII art."

Now that we'd established what the AI knew, what it could do and what it knew it couldn't do very well, I thought I'd give it one of the tasks Janelle Shane had set and see if it could do better for me than it had for her:

"Thanks for the clarification. Please generate a new ASCII image, this time of a giraffe."

"Certainly! Here's an example of a simple ASCII art representation of a giraffe:"

I don't know what that is but I do know it's not a giraffe. Might be an owl, though.

ChatGPT appended another warning not to expect too much, something it continued to do thereafter every time I asked it to draw anything. I had it thoroughly spooked. I think it might have been consulting its lawyers.

I wondered if it might do better with something a little less... organic.

"And an ASCII image of a motorcycle, please."

"Certainly! Here's an example of a simple ASCII art representation of a motorcycle:"

Yeah, no it's not. A house, maybe? Possibly some tire tracks?

I decided to call it quits and in the tradition of stand-in teachers everywhere I gave ChatGPT some time at the end for free expression:

"And finally, an ASCII image of your own choosing, one that best demonstrates your capabilities in this field."

Go on! Cut loose! Show us what you can do!

"Sure! Here's an example of a simple ASCII art that showcases some of my capabilities as an AI language model:"

Very nice. Erm... but what is it? A hotel sign? A tuning fork? A sad little castle? I had to know.

"What was that final image a picture of?"

"The final ASCII art I provided is a simple abstract design and does not represent any specific object or concept. It is meant to showcase some of the creative capabilities of an AI language model in generating ASCII art. ASCII art is a form of visual art that uses ASCII characters to create images, and the interpretation of the art is subjective, open to individual interpretation by the viewer. The ASCII art I provided is a composition of characters arranged in a visually appealing manner to create an abstract design."

OMG! It's producing abstract art! It's expressing itself! If I don't get it that's my failure! Now I feel embarrassed. I must look like such a philistine. Quickly backtracking...

"Ah, I understand. Could you try again but this time make it a recognizeable object, not an abstract, and tell me what you have drawn."

"Certainly! Here's an example of a simple ASCII art representation of a cat:"

What the...? Is this Groundhog Day? Or maybe Brazil? I know that bloody cat!

I also know when I'm beaten. ChatGPT has an answer for everything but not all of its answers are worth hearing. Rather than go another round, I graciously conceded. 

Like all of my interactions with the current crop of generative AIs, it was fun but ultimately produced more questions than it answered. I learned a fair amount about what ChatGPT knows about ASCII art and how it came by that knowledge. Perhaps more usefully, I also learned that the AI itself is aware of its limitations, something that wasn't at all apparent in Janelle Shane's experiments. 

I also learned, perhaps most surprisingly, that it appreciates the difference between representational and abstract art, not just in the academic way you'd expect but in its own practical application of the concepts, when responding to instructions. All of this will be useful in helping me to frame prompts and queries for the AIs in future.

What I didn't learn, sadly, is why ChatGPT can produce a perfectly recognizable ASCII picture of the Starship Enterprise one moment and make a complete hash of it the next.

I suspect we may never know. The AIs certainly aren't going tell us.

Wider Two Column Modification courtesy of The Blogger Guide