Coconut Monkey Cornerclub

[Brian Boru]

Quite the story…

Wojtek: The bear who was a private in the Polish army

The bear was enlisted into the Polish army and served alongside soldiers during World War Two.

www.bbc.com

 

[ZedClampet]

We're finally supposed to have our first freeze this evening, so yesterday I spent the morning harvesting my garden for the last time, as well as bringing in my sprinklers, hoses and whatnot.

An alright harvest, but my tomatoes, cantaloupes and peas didn't survive long after planting this year, so mostly just peppers and cucumbers.

https://imgur.com/a/H1KNuYH

The last image shows my efforts this morning. I'm pickling the less ripe peppers and making an attempt at dehydrating the over ripe, wrinkled ones in order to grind them into chili powder. I did also harvest my basil, which I will also dehydrate, but it was heavily bolted, so I'm not sure I'll get much out of it.

I emptied the pool lines about a month ago. Probably the last time for that pool. If I get it filled in I'm going to do a garden or maybe a greenhouse. Pool brought us a lot of fun through the years, but it's too much money and work for the use we get out of it since the kids are gone and the nephews and nieces are all grown.

 

[ZedClampet]

In photography, do you know what the instruction "cowboy shot" means? Don't look it up. Just guess if you don't know.

 

[Brian Boru]

"cowboy shot" … Just guess

Dunno, so:

♣ Driving along in passenger seat, whip camera out of pocket and fire off a few at a fleeting opportunity.
♦ Take a naughty candid, then jump on your horse and hightail it.
♥ Training to be a cattle imagery specialist.
♠ Camera hidden in hat.

Now to look it up…

Well that was a big let-down

 

[ZedClampet]

Well that was a big let-down

Only because of your dirty mind. Did you see why it is that way? I wanted to see if people could guess it by thinking about cowboys and old westerns. Since Brian felt it was a letdown, I'll go ahead and tell everyone it's a photo taken from the middle of the thigh to the top of the head. This way, when they were taking photos of cowboys, they always got the holsters and guns in the picture.

I post the same way I give gifts. I make everyone say, wut? Unintentionally of course. That's why I'm only giving cash this year and trying not to post as much

 

[Frindis]

@ZedClampet Do you have a favorite western movie? I love the genre and there are so many I could have put as number one, but the one for me that I always keep coming back to is Unforgiven, directed by the legendary Clint Eastwood. He stars in it, Morgan Freeman and another great actor I like a lot, Gene Hackman. The plot, locations, shooting, actors, props, everything is brilliantly done.

 

[Brian Boru]

🎃 Happy Halloween! 💀​

…

Samhain to Soulmass: The Pagan origins of familiar Halloween rituals

From outrageous costumes to trick or treat: the unexpected ancient roots of Halloween's most popular – and most esoteric – traditions.

www.bbc.com

…

Sex, love and immortality: Behind the obsession with vampires

Mark Gatiss, from BBC’s Dracula, and Rolin Jones, from Interview with the Vampire, on why vampires are so popular.

www.bbc.com

 

[ZedClampet]

@ZedClampet Do you have a favorite western movie? I love the genre and there are so many I could have put as number one, but the one for me that I always keep coming back to is Unforgiven, directed by the legendary Clint Eastwood. He stars in it, Morgan Freeman and another great actor I like a lot, Gene Hackman. The plot, locations, shooting, actors, props, everything is brilliantly done.

There's a bunch. One thing I know for sure is that one of the most iconic movie songs, if not THE most iconic movie song, of all time comes from a western:

View: https://www.youtube.com/watch?v=enuOArEfqGo


Is there anything that say western more than that? It's brilliant.

🎃 Happy Halloween! 💀​

Sex, love and immortality: Behind the obsession with vampires

Mark Gatiss, from BBC’s Dracula, and Rolin Jones, from Interview with the Vampire, on why vampires are so popular.

www.bbc.com

Whenever I play V Rising, I have my human blood slaves that I keep in cages, and I never take enough blood to actually kill them. But when I'm finished with a game, the last thing I do is go to them and turn them into vampires and let them go. My gift for their service. I figure that makes us good. Don't owe them anything after that.

*****
I've got two months left sitting on the board of the literacy charity then I'm FREEEEEEE! Or I would be, but now I'm going to be the Finance Chair and board treasurer I'm a sucker, yep...

However, one thing I want to do is to start delivering meals once a week to elderly and handicapped folks for Mobile Meals. I found out last week that my kids are doing it together in their little college town, which makes me proud, and I've decided to look into it here. I heard on the radio they needed more drivers. My mom did it when I was little, and I, being a boy, used to get so hungry smelling those hot meals. I never got to eat one haha.

Don't worry. I'm not a good person. I'm just bored.

 

[Brian Boru]

one of the most iconic movie songs

I saw the title in the video pic and was all set to extol the virtues of The Danish National Symphony Orchestra version—and that's the one you picked! Gotta love the Wah-wah-wah girl

I know John Williams is The Man, but I prefer Ennio Morricone scores—put him with Sergio Leone's style and you get a sensual feast. The Dollars movies were so refreshing at the time, and the very different Once Upon a Time in the West is also one of my favs.

The Magnificent Seven deserves credit for one rare achievement—being a Hollywood adaptation of a foreign classic …Kurosawa’s Seven Samurai… which isn't a dog's dinner. For me TMS is the more enjoyable of the two, more lighthearted and accessible.

For a little change of pace, i want to see the animated Western Rango sometime—the voice cast is ridiculous, @Frindis it even has the inimitable Aeryn Sun in it!

 

[Brian Boru]

Can AI do what the monkeys can't?

'Infinite monkey theorem' challenged by Australian mathematicians

Australian researchers have poked holes in an old thought-experiment known as the "infinite monkey theorem".

www.bbc.com

…

Aah, recognition at last—all bow:

Charli XCX's Brat crowned Collins Dictionary word of the year

It describes a person who is "characterised by a confident, independent and hedonistic attitude".

www.bbc.com

 

[ZedClampet]

Before I tell you something, let's review ME

I constantly admit to being bad at games.
I constantly admit to being an absolute moron.
I'm the god-king of maudlin.
I am constantly admitting to all sorts of negative things.

So now I would like to say that I'm better at training diffusion LoRAs than just about any damn body. I've yet to find anyone better.

Feels good. Won't last, though. There's something I don't know and it's all going to blow up. At least that's my assumption.

 

[Brian Boru]

^ You started so well…

diffusion LoRAs

"Diffusion"—so sth to do with light maybe:
Lenses of Radical Auras?
League of Refractive Alliances?
Lessons on Reflective Aluminum?

I give up 🤷‍♂️

 

[ZedClampet]

^ You started so well…



"Diffusion"—so sth to do with light maybe:
Lenses of Radical Auras?
League of Refractive Alliances?
Lessons on Reflective Aluminum?

I give up 🤷‍♂️

In AI, diffusion models learn by taking pictures and adding noise to them until the entire screen is nothing but static, and from that, and with endless training, they learn to reverse the process and take static and whittle away noise until they have a picture. Every picture I've posted started out looking like the screen of an old television that's on an empty channel--just static. And then AI just removed stuff until nothing was left but the picture.

There are other methods, for instance, adversarial, where two AI argue until they've made a picture (to vastly over-simplify it)

Adversarial method is also now being used to attack AI. It's a new cybersecurity concern. Malicious AI subverting "good" AI.

 

[ZedClampet]

Be careful letting your significant other around your computer cables.

I don't know. Maybe it's hindsight, but I feel like your first impression of this woman should be, "She might kill me."

 

[ZedClampet]

I have a problem. The way I naturally use some websites makes it seem like I'm a bot. Both Pinterest and Instagram have limited my access. I guess I need to come up with a new technique.

 

[Brian Boru]

The way I naturally use some websites makes it seem like I'm a bot

🟠

Please half-fill the circle with the appropriate color for your team.

 

[Brian Boru]

"If it weren’t for the noble efforts of Brian Boru we’d all be wearing “Kiss me I’m Danish” t-shirts."

Boru - Age verification

 

[ZedClampet]

I don't think I told this crazy story that happened here a few weeks ago. Just down the road from us is a metal recycling center. A lot of the metal comes boxed. So this big box had been sitting there for a couple of weeks, and an employee got a blow-torch and set it on fire. Basically, they burn the box off and then grab the metal with a forklift. But this box reacted differently. The box was burning, but whatever was inside was smoldering. The guy opened the box to see what was smoking and found 200 pounds of dynamite. He somehow put out the fire in the box, which I think was a pretty heroic thing to do. I think most people would have got the heck out of there. Anyway, the police came and made everyone within half-a-mile evacuate. The TNT was determined to be commercial grade, and to make matters worse, it had been packed along with the blast caps, which is against the law.

ATF let it sit for over 24 hours and then packed it up and disposed of it. Not sure what the procedure is for that. Federal authorities said they were going to figure out where it came from, but I haven't heard anything. I just asked on a local forum. If someone knows the answer I'll post it just to wrap up the story.

 

[ZedClampet]

PSA

If you know a young person who suffers from depression and anxiety or even worse, introduce them to Ren. It could save their life.

View: https://www.youtube.com/watch?v=s_nc1IVoMxc

 

[BeardyHat]

PSA

If you know a young person who suffers from depression and anxiety or even worse, introduce them to Ren. It could save their life.

View: https://www.youtube.com/watch?v=s_nc1IVoMxc

Good song, thanks. Is that Spanish Guitar? I've always loved that sound.

I've been struggling with my depression again for several weeks now, but at this point in my life I've been through it so many times, it doesn't bother me so much. I know I'll have a period of happiness again at some point when the pendulum swings back for me.

But the harder struggle is anxiety, which I've had in leagues the past several days. It's so much harder to cope with, especially when it's been waking me in the middle of the night as it has recently.

 

[neogunhero]

Finally got beta access to Apple’s AI image generator.

A few interesting things: one is that it encourages you to use a real persons face. I’ve found it nearly impossible to generate a random human without being asked to choose a photo as a starting point. I’m not a huge AI connoisseur, I've only used ChatGPT, Copilot and Stable Diffusion, but I don’t think it’s common for an AI to encourage you strongly to use a real persons photo. I thought we were having major issues with that exactly.

Second, which is probably a good counter to the first point, is that it is so strict with your prompts. For example, it will not accept the words/phrases “zombie”, “holding hands”, “wrestling”, etc. It would not take “two guinea pigs holding hands”, but it would take “two guinea pigs”. They’ve put some of the strongest guardrails in any AI I’ve used. I think it is good though since they keep asking to use a photo of a real person, it makes it harder to put them in provocative situations.

Third, it’s very fast since it is being generated on-device. As it generates a prompt, you can swipe through them, and it will keep generating more and more until it reaches about 10 or so. However, each photo is a slight riff off of the first image generated so there isn’t too much differences between each one.

Lastly, the overall interface is pretty nice. You can start with a photo or type a prompt, and it gives a bunch of suggestions to throw into it. It practically takes up to 6 separate prompts to create one image, as long as it makes somewhat sense. For example, I started with “silly monkey playing on computer” and it generated it. It then gave a “sunglasses” option, which gets added to the circular array of prompts, and boom the money has sunglasses. It gives a small amount of different things to add on top of your main prompt.

Overall the main appeal to me is just how it is done on-device. This is the first time I’ve used AI that doesn’t generate the image over the internet, so that alone is really neat to me. It is by far not the best for creating lifelike images. As you can see in the pics I’ll post below, the general aesthetic is quite cartoonish, perhaps for a reason since you’re encouraged to use real people’s faces. It’s a fun little toy more than it is a practical image generator.

Spoiler: Apple Intelligence Images

 

[ZedClampet]

Good song, thanks. Is that Spanish Guitar? I've always loved that sound.

I've been struggling with my depression again for several weeks now, but at this point in my life I've been through it so many times, it doesn't bother me so much. I know I'll have a period of happiness again at some point when the pendulum swings back for me.

But the harder struggle is anxiety, which I've had in leagues the past several days. It's so much harder to cope with, especially when it's been waking me in the middle of the night as it has recently.

Anxiety is the worst. I've had my share of that. Hopefully it will pass quickly for you.

Finally got beta access to Apple’s AI image generator.

A few interesting things: one is that it encourages you to use a real persons face. I’ve found it nearly impossible to generate a random human without being asked to choose a photo as a starting point. I’m not a huge AI connoisseur, I've only used ChatGPT, Copilot and Stable Diffusion, but I don’t think it’s common for an AI to encourage you strongly to use a real persons photo. I thought we were having major issues with that exactly.

Second, which is probably a good counter to the first point, is that it is so strict with your prompts. For example, it will not accept the words/phrases “zombie”, “holding hands”, “wrestling”, etc. It would not take “two guinea pigs holding hands”, but it would take “two guinea pigs”. They’ve put some of the strongest guardrails in any AI I’ve used. I think it is good though since they keep asking to use a photo of a real person, it makes it harder to put them in provocative situations.

Third, it’s very fast since it is being generated on-device. As it generates a prompt, you can swipe through them, and it will keep generating more and more until it reaches about 10 or so. However, each photo is a slight riff off of the first image generated so there isn’t too much differences between each one.

Lastly, the overall interface is pretty nice. You can start with a photo or type a prompt, and it gives a bunch of suggestions to throw into it. It practically takes up to 6 separate prompts to create one image, as long as it makes somewhat sense. For example, I started with “silly monkey playing on computer” and it generated it. It then gave a “sunglasses” option, which gets added to the circular array of prompts, and boom the money has sunglasses. It gives a small amount of different things to add on top of your main prompt.

Overall the main appeal to me is just how it is done on-device. This is the first time I’ve used AI that doesn’t generate the image over the internet, so that alone is really neat to me. It is by far not the best for creating lifelike images. As you can see in the pics I’ll post below, the general aesthetic is quite cartoonish, perhaps for a reason since you’re encouraged to use real people’s faces. It’s a fun little toy more than it is a practical image generator.

They aren't the first to go the draconian prompt police route. CoPilot wouldn't have made that picture I used for the monthly post because there was a child in it. At least it wouldn't have some time back. I haven't tried to use it in awhile. I've got Flux on my own PC now, so there aren't any prompt restrictions on that that I'm aware of, but that's because it is marketed for commercial use and use with LoRAs, which are additional training to fine tune the model. You can't say "No handholding" and expect a marketing firm to be impressed by your moral code. And you can train it on real people's faces, as well, but you have to have permission from the person. The point of putting real people in is if you wanted one person to be the face of all your advertising efforts, this would allow you to do this. You could have Flo from Progressive holding hands with Mr. Clean! Or, as it happens sometimes, Flo's third leg sticking through Mr. Clean's face.

Edit: I found a training LoRA (it was the first one that came up when I searched "art") and the trigger words for the LoRA are "debauchery, depraved, unhinged, excess". I'm guessing Apple and Microsoft would have heart attacks just thinking about that. It's a LoRA trained on the "Decadence" art movement.

 

[neogunhero]

I've got Flux on my own PC now, so there aren't any prompt restrictions on that that I'm aware of…

How does that work for a normal/gaming PC without dedicated AI hardware? Does it depend on what GPU you have? I have a RTX 2060, so not the most powerful but decent for most 1080p gaming.

And you can train it on real people's faces, as well, but you have to have permission from the person.

With Apple’s AI, I could not find a way to pick the specific photo I wanted. It seems to scan your photos library for any faces it can pick up on, and gives you a random selection of about 8 to choose from. It surely knows what I look like so it keeps suggesting photos of myself, which again is probably a reasonable guardrail. However though just because you can’t choose the specific photo yourself doesn’t mean it won’t pick up any photos of people you may have on your phone without their permission.

EDIT: Actually I looked a bit harder and found the button where you can upload a photo. I’ll test it on many different kinds of photos to see what it can do.

 

[ZedClampet]

How does that work for a normal/gaming PC without dedicated AI hardware? Does it depend on what GPU you have? I have a RTX 2060, so not the most powerful but decent for most 1080p gaming.

The 2060 would work. These aren't going to be as fast as the Apple thing. The Apple has a tiny model compared to Flux/Stable Diffusion. In fact, the reason "holding hands" is prohibited may actually be because it can't do it even if it wanted to. Hands are the hardest thing for AI.

But the exact speed will depend on the model you choose. Streamlined models like you have on your Apple are available, but most people use the best models available and just wait longer for the pictures.

Anyway, I recommend 32 GB of RAM. If you only have 16, it uses it all and then you can't do much while waiting for your picture. Plus, it takes longer if you only have 16. On my laptop with 32 GB it takes about a minute to make a picture. But on the one with 16, it takes twice as long.

Other things regarding speed that you might want to know: The size of the picture makes a huge difference in the speed AND in what picture the AI draws. You can always make a small picture and have the AI enlarge it afterward, but if it's refusing to draw exactly what you requested, you may have to enlarge the working area for it and make a bigger picture. Also, the number of passes it makes per pixel make a big difference, and I suggest going with the lowest number that provides you the clarity that you are happy with, usually this is 15 to 20 passes.

Also, the first picture you make after firing up the program can take 3-5 minutes. But after that, it speeds up. If you load a custom LoRA, the first picture made with that may take a little longer. I'm guessing it's filling up your RAM with model images.

As for LoRA, it depends on which diffusion model (Flux or the new Stable Diffusion 3.5) and how much VRAM you have as to whether you could train your own, but there are also plenty of sites that will let you train a LoRA for cheap on their computers.

Edit: On the times, remember I'm using crappy laptops.

 

[ZedClampet]

By the way, there's a local image generator included with the Corel bundle you can get on Humble. I haven't tried it.