I, ChatGPT

DogsB

https://www.youtube.com/watch?v=OWIxzE2D7Xk

I think I might die laughing.

sockpuppet7

@boomzilla said in I, ChatGPT:

@Carnage said in I, ChatGPT:

@DogsB said in I, ChatGPT:

Kyle Orland / May 13 / AI

Disarmingly lifelike: ChatGPT-4o will laugh at your jokes and your dumb hat

It's amazing what a few well-placed chuckles and vocal tone shifts can do.

How long will this hype cycle last? They do appear to be getting shorter.

Everyone is still praising the emperors clothes and ignoring the boys saying he's naked, but the house of VC cards might collapse soon-ish I hope.

As soon as there's another bubble to pump up.

Programmers are also human nailed it once again:

*Everyone on AI twitter 2023*

https://youtu.be/sxdDVAv6QDY?si=xbHwQNLT4iwnzCCi

I was trying to find where I got linked to this channel, this guy is so fun, I couldn't stop watching his videos

HardwareGeek

@Bulb said in I, ChatGPT:

@Zecc I think it mainly makes difference in social intelligence. Dogs live in tight packs that cooperate and that requires a lot of brain-power. Cats are probably a little better in planning their own individual actions, but learning to hunt or fight in a coordinated pack is way out of their league.

I remember reading somewhere^{[citation needed]} that dog have intelligence roughly equivalent to a 3-yo child (by whatever sort of intelligence measurement one could use for such a comparison). Cats are equivalent to a 2.5-yo child, IIRC. I might be a bit off on the numbers, but whatever the actual numbers, a dog was equivalent to a slightly older, more mature child. If you can believe such quantitative measurements of intelligence.

HardwareGeek

@DogsB said in I, ChatGPT:

https://www.youtube.com/watch?v=TGfQu0bQTKc

This guy is gold.

I think that alone is enough humor for me; I won't bother watching the video.

dkf

@HardwareGeek said in I, ChatGPT:

@Bulb said in I, ChatGPT:

@Zecc I think it mainly makes difference in social intelligence. Dogs live in tight packs that cooperate and that requires a lot of brain-power. Cats are probably a little better in planning their own individual actions, but learning to hunt or fight in a coordinated pack is way out of their league.

I remember reading somewhere^{[citation needed]} that dog have intelligence roughly equivalent to a 3-yo child (by whatever sort of intelligence measurement one could use for such a comparison). Cats are equivalent to a 2.5-yo child, IIRC. I might be a bit off on the numbers, but whatever the actual numbers, a dog was equivalent to a slightly older, more mature child. If you can believe such quantitative measurements of intelligence.

It's not really a helpful measure. Dogs are intelligent in a way that is suited to being a dog, and cats are in ways for cats. You can't separate mind from environment. Children of those ages are developing intelligence for humans, which is far more socially loaded than even for dogs. (They're also working hard on language and have a far greater command of that by 3 than dogs ever will.)

Some of the computation in the brain is much more complex than is reachable with a sane number of ANN layers. (It's the pyramidal neurons, I tell you!) This is true of mice, cats, dogs and humans; possibly of all mammals and likely of a lot of other vertebrates as well (dunno about octopuses and squid; theyre molluscs but they seem to have some smarts too). There are some types of worm that we can simulate fully and have been able to for many years, and maybe spiders are simulatable or not. Nobody knows where the line really is.

It's very complicated and difficult!

Zerosquare

@HardwareGeek said in I, ChatGPT:

a dog was equivalent to a slightly older, more mature child.

: More mature? You've not met some of my dogs.

blek

@HardwareGeek said in I, ChatGPT:

Cats are equivalent to a 2.5-yo child, IIRC.

I am now imagining a 2.5 year old child climbing up the back of my chair to perch itself on the top.

homoBalkanus

@blek they all do that

loopback0

@HardwareGeek said in I, ChatGPT:

@DogsB said in I, ChatGPT:

https://www.youtube.com/watch?v=TGfQu0bQTKc

This guy is gold.

Speaking of the intelligence of a 3 year old...

CodeJunkie

@kazitor said in I, ChatGPT:

TIL:

Dare I ask how anyone even comes across these questions?

Hmm.... assuming the highlighted part is accurate (which it probably is) ... who got paid 5 million dollars a year for 10 years to eventually give a yes or no answer to a stupid fucking question? Also, I'm in the wrong damn business ... as usual.

EDIT: I could have milked that project for that long and have been like "yeah sure, why not" after 10 years. (pun intended)

CodeJunkie

@HardwareGeek said in I, ChatGPT:

@DogsB said in I, ChatGPT:

https://www.youtube.com/watch?v=TGfQu0bQTKc

This guy is gold.

I think that alone is enough humor for me; I won't bother watching the video.

The video is brilliant.

EDIT: Check this one out: https://www.youtube.com/watch?v=lE4UXdJSJM4

topspin

@boomzilla said in I, ChatGPT:

Programmers are also human nailed it once again:

*Everyone on AI twitter 2023*

https://youtu.be/sxdDVAv6QDY?si=xbHwQNLT4iwnzCCi

Most times I'm wrong
But when I'm in the right, I refer to that point forever.

Interview with Senior WTDWTF member when?

blek

@topspin Whenever you're free

Arantor

Tarek Ziadé

Experimenting with local alt text generation in Firefox Nightly – Mozilla Hacks - the Web developer blog

Firefox 130 will feature an on-device AI model that automatically generates alt-text for images, integrated into its built-in PDF editor.

topspin

@Arantor I’m sure that’ll be as useful as the automatic description PowerPoint inserts which I then have to manually remove.

DogsB

Gern_Blaanston

People complaining about ChatGPT and AI in general can't seem to make up their mind what their complaint actually is.

This article seems to be a complaint that publishers are now making deals to license their content to AI companies:

Over the past several months, publishing companies representing more than 70 newspapers, websites, and magazines have licensed their content to OpenAI

Ever since AI became the hot, new thing, people have been complaining that AI companies are using copyrighted material without permission or compensation. OK, that's a fair and reasonable complaint.

So now, the AI companies have signed deals where they have gotten permission, and they are paying licensing fees for the copyrighted material they are using for training.

And now people are complaining about THAT. What the fucking fuck. Make up your mind.

topspin

@Gern_Blaanston said in I, ChatGPT:

People complaining about ChatGPT and AI in general can't seem to make up their mind what their complaint actually is.

Knowing about it. Reading about it. This whole plane of existence...
Pick yours.

BernieTheBernie

@DogsB said in I, ChatGPT:

Polish w is in in English.

sockpuppet7

@Gern_Blaanston said in I, ChatGPT:

People complaining about ChatGPT and AI in general can't seem to make up their mind what their complaint actually is.

This article seems to be a complaint that publishers are now making deals to license their content to AI companies:

Over the past several months, publishing companies representing more than 70 newspapers, websites, and magazines have licensed their content to OpenAI

Ever since AI became the hot, new thing, people have been complaining that AI companies are using copyrighted material without permission or compensation. OK, that's a fair and reasonable complaint.

So now, the AI companies have signed deals where they have gotten permission, and they are paying licensing fees for the copyrighted material they are using for training.

And now people are complaining about THAT. What the fucking fuck. Make up your mind.

the copyright was always an excuse, these are common luddites

Arantor

@Gern_Blaanston whose permission is a key question here.

Take SlackOverflow - AI scraping without permission is a problem, and the contributors feel that SO profiting off their work for a use they don’t agree with is also bad. It’s not either/or here.

They don’t really have a leg to stand on, SO contributions are suitably licensed but they can feel upset that it’s for a use they didn’t approve of and don’t appreciate.

Kind of like deviantArt sharing artists’ works without permission, and then doing a dirty on artists to make it hard to not give permission.

topspin

@sockpuppet7 said in I, ChatGPT:

the copyright was always an excuse, these are common luddites

You mean an excuse for the companies that sued you for 3 trillion dollars when it was theirs, but don't care when they break everybody's copyright en masse? Probably.

Bulb

@Arantor said in I, ChatGPT:

SO contributions are suitably licensed

is the output of a LLM derivative work of the training material? Because SO content is licensed CC-BY-SA

izzion

@Bulb said in I, ChatGPT:

@Arantor said in I, ChatGPT:

SO contributions are suitably licensed

is the output of a LLM derivative work of the training material? Because SO content is licensed CC-BY-SA

An LLM is a blender that removes all copyright and trademarks and gives the AI company the ability to do whatever the hell they want without legal penalty _{because they don't ingest any license agreements into the training material}

sockpuppet7

@izzion said in I, ChatGPT:

@Bulb said in I, ChatGPT:

@Arantor said in I, ChatGPT:

SO contributions are suitably licensed

is the output of a LLM derivative work of the training material? Because SO content is licensed CC-BY-SA

An LLM is a blender that removes all copyright and trademarks and gives the AI company the ability to do whatever the hell they want without legal penalty _{because they don't ingest any license agreements into the training material}

I don't think that is how the law should be applied, for pragmatic reasons. You'd have a hard time trying to prove what you used to train your models. I think you must judge the final product as if it was produced by hand.

I have the all tools to print a perfect clone of mickey mouse with whatever I can access on the internet. It's me using this tools to produce some material that would get me copyright problems. The same should be applied to AI, it's a worth of energy trying to argue how a model was trained, go after someone that make fake mickey mouses with it

boomzilla

My Facebook feed is absolutely flooded with crap from AI picture groups / accounts "recommended for me." It's all women, in various states of undress, though mostly at least in bikinis. Lots of actresses, volleyball players, women "breastfeeding."

The breastfeeding ones are the most egregious, since they show lots of tit. Went to report some for "adult nudity," and there's an explicit "Except breastfeeding" exception, which explains the genre, though I'm certain this wasn't what they had in mind.

Zerosquare

@boomzilla said in I, ChatGPT:

My Facebook feed is absolutely flooded with crap from AI picture groups / accounts "recommended for me." It's all women, in various states of undress, though mostly at least in bikinis. Lots of actresses, volleyball players, women "breastfeeding."

I'm sorry, but I believe @Polygeekery already used that excuse.

Arantor

Oh Wizards of the Coast, we thought you'd learned.

https://hasbro.eightfold.ai/careers?department=WIZARDS %26 HASBRO GAMING&pid=68740107168&domain=hasbro.com&sort_by=relevance

Reproduction:

Job Title: Principal AI Engineer

Job Description: (emphasis theirs)

At Wizards of the Coast, we connect people around the world through play and imagination. From our genre-defining games like Magic: The Gathering® and Dungeons & Dragons® to our growing multiverse, we continue to innovate and build new ways to foster friendship and connection. That’s where you come in!

We are building a central technology team to create high-value software and processes in direct support of our development teams. We are led by the belief that providing industry-leading tools, innovative technology, and excellent support allows our teams to focus on delivering outstanding player experiences first and foremost!

We’re looking for someone who believes that crafting high-value, shared technology to accelerate creation and testing of content with ML systems engineering is key to delivering high-quality player experiences. To begin building an outstanding team we are looking for a versatile and dedicated Lead Software Engineer, AI/ML. In this role, you will be the most senior engineer within the team and partner with the Technical Director to ensure all of the products at WotC are supported by our engineering efforts. You will lead this new team to explore uses of AI in game development, asset creation and automated frameworks.

What you'll do: (emphasis mine)

Work closely with internal and external teams to define and drive the strategic roadmap for AI integrations in our tools, pipelines, and game products

Be an advocate with vision in AI, working closely with other teams to build roadmaps and timelines for usage across our products.

Explore new uses of AI for all levels of game development.

Design, build, and deploy systems for intelligent generation of text dialog, audio, art assets, NPC behaviors, and real time bot frameworks.

Design and develop libraries for teams to easily consume and deploy AI tools and enhancements.

What you'll bring (emphasis mine)

7+ years experience in an engineering lead capacity.

Proven experience with AI/ML systems in at least one of the following areas: simulation, asset creation, and generative content.

Ability to drive consensus on approaches and solutions for creating consistently high-quality game software.

Knowledge of major advancements in AI, including LMMs, 2D and 3D content creation, and audio generation.

Understanding the ethical ramifications of using AI in our work and being able to speak to how to ensure ethics are embedded in what we do.

Ability to evaluate external software for decision-making on buying or building.

But please tell me again how I'm the one who's a Luddite for being concerned that actual creative people are being pushed out in favour of AI-generated content. This is already after WotC "listened to the complaints" the previous time they were rapped over the knuckles for trying to pass of AI work as artists' work. (And note that I'm not calling for an end to use of AI, nor even a moritorium on it at this point, simply 'can't we even have a meaningful conversation that doesn't begin with defending shitting on people')

Remember: this is a leadership role wanting demonstrated experience in AI content generation, with a focus on using AI content generation in more not less of the development process. So the plan will be to install leadership who uses this tech and encourages everyone below them to do the same.

izzion

@Arantor said in I, ChatGPT:

and encourages everyone below them to do the same.

that's a strange way to spell "compels"

sockpuppet7

@Arantor said in I, ChatGPT:

But please tell me again how I'm the one who's a Luddite for being concerned that actual creative people are being pushed out in favour of AI-generated content.

You don't need to be wrong to be a luddite (I still think you are)

You get extra points for being English

About your point, demanding that the creatives use the most productive tool available isn't pushing them out. The current crop of AI doesn't produce good content without human input

The things I'm playing with are mostly because I found amusing to code it.

Arantor

@sockpuppet7 that makes a huge assumption that volume of output is the primary criteria for a creative person, and that the “most productive tool” is AI-assist.

Bear in mind that you’re talking to someone who uses Sublime in favour of “better” tools such as PHPStorm or VSCode and was still demonstrably more productive than my peers actively using those tools. “Better tools” does not equal more productivity.

Zerosquare

@Arantor said in I, ChatGPT:

still demonstrably more productive than my peers

That may say more about your peers than about you

boomzilla

@Arantor said in I, ChatGPT:

But please tell me again how I'm the one who's a Luddite for being concerned that actual creative people are being pushed out in favour of AI-generated content.

You're a Luddite.

I think we're at the stage where everyone has to come up with some kind of AI strategy because they're afraid they'll be left behind if they don't. I also think that's going to fizzle a bit when it's not the silver bullet people are hoping / fearing it is.

That said, it will upset certain fields more than others. I think the potential for upsetting the visual arts status quo is much higher than anything else, based on what I've seen.

Arantor

@boomzilla you did read the part where I don't... actually... want to stop AI, right?

It's also not so much 'that they're afraid they'll be left behind', we're already collectively leaving people behind. It's not a thing to be afraid of happening in a future but a thing to be happening actively right now.

And the part where you were complaining about how Facebook is flooded with crap? This is the result of people like you complaining about people like me (who, again, don't even want to stop it, but want to talk about the things that are problematic about it and see if we can fix them) and rejecting the debate as 'eh just another Luddite', well, you're welcome, that's your new future. Enjoy.

boomzilla

@Arantor said in I, ChatGPT:

@boomzilla you did read the part where I don't... actually... want to stop AI, right?

Yes, but you also asked to get called a Luddite. Sheesh, make up your mind.

It's also not so much 'that they're afraid they'll be left behind', we're already collectively leaving people behind. It's not a thing to be afraid of happening in a future but a thing to be happening actively right now.

No, companies are afraid they'll be left behind other companies if they lose the "AI race."

And the part where you were complaining about how Facebook is flooded with crap? This is the result of people like you complaining about people like me (who, again, don't even want to stop it, but want to talk about the things that are problematic about it and see if we can fix them) and rejecting the debate as 'eh just another Luddite', well, you're welcome, that's your new future. Enjoy.

I mean...it gets flooded with some weird thing every week or two (for a while it was Coast Guard cooking, of all things). This week's is AI generated soft core porn. It's not an AI problem, per se, just dumb FB algorithm shit that I'm really complaining about.

Zecc

@boomzilla said in I, ChatGPT:

Yes, but you also asked to get called a Luddite.

But you half-assed it.

Here, @Arantor: you're a Luddite for being concerned that actual creative people are being pushed out in favour of AI-generated content.

Zerosquare

1NBSP.

sockpuppet7

@Arantor said in I, ChatGPT:

@sockpuppet7 that makes a huge assumption that volume of output is the primary criteria for a creative person, and that the “most productive tool” is AI-assist.

Bear in mind that you’re talking to someone who uses Sublime in favour of “better” tools such as PHPStorm or VSCode and was still demonstrably more productive than my peers actively using those tools. “Better tools” does not equal more productivity.

You are allowed to like sublime better, but if you say that you can't/won't/refuse to use PHPStorm and VSCode in an interview for a PHP job that will probably reduce your chances

edit: Also, the productivity that matters is measured by long term profit. I believe that AI in the hands of an artist that is already competent without AI will usually improve their output without losing quality (because they already can do it without AI, they can fix or even not use it where it doesnt fit)

Bulb

@Zerosquare said in I, ChatGPT:

Some mushrooms can only be reliably told apart by also looking at the underside. In this case, (sufficiently mature) button mushroom has brown gills whereas the destroying angels have white, at most light pink gills. So this image is not enough for identification. Though the stipe looks too thin and long for a button mushroom, so it most likely is indeed a destroying angel (there is one another very similar species of amanita, but one doesn't need to bother telling those apart since they are both poisonous anyway).

HardwareGeek

@Bulb Forget AI, I wouldn't go foraging with any source of IDs less authoritative than a trained, experienced mycologist, and probably not even then. I buy my mushrooms at the grocery story, TYVM.

DogsB

@sockpuppet7 said in I, ChatGPT:

edit: Also, the productivity that matters is measured by long term profit.

There's no actual proof of that yet and there's already evidence on the forum of more time lost to fixing its output for anything non-trivial.

I believe that AI in the hands of an artist that is already competent without AI will usually improve their output without losing quality (because they already can do it without AI, they can fix or even not use it where it doesnt fit)

You're getting back into correcting the ai output vs just doing it yourself. In the case of artists, it's not just correcting a jpeg. It's building up a series of psds with reusable and editable layers. The "send us the PSD" meme is funny for a reason. AI art is probably a poor starting point because they'll have to spend hours or days building the PSD. The starting resolution probably has to be at least 8k too.

DogsB

@sockpuppet7 said in I, ChatGPT:

the copyright was always an excuse, these are common luddites

I want to see an unfiltered AI with everything in it to see what it creates, but the reality is that if you're not going to enforce copyright and have a way for creatives to make rent with their work, you're not going to have anything worth consuming.

ixvedeusi

@sockpuppet7 said in I, ChatGPT:

I believe that AI in the hands of an artist that is already competent without AI will usually improve their output without losing quality

The big caveat here is that they need to be allowed to spend the time necessary to produce good output, which in turn means they have to be able to sell their work for a price that can pay for their living expenses. The higher the price difference between quality content and crap, the harder it gets for the people producing quality content to compete against the people producing crap.

Considering how the price / quality trade-off has been playing out with web pages (and software in general) recently, I admit I'm a bit concerned.

Bulb

@HardwareGeek said in I, ChatGPT:

@Bulb Forget AI, I wouldn't go foraging with any source of IDs less authoritative than a trained, experienced mycologist, and probably not even then.

Foraging for mushrooms is very common around here, especially among the older folks. Most people can recognize at least some common edible mushrooms and don't pick anything they are not sure what it is.

But you can learn it from a good mushroom atlas. The key is that for each species it list which other species it could be confused with and how to reliably tell them apart—and that often requires looking for features that can't be seen on one snapshot like that.

I buy my mushrooms at the grocery story, TYVM.

Yeah, that's much easier, and the farmed button mushrooms are safe. But button mushrooms are usually the only ones that are cultivated. Other kinds like ceps need to be foraged for.

HardwareGeek

@Bulb said in I, ChatGPT:

But button mushrooms are usually the only ones that are cultivated.

Around here, we can typically buy button, portabello, shiitake, oyster, and maybe sometimes one or two others. I assume all of these are farmed, but I don't really know for certain.

Incidentally, I used to live 10 miles or so from a mushroom farm, and I'd occasionally ride my bike that way. Man, what a stench! The manure in which they grow the mushrooms is very fresh; it may be called compost, but it's not really composted.

Atazhaia

@HardwareGeek said in I, ChatGPT:

@Bulb Forget AI, I wouldn't go foraging with any source of IDs less authoritative than a trained, experienced mycologist, and probably not even then. I buy my mushrooms at the grocery story, TYVM.

I would go out and pick some mushrooms, but this is limited to a very few distinctive species that does not have any similar-looking poisonous "friends" over here. A good example is this: