The AI discussion thread

[Red Squirrel]

I sometimes have fun with that, it can't edit images you give it yet but it will edit ones it generates. I assume at some point it will be able to edit any image. That could be really cool if you want to change a IRL image like change the background etc.

There's this meme going around where Mark Carney is portrayed in heroic situations and it's pretty funny the stuff some people have come up with.

 

[Kaido]

Sesame has a new conversational AI...absolutely insane:

https://www.tiktok.com/video/7476560723951373573

 

[[DHT]Osiris]

So, uh, AI-generated classic sheet music...isn't bad?

https://twitter.com/x/status/1896247944124539343

Maybe I don't like classical music but that sounds terrible. There's no flow, it's just random instruments playing semi random notes with inconsistent start/stops.

 

[Kaido]

Maya & 3ds Max Developer Autodesk Fires 1,350 Workers to Accelerate Investments in AI:

Around 9% of the company's entire workforce have lost their jobs.

Maya Developer Autodesk Lays Off 1,350 Workers to Invest in AI

Around 9% of the company's entire workforce have lost their jobs.

80.lv

Autodesk Upsets 3D Artists by Introducing a New Generative AI

"Wait you know what your customers do for a living right?"

80.lv

 

[Kaido]

"World Wide Wasteland":

Google's new AI mode for search might turn the Web into a World Wide Wasteland

Google's experimental AI mode pushes search toward a closed ecosystem, mirroring OpenAI's ChatGPT search and Perplexity while potentially delivering a fatal blow to the open web.

the-decoder.com

Recent research from content-licensing company TollBit shows AI search engines like OpenAI and Perplexity reduce referral traffic to news sites and blogs by 96% compared to traditional Google search.

 

[MrSquished]

So the MAGA owner of the Los Angeles times decided to have an AI automatically give an opposing viewpoint to any opinion column. Shortly after that, the AI defended the KKK (basically taking the Trump party side) so they have disabled that feature for now lol When you go MAGA, you go racist, and AI knows it. Amazing!

MAGA Newspaper Owner’s AI Bot Defends KKK

An AI-generated summary tried to offer “different views” on the hate group.

www.thedailybeast.com

 

[Kaido]

So the MAGA owner of the Los Angeles times decided to have an AI automatically give an opposing viewpoint to any opinion column. Shortly after that, the AI defended the KKK (basically taking the Trump party side) so they have disabled that feature for now lol When you go MAGA, you go racist, and AI knows it. Amazing!

MAGA Newspaper Owner’s AI Bot Defends KKK

An AI-generated summary tried to offer “different views” on the hate group.

www.thedailybeast.com

Apple added AI News to their news app & had to yank it because it was generating fake headlines:

https://www.cnn.com/2025/01/16/media/apple-ai-news-fake-headlines/index.html

Many problems with AI these days:

1. Fake news being seen by millions if not billions of people
2. Hateful content being pushed out into public, which helps gives people harboring those negative beliefs permission to be vocal about it & act
3. AI images can be photorealistic
4. AI video can be nearly photorealistic; many people (ex. grandma on Facebook) cannot tell the difference

This is a fake AI video...most of the comments think it's real:

FACT CHECK: Video Showing Nancy Pelosi Falling On House Floor Is AI-Generated

A video shared on X purports to show Speaker Emerita Nancy Pelosi falling down while using a walker on the House floor during the recent certification of the 2024 presidential election. Verdict: False The video has been generated with artificial intelligence (AI) and was originally shared on X...

www.aol.com

We've been in 1984 territory for a long time:

Facebook–Cambridge Analytica data scandal - Wikipedia

en.wikipedia.org

The app harvested the data of up to 87 million Facebook profiles. Cambridge Analytica used the data to analytically assist the 2016 presidential campaigns of Ted Cruz and Donald Trump.

Today's reality:

1. Data is the new currency
2. The data is used to shape influence
3. The data can be fake & be impossible or near impossible to detect (headlines, images, videos, etc.)

Live recreation requires just 4 photos now:

https://twitter.com/x/status/1896805188335522143

Scary times...

 

[Kaido]

Alibaba bringing the heat:

https://www.cnn.com/2025/03/06/tech/china-alibaba-ai-model-deepseek-hnk-intl/index.html

Alibaba touted its new model, QwQ-32B, in an online statement as delivering “exceptional performance, almost entirely surpassing OpenAI-o1-mini and rivaling the strongest open-source reasoning model, DeepSeek-R1.” OpenAI-o1-mini is the American company’s cost-efficient reasoning model released last year.

Alibaba added the model has achieved a “qualitative leap in mathematics, coding, and general capabilities, with overall performance on par with DeepSeek R1,” it said in the statement.

Last week, Alibaba pledged to invest at least 380 billion yuan ($52.4 billion) in its AI and cloud computing infrastructure over the next three years. It said the amount exceeded what it had invested in those areas over the past decade.

 

[Kaido]

Real-time rendering with micro-expressions:

https://twitter.com/x/status/1897721581939679350

 

[William Gaatjes]

Singing : "We are the world".

Nice piece of AI generated imagery :

 

[Chaotic42]

I asked ChatGPT 4 Turbo to create word art of the word "benzene" in a circle, like a ring. It couldn't do it. It never spelled it correctly once even though I repeatedly told it to check the spelling, and most of them weren't even symmetrical, let alone circular. It kept apologizing and saying it was definitely going to get it right before responding, and it never did.

If this is supposed to be marketing for how great a product it is, they need to put some work in. I don't think it's ever given me a decent response to anything that isn't remarkably trivial.

 

[William Gaatjes]

Maya & 3ds Max Developer Autodesk Fires 1,350 Workers to Accelerate Investments in AI:

Maya Developer Autodesk Lays Off 1,350 Workers to Invest in AI

Around 9% of the company's entire workforce have lost their jobs.

80.lv

Autodesk Upsets 3D Artists by Introducing a New Generative AI

"Wait you know what your customers do for a living right?"

80.lv

I fear for the day AI is introduced to draw electronic schematics and printed circuit boards by inputting text. That is going to be a mess, to debug.
I hope that day never comes. But i know better. I dread thy futurah where companies will only be presented by "the man" -agers that input text in AI textboxes. Dirie Dirie Dire Dread man.

 

[igor_kavinski]

I dread thy futurah where companies will only be presented by "the man" -agers that input text in AI textboxes. Dirie Dirie Dire Dread man.

A fool with a powerful tool just becomes a more powerful fool, i.e. a potentially more dangerous fool.

 

[RnR_au]

I asked ChatGPT 4 Turbo to create word art of the word "benzene" in a circle, like a ring. It couldn't do it.

Try using a dedicated image generator. I believe there are some out there that are much better at generating words in images than others.

It kept apologizing and saying it was definitely going to get it right before responding, and it never did.

LLM's tend not to be aware of what they can do and can't do. They will keep generating text at you as long as you keep asking it to. AI agents will have more smarts, but only because they have been coded by humans to satisfy certain requirements in a niche field.

If this is supposed to be marketing for how great a product it is, they need to put some work in. I don't think it's ever given me a decent response to anything that isn't remarkably trivial.

You may also have been using AI models wrong. There are plenty of folk in the software industry who are alarmed at how good these can be, used correctly ofcause.

Edit: this is what I got when I just used the free chatgpt... prompt: "create word art of the word "benzene" in a circle, like a ring"



So it can spell... but these AI's are effectively dreaming these images. They are not designing them as such. They have no understanding of the image themselves.


Edit #2: without words, the AI tools are decent. With the prompt: "generate an image of a happy viking hugging a happy gruffalo"

 

[Kaido]

Edit: this is what I got when I just used the free chatgpt... prompt: "create word art of the word "benzene" in a circle, like a ring"

Ideogram is the ticket for art with text:

https://ideogram.ai/t/explore

 

[[DHT]Osiris]

Ideogram is the ticket for art with text:

https://ideogram.ai/t/explore

Yup, it's all about how they model is trained. Art vs anime art vs steamboat Willy art. Text vs stylized artistic text vs newspaper text. Spatial awareness vs combat spatial awareness vs high speed photography spatial awareness.

 

[Kaido]

Yup, it's all about how they model is trained. Art vs anime art vs steamboat Willy art. Text vs stylized artistic text vs newspaper text. Spatial awareness vs combat spatial awareness vs high speed photography spatial awareness.

We're right on the cusp of photorealistic video:

https://twitter.com/x/status/1899819911234617408

We crossed the photorealistic line for photos earlier this year. Voice chat is about 95% there. Try the Sesame demo:

Crossing the uncanny valley of conversational voice

At Sesame, our goal is to achieve “voice presence”—the magical quality that makes spoken interactions feel real, understood, and valued.

www.sesame.com

 

[Kaido]

Sesame has a new conversational AI...absolutely insane:

https://www.tiktok.com/video/7476560723951373573

Sesame just dropped their CSM: (audio generation AI, not an LLM)

https://twitter.com/x/status/1900304515376799915

Try it here:

Sesame CSM - a Hugging Face Space by sesame

Conversational speech generation

huggingface.co

This is Jarvis!! One of the biggest AI technology releases in 2025 imo!

 

[Kaido]

Yay character consistency achieved:

https://twitter.com/x/status/1900231536336921040

 

[Kaido]

Hedra released Character 3:

https://twitter.com/x/status/1899484291643605283

https://twitter.com/x/status/1899830267998593261

 

[Kaido]

Hedra released Character 3:

https://twitter.com/x/status/1899484291643605283

https://twitter.com/x/status/1899830267998593261

It's incredible:

https://x.com/azed_ai/status/1898048037509075212

https://x.com/blizaine/status/1899643188190556667

 

[Kaido]

It's incredible:

https://x.com/azed_ai/status/1898048037509075212

https://x.com/blizaine/status/1899643188190556667

Create a digital double of yourself:

https://twitter.com/x/status/1900300353696129367

I think we're up to about 90 to 95% vide realism now:

https://twitter.com/x/status/1898428597553246670

 

[Kaido]

Pika Labs has Pikaswaps now. Fashion:

https://twitter.com/x/status/1899625842290155610

Hair:

https://twitter.com/x/status/1900359355884855534

 

[Kaido]

Filmmaking signature style cloning:

https://twitter.com/x/status/1899898381515853839

Wes Anderson:



Denis Villeneuve:



Christopher Nolan:



Quentin Tarantino:



Terrence Malick:



These were all generated on a computer

 

[Kaido]

AI enhanced textures on the new Half-Life 2 RTX release:

NVIDIA RTX Remix Officially Released with DLSS 4 and RTX Neural Shaders, Half-Life 2 RTX Playable Demo Available March 18th

NVIDIA’s RTX-powered modding platform has been officially released, adding DLSS 4 with Multi Frame Generation and a full suite of RTX Neural Rendering technology to multiply performance, enabling modders to bring more beautiful worlds and characters to their ray traced remasters of classic titles.

www.nvidia.com