The AI discussion thread

[Kaido]

Repainting 3D assets. Scary scary SCARY amazing stuff:

https://twitter.com/x/status/1706637837775368621

For anyone who has ever done any texture-mapping or 3D painting (Adobe Substance, ArmorPaint, Zbrush, VR, etc.), I mean...dang:

Breathing New Life into 3D Assets with Generative Repainting

www.obukhov.ai

It's somewhat rudimentary at this point, but like everything else in AI, we're gonna see some RAPID acceleration of this technique of applying the 2D generative power of stuff like Stable Diffusion to 3D models:



Imagine how cool this will be for upscaling old games! For example, there's a fan-made remake of Half-Life 1 called Black Mesa that took 15 YEARS to make!

Save 75% on Black Mesa on Steam

Relive Half-Life.

store.steampowered.com

 

[Kaido]

Make children's coloring books using AI:

https://twitter.com/x/status/1708446481064267919

 

[dank69]

AI creates simulated video game development company that develops actual video games:

Now working on creating a sim game where the sims do the same inside the game they create. Sims all the way down.

 

[Kaido]

There's a guy I follow on Twitter (@MrUgleh) who does cool art using "controllism" mashed up with A.I.:

https://twitter.com/MrUgleh

Original post from reddit:

https://www.reddit.com/r/StableDiffusion/comments/16ew9fz

He figured out how to have A.I. grow art into patterns:

View attachment 86260

Shapes in motion to AI video code cracked:

https://twitter.com/x/status/1710343775392829636

 

[Kaido]

I was super interested in animation in high school, but ended up on the filmmaking side of things. AI is going to be...unbelievable for future productions:

 

[JEDI]

What happens to society when you can't distinguish AI videos with real videos?

Will it be like Ghost in the Shell where videos are no longer admissible in court?

We're getting to that point real fast, like in a few years

 

[igor_kavinski]

Will it be like Ghost in the Shell where videos are no longer admissible in court?

Hopefully, people will be plugged into the Matrix by then.

Crime rate will drop to zero.

 

[Captante]

 

[Kaido]

What happens to society when you can't distinguish AI videos with real videos?

Will it be like Ghost in the Shell where videos are no longer admissible in court?

We're getting to that point real fast, like in a few years

Check out the latest AI language translation stuff:

1. Replicates your voice
2. Language translation
3. Video face morph to match lip-syncing

English:

https://twitter.com/x/status/1710485964307706035

AI-translated into pretty decent Spanish!

https://twitter.com/x/status/1710488080761950265

 

[Kaido]

Another peek into the future of AI-assisted independent filmmaking:

1. Shoot the video on your smartphone
2. Apply an AI video effects overlay
3. Apply Deepfake facial reconstruction
4. Apply AI voice-matching technology

Very, very early stuff but SUPER promising!

https://twitter.com/x/status/1710647097991504126

 

[Kaido]

Shapes in motion to AI video code cracked:

https://twitter.com/x/status/1710343775392829636

1. Animated vector shape
2. AI video overlay
3. Animation applied

I'm constantly amazed at how fast iterative improvements are happening in the world of AI. Just bonkers:

https://twitter.com/x/status/1710737499688219072

 

[Kaido]

Another peek into the future of AI-assisted independent filmmaking:

1. Shoot the video on your smartphone
2. Apply an AI video effects overlay
3. Apply Deepfake facial reconstruction
4. Apply AI voice-matching technology

Very, very early stuff but SUPER promising!

https://twitter.com/x/status/1710647097991504126

The workflows for video to animation are getting pretty solidly established. Next improvement will be quality:

https://twitter.com/x/status/1711124369911509133

Using multiple control nets. This stack was:

1. Depth
2. OpenPose
3. MLSD (for room structure)
4. Revision_ClipVision

https://twitter.com/blizaine/status/1711124375506715113

 

[[DHT]Osiris]

Gonna have to hook AI up to the Internet to keep it from getting tricked into doing the illegal.

https://twitter.com/x/status/1710321733184667985

 

[Kaido]

Adobe Firefly 2 has been released: (web-based AI generator)

The future is Firefly: Unlock new levels of creativity with the latest generative AI innovations | Adobe Blog

What’s next for Adobe Firefly and the latest generative AI innovations.

blog.adobe.com

Adobe is building high-quality, USER FRIENDLY AI tools. For example, it now has camera tools instead of just text prompts:



They are staying on top of all of the major developments:



The photorealism is incredible:

link to artist



link to artist



link to artist



They've finally figured out fingers:

link to artist



link to artist



Illustration work:

link to artist



link to artist



The technology is going into the other apps as well, including Text to Vector Graphic AI art generation in Illustrator:



Brief video on AI Generative Fill & Expand in Photoshop:

https://twitter.com/x/status/1712120708325929245

 

[Kaido]

A good thread on ChatGPT 4 for video projects (may need to login to Twitter to see the full thread). "CapCut AI will automatically generate a script, source matching stock videos or photos, and create a video for you."

https://twitter.com/x/status/1711728992548470823

 

[Kaido]

Make a video game using DALL-E 3 for art & ChatGPT 4 for programming:

https://twitter.com/x/status/1712179990144426442

 

[JEDI]

I think I need to take a graphic artists AI class

 

[JEDI]

I feel lost and confused with AI, kinda like how my parents felt with computers in the 1990's.

it was only with the advent of smartphones (2010's) that my parents finally felt somewhat comfortable with computing devices. 20yrs...

Will it be 2040's for me? and what will be the catalyst?
(Besides Skynet taking over the world?)

 

[Kaido]

I feel lost and confused with AI, kinda like how my parents felt with computers in the 1990's.

it was only with the advent of smartphones (2010's) that my parents finally felt somewhat comfortable with computing devices. 20yrs...

Will it be 2040's for me? and what will be the catalyst?
(Besides Skynet taking over the world?)

AI is a crock. Here's the scoop:

1. The whole setup is the standard if/then relationship
2. Then it vacuums up all the data in the world
3. Then it cobbles something neat together when prompted!

The neat thing comes in the combinations of:

1. Inputs
2. Processing
3. Outputs

We now have multi-modal inputs & outputs:

1. Text
2. Images
3. Videos

Three things are driving growth now:

1. The convenience & power of the inputs (prompts such as text, images, video, motion capture, as well as visual slider tools, etc.), i.e. making it easier to use. For example, Microsoft added Copilot to Office 365, so now you can "chat" with the AI engine to ask it for what you want & tweak it down to what you need!
2. The types of output (text, photo, video, real-time chat, audio, etc.)
3. The quality of the processing (the various AI engines are literally getting smarter & more detailed every day!)

For example, Adobe just released Firefly 2, which has a couple key features:

1. Very easy to use
2. Includes hyper-realistic output

These are sample generative image outputs from the latest software release (not real photos, all are 100% AI-generated!). But once you "see behind the curtain" (re: Wizard of Oz) & understand how things work, it's all the same formula:

1. Inputs (ask the AI engine to do something, in as much detail as possible)
2. Processing (this is on v2 of the Adobe Firefly generative AI engine)
3. Output (hyper-realistic generated imagery)

Current state of affairs:

1. It can craft any image you want
2. We are on the cusp of overcoming the "uncanny valley" for photos. The images below are just right there on the border of being totally realistic!
3. Raster (Photoshop) & Vector (Illustrator) already have generative AI baked in at this point. Draw anything you want, in any color or style you want!
4. It can clone any voice
5. As as well any song. Music. Instruments. Soundtracks. Vocals. You name it!
6. Video is still in the early stages, but it's RAPIDLY advancing! Motion capture, CGI, video editing, etc.

 

[Kaido]

Concept art generation:

https://twitter.com/x/status/1713877831594435047

Incredible detail:

https://twitter.com/x/status/1713904476313063446

 

[GodisanAtheist]

What happens to society when you can't distinguish AI videos with real videos?

Will it be like Ghost in the Shell where videos are no longer admissible in court?

We're getting to that point real fast, like in a few years

-Forget court admissible evidence, you can shape public opinion and society as a whole thanks to stuff like this.

No need for bad actors, false flags, non-contextualized images, etc.

Just make the everything up from whole cloth and watch society crumble around you. This is beyond the power of nuclear bombs. We haven't, as a society, even grappled with things like social media's effects on culture and perception. AI is going to upend things in a way the human animal is absolutely not ready for.

 

[Kaido]

-Forget court admissible evidence, you can shape public opinion and society as a whole thanks to stuff like this.

No need for bad actors, false flags, non-contextualized images, etc.

Just make the everything up from whole cloth and watch society crumble around you. This is beyond the power of nuclear bombs. We haven't, as a society, even grappled with things like social media's effects on culture and perception. AI is going to upend things in a way the human animal is absolutely not ready for.

Oh sure. This was 5 years ago:

A low-quality deepfake of Zelesky "surrendering" went viral last year:

AI can now replicate your voice & deepfake lip sync on any uploaded video in tons of different languages:

https://twitter.com/joshua_xu_/status/1687129787267973123

The reality is:

1. People read headlines. Videos are this concept on steroids
2. People suffer from confirmation bias & already subscribe to their news & entertainment sources of choice
3. Perception beats perspective every time. Things can go viral so easily because people read headlines & want to hear what they want to hear.

Guaranteed scary times ahead!!

 

[[DHT]Osiris]

AI is a crock. Here's the scoop:

1. The whole setup is the standard if/then relationship
2. Then it vacuums up all the data in the world
3. Then it cobbles something neat together when prompted!

The neat thing comes in the combinations of:

1. Inputs
2. Processing
3. Outputs

We now have multi-modal inputs & outputs:

1. Text
2. Images
3. Videos

Three things are driving growth now:

1. The convenience & power of the inputs (prompts such as text, images, video, motion capture, as well as visual slider tools, etc.), i.e. making it easier to use. For example, Microsoft added Copilot to Office 365, so now you can "chat" with the AI engine to ask it for what you want & tweak it down to what you need!
2. The types of output (text, photo, video, real-time chat, audio, etc.)
3. The quality of the processing (the various AI engines are literally getting smarter & more detailed every day!)

For example, Adobe just released Firefly 2, which has a couple key features:

1. Very easy to use
2. Includes hyper-realistic output

These are sample generative image outputs from the latest software release (not real photos, all are 100% AI-generated!). But once you "see behind the curtain" (re: Wizard of Oz) & understand how things work, it's all the same formula:

1. Inputs (ask the AI engine to do something, in as much detail as possible)
2. Processing (this is on v2 of the Adobe Firefly generative AI engine)
3. Output (hyper-realistic generated imagery)

Current state of affairs:

1. It can craft any image you want
2. We are on the cusp of overcoming the "uncanny valley" for photos. The images below are just right there on the border of being totally realistic!
3. Raster (Photoshop) & Vector (Illustrator) already have generative AI baked in at this point. Draw anything you want, in any color or style you want!
4. It can clone any voice
5. As as well any song. Music. Instruments. Soundtracks. Vocals. You name it!
6. Video is still in the early stages, but it's RAPIDLY advancing! Motion capture, CGI, video editing, etc.

View attachment 87258

View attachment 87259

View attachment 87260

View attachment 87261

'ai is a crock'
*Proceeds to show dozens of ways AI is awesome and will disrupt industries, while still only touching the tip of the iceberg*

 

[Red Squirrel]

Deep fake stuff is going to get scary. Imagine how useful it will be for extortion. Deep fake vid of you doing or saying something and if you don't send bitcoin they'll post it everywhere. It will also make legal cases interesting because people will be able to refute evidence and say it's deep fake.

 

[Kaido]

'ai is a crock'
*Proceeds to show dozens of ways AI is awesome and will disrupt industries, while still only touching the tip of the iceberg*

To clarify, AI is a "crock" as in Hollywood-style self-aware software. It's just fancy if-then software.

As far as impact goes, it will replace a LOT of jobs in coming years. The first waves of layoffs has already started TODAY due to increased automation & decreased advertising. StackOverflow just announced axing 100 people, unfortunately:

Stack Overflow lays off over 100 people as the AI coding boom continues

Stack Overflow just doubled its staff last year.

www.theverge.com

LinkedIn just announced 660 jobs cut:

https://www.axios.com/2023/10/16/linkedin-layoffs-restructuring-artifical-intelligence

They are focusing on AI recruiting. Every major business I work with is using AI for HR hiring assistance due to the sheer amount of time it saves in the recruitment process:

2024 Recruiter Release | LinkedIn Talent Solutions

Discover the newest capabilities and enhanced features to help you hire the right people, faster.

business.linkedin.com

I use LinkedIn Learning (formerly Lynda) quite a bit. They're adding an AI-powered learning coach, which includes a chatbot that gives real-time advice & tailored content recommendation based on everything in your professional life (your job title, career goals, skills you follow, etc.). One of the reasons I've been able to stay in IT long-term is through ongoing professional development; I'm very curious to see how AI affects personal education in practice:

Reimagining hiring and learning with the power of AI

As the skills required to do our jobs change by a staggering 65% by 2030, the role of HR has never been more important, as our CEO, Ryan Roslansky, shared at our 12th annual Talent Connect event this morning. In fact, 90% of HR professionals say their role has become more strategic in the last year.

www.linkedin.com

Disruptive? Absolutely. The genie is out of the bottle; the cat is out of the bag. Pandora's box is open. It's going to be a thrilling & scary time for the next few years as AI grows up & settles into our culture. Reminds me a bit of Jane from Ender's game, how it's going to be pervasive worldwide. So much potential to fix so many worldwide problems, and yet we're still fighting over strips of land. We're a crazy species lol.