General AI Crap

[HAL 9000]

 

[HAL 9000]

 

[HAL 9000]

Microsoft claims "Community First" AI data center infrastructure — also known as maximum corpo-washing - Windows Central​

Microsoft has begun to realize the backlash to its AI infrastructure push is serious enough to create an entire initiative around "giving back" to communities it impacts with its data center glut.

Microsoft continues to not "get it," but hey, at least it's trying. Sorta.

Microsoft finds itself at another computing crossroads it is largely ill-prepared for, as artificial intelligence threatens to upend everything and anything. Jobs destroyed, tech prices exploding, climatological commitments brushed aside, industrial-scale disinformation, deep fakes, and abuse — artificial intelligence certainly isn't the PR win Microsoft was likely hoping it would be. In fact, the company has earned itself the moniker "Microslop" for its haphazard stumbling through the hype.

Yet, of course, that certainly doesn't mean it's going to slow down.

AI hype is driving a historic rush on the share prices of all associated companies. Microsoft enjoyed the second-highest market capitalization as it rode the early wave hard, but Google's resurgence has seen the Windows firm knocked all the way down to fourth place.

Indeed, Google Gemini and its ownership of the entire stack seem to be the front-runner in AI now, as Microsoft CEO Satya Nadella prioritized short-termism, shareholder fads, and complacency over long-term planning and quality.

Either way, Microsoft will find itself powering the back end for the real innovators leveraging the tech, owing to its industry-leading infrastructure investments in Azure. But there are new tensions rising as it continues down this infrastructural path.

Microsoft, of course, has absolutely no choice but to continue investing in Azure data center infrastructure if it hopes to have any role to play whatsoever in the post-AI world.

With Google now dominating AI on both iOS and Android, ChatGPT declining, and Copilot nowhere to be found ... it seems Microsoft's role is to play data center only.

Increasingly, though, communities are beginning to realize how painful having a Microsoft data center in their backyard can really be.

Microsoft's "Community-First" AI data centers claims mocked

Microsoft has begun to realize the backlash to its AI infrastructure push is serious enough to create an entire "Community First" data center plan.

www.windowscentral.com

 

[OnlyOneKenobi]

I’ve been experimenting with a prompt style that approaches image generation like a fictional Hollywood biopic... casting an actor to portray a real person rather than directly replacing them. I haven’t come across much discussion on this particular framing, so I thought it would be interesting.

On the left - a person from a stock photo. On the right - Brad Pitt, playing the role of that person, on a set that resembles his environment - and talking to the "director" of that film in a BTS photo.

 

[ToxicBunny]

I'm assuming that you didn't define who the "director" is, so I find it interesting that the "director" role seems to tend towards looking like Steven Spielberg.

 

[OnlyOneKenobi]

I'm assuming that you didn't define who the "director" is, so I find it interesting that the "director" role seems to tend towards looking like Steven Spielberg.

I did. It's supposed to be David Fincher, but he does look like Spielberg in that shot.

 

[satanboy]

X says it’s rolling out new restrictions on Grok to block sexually explicit deepfakes, including edits of real people in revealing images. The changes come as U.K. regulator Ofcom continues investigating whether the platform broke the law.

The tool will now refuse illegal image requests, geoblock features where banned, and limit image creation to paid users. Elon Musk says any issues found will be fixed immediately.

 

[OnlyOneKenobi]

Following on from the actor-casting experiment I posted earlier, I’ve been playing with a first-person POV approach where the model is forced to render a scene from inside a specific character’s head rather than observing them. Instead of just changing the camera angle, the prompt anchors the viewer as an actual character in the scene and makes the model describe everything relative to that position. As a stress test I used the “Here’s Johnny” moment from The Shining, with the reference images on the left and the generated result on the right, this time seen entirely through Jack Torrance’s eyes. What surprised me is how differently the model behaves once it’s locked into being someone rather than watching them. It seems to handle spatial relationships, framing, and character positioning much more consistently, almost like it’s directing the scene instead of assembling an image.

 

[Quicks]

AI hype meets reality as majority of CEOs report no financial returns

https://www.techspot.com/news/110983-ai-hype-meets-reality-majority-ceos-report-no.html

 

[OnlyOneKenobi]

View attachment 1879694

Following on from the actor-casting experiment I posted earlier, I’ve been playing with a first-person POV approach where the model is forced to render a scene from inside a specific character’s head rather than observing them. Instead of just changing the camera angle, the prompt anchors the viewer as an actual character in the scene and makes the model describe everything relative to that position. As a stress test I used the “Here’s Johnny” moment from The Shining, with the reference images on the left and the generated result on the right, this time seen entirely through Jack Torrance’s eyes. What surprised me is how differently the model behaves once it’s locked into being someone rather than watching them. It seems to handle spatial relationships, framing, and character positioning much more consistently, almost like it’s directing the scene instead of assembling an image.

This takes some doing for the model to get right, but when it works it's fun. I've even converted a screenshot from MK2 into first person...

 

[Quicks]

Why Silicon Valley is losing faith in the American dream

The same forces driving record valuations are fueling doubts about middle-class survival

https://www.techspot.com/news/110990-why-silicon-valley-losing-faith-american-dream.html

 

[HAL 9000]

 

[satanboy]

OpenAI is facing a huge legal risk that almost no one is talking about

Legal experts say AI chatbots are bringing a new legal risk for tech companies.

www.sfgate.com

The artificial intelligence boom has fueled sky-high valuations for companies like OpenAI and Google, in an investment blitz so large it’s propelling the American economy. But a growing legal threat could cost AI companies billions of dollars in damages — and slow their rapid releases of the nascent tech.

This legal threat came into clearer focus last week, when Google and the chatbot startup Character.AI suddenly reached tentative settlements with five families who said chatbots created by the Google-funded startup had harmed their children. The complaints in those cases described a suicide, attempted suicide and another mental health crisis; citing chat logs, the families alleged that Character.AI’s relationship-based chatbot was dangerously and negligently designed.

In holding the startup to account for these tragedies, the families were testing a legal argument that hasn’t yet won in court. But the mere prospect of the families winning likely pushed Google to settle, according to Vincent Joralemon, the director of the Berkeley Law Life Sciences Law & Policy Center. Joralemon said a single wrongful death lawsuit can result in a $20 million to $100 million judgment because the damages are meant to punish the companies, not just pay the wronged plaintiffs. Joralemon said that level of legal risk extends to the entire AI industry.

 

[satanboy]

OpenAI CEO Sam Altman flags growing dangers as AI

OpenAI Chief Executive Officer Sam Altman has admitted that advanced AI agents are starting to pose serious real-world risks, especially around security.

etedge-insights.com

In a recent post on the social media platform X, owned by Elon Musk, Altman said AI systems are entering a more dangerous phase. While companies have become better at measuring how powerful AI models are, understanding how those powers could be misused is becoming much harder.

Altman explained that the industry is moving into unfamiliar territory. “We have a strong foundation of measuring growing capabilities, but we are entering a world where we need a more nuanced understanding of how those capabilities could be abused,” he wrote. He added that limiting the risks without blocking the benefits is a tough challenge, mainly because there are very few past examples to learn from.

 

[satanboy]

OpenAI has announced at the Davos forum that it plans to launch its first-ever consumer device in the second half of 2026, calling it the highlight of the year for the company.

According to multiple leaks and industry speculation, the upcoming device is expected to be a small, wearable form factor designed to seamlessly integrate AI into everyday life. Unlike traditional gadgets, it may not rely on screens. Instead, the device is rumored to listen, look, and chat, using advanced sensors, microphones, and AI models to understand its surroundings and interact naturally with users.

The vision appears to be an always-available AI companion—one that can observe context, respond in real time, and assist without requiring constant manual input. If these leaks are accurate, the device could represent a major shift away from phones and toward ambient, hands-free AI experiences.

aicouncillor

 

[HAL 9000]

View attachment 1880223
OpenAI has announced at the Davos forum that it plans to launch its first-ever consumer device in the second half of 2026, calling it the highlight of the year for the company.

According to multiple leaks and industry speculation, the upcoming device is expected to be a small, wearable form factor designed to seamlessly integrate AI into everyday life. Unlike traditional gadgets, it may not rely on screens. Instead, the device is rumored to listen, look, and chat, using advanced sensors, microphones, and AI models to understand its surroundings and interact naturally with users.

The vision appears to be an always-available AI companion—one that can observe context, respond in real time, and assist without requiring constant manual input. If these leaks are accurate, the device could represent a major shift away from phones and toward ambient, hands-free AI experiences.

aicouncillor

 

[satanboy]

Sweden's charting body has banned a track from its rankings over its use of AI tools.

Pop-folk single ‘I Know, You’re Not Mine – Jag Vet, Du Är Inte Min’ has already become one of Sweden's biggest tracks of 2026, garnering millions of streams and soaring to the top of the country's Spotify chart.

However, an investigation by Swedish journalist Emanuel Karlsten found that "Jacub" - the artist listed as the single's creator - was actually a team of producers connected to music marketing company Stellar Music, two of whom reportedly worked for the firm's AI department.

Following this revelation, the IFPI, which runs Sweden's official music chart Sverigetopplistan, said the single would not qualify for inclusion in its singles chart, with IFPI head Ludvig Werner telling BBC News: "Our rule is that if it is a song that is mainly AI-generated, it does not have the right to be on the top list."

mixmag.net

 

[Naaimasjien]

View attachment 1879694

Following on from the actor-casting experiment I posted earlier, I’ve been playing with a first-person POV approach where the model is forced to render a scene from inside a specific character’s head rather than observing them. Instead of just changing the camera angle, the prompt anchors the viewer as an actual character in the scene and makes the model describe everything relative to that position. As a stress test I used the “Here’s Johnny” moment from The Shining, with the reference images on the left and the generated result on the right, this time seen entirely through Jack Torrance’s eyes. What surprised me is how differently the model behaves once it’s locked into being someone rather than watching them. It seems to handle spatial relationships, framing, and character positioning much more consistently, almost like it’s directing the scene instead of assembling an image.

The POV is incorrect. She was not in front of the door as hallucinated by the AI image generator.

 

[OnlyOneKenobi]

The POV is incorrect. She was not in front of the door as hallucinated by the AI image generator.

I'm well aware it's not 100% true to where the person was positioned within the scene. It is still a reasonable first person re-creation of what was seen during that sequence. Subsequent generations might place the person in the correct position, but I wasn't concerned with nailing it to that degree.

 

[tetrasect]

View attachment 1880294
Sweden's charting body has banned a track from its rankings over its use of AI tools.

Pop-folk single ‘I Know, You’re Not Mine – Jag Vet, Du Är Inte Min’ has already become one of Sweden's biggest tracks of 2026, garnering millions of streams and soaring to the top of the country's Spotify chart.

However, an investigation by Swedish journalist Emanuel Karlsten found that "Jacub" - the artist listed as the single's creator - was actually a team of producers connected to music marketing company Stellar Music, two of whom reportedly worked for the firm's AI department.

Following this revelation, the IFPI, which runs Sweden's official music chart Sverigetopplistan, said the single would not qualify for inclusion in its singles chart, with IFPI head Ludvig Werner telling BBC News: "Our rule is that if it is a song that is mainly AI-generated, it does not have the right to be on the top list."

mixmag.net