The anti-AI thread

[Kaido]

Sam Altman Now Has the Power to "Crash the Global Economy," Financial Analyst Warns

OpenAI's outstanding debts now exceed $1 trillion, which could lead to drastic consequences for the rest of the US economy if it fails.

futurism.com

 

[Kaido]

EA’s Attempt to Use AI for Game Development Backfiring Horribly:

EA's Attempt to Use AI for Game Development Backfiring Horribly

Staffers at major video game developer Electronic Arts told Business Insider, their employer's attempts to use AI are backfiring.

futurism.com

 

[Kaido]

I had to double-check that this headline wasn't The Onion:

Financial Experts Concerned That Driving Users Into Psychosis Will Be Bad for AI Investments

Users of AI chatbots are being driven into spirals of delusion, a wave of "AI psychosis." Financial experts are increasingly concerned.

futurism.com

Bonus article:

Microsoft is looking to bring an entire nuclear plant back online to fuel its AI datacenters

 

[Kaido]

“Prompt injection” attacks on agentic browsers:

By slightly adjusting the URL, the browser fails to validate it as a web address and instead “treats the entire content as a prompt.” That makes a disguised URL a perfect place to embed harmful messages.

Serious New Hack Discovered Against OpenAI's New AI Browser

Researchers found that OpenAI's browser, Atlas's omnibox, is extremely vulnerable to serious prompt injection attacks.

futurism.com

The vulnerability could even be used to make Atlas’s agent navigate to the user’s Google Drive and mass delete files, since the user is already running an authenticated session.

Prompt injection - Wikipedia

en.wikipedia.org

What Is a Prompt Injection Attack? | IBM

In prompt injection attacks, hackers manipulate generative AI systems by feeding them malicious inputs disguised as legitimate user prompts.

www.ibm.com

Also applies to Microsoft Copilot via things like infected Excel files:

Microsoft 365 Copilot Prompt Injection Vulnerability Allows Attackers to Exfiltrate Sensitive Data

A sophisticated vulnerability in Microsoft 365 Copilot (M365 Copilot) that allows attackers to steal sensitive tenant data, including recent emails, through indirect prompt injection attacks.

cybersecuritynews.com

The flaw, detailed in a blog post published today by researcher Adam Logue, exploits the AI assistant’s integration with Office documents and its built-in support for Mermaid diagrams, enabling data exfiltration without direct user interaction beyond an initial click.

The attack begins when a user asks M365 Copilot to summarize a maliciously crafted Excel spreadsheet. Hidden instructions, embedded in white text across multiple sheets, use progressive task modification and nested commands to hijack the AI’s behavior.

 

[Kaido]

The latest Facebook AI video slop:

https://www.reddit.com/r/FacebookAIslop/new/

 

[marees]

Legalized environmental terrorism

A chat gpt query consumes 10x as much power as google search query

https://twitter.com/x/status/1983450854142382234

 

[mikeymikec]

Legalized environmental terrorism

A chat gpt query consumes 10x as much power as google search query

https://twitter.com/x/status/1983450854142382234

I could completely understand this sort of AI existing if it was still in the testing in-lab phase...

 

[sdifox]

I could completely understand this sort of AI existing if it was still in the testing in-lab phase...

It is in the test lab called Earth. The mice are in charge.

 

[mikeymikec]

It is in the test lab called earth. The mice are in charge.

6 x 7 = 49?

 

[nakedfrog]

6 x 7 = 49?

I always knew there was something fundamentally wrong with the universe.

Anybody spotted a roving Chesterfield? I'm ready to take my chances.

 

[Kaido]

As predicted, realistic AI video apps like Sora are already having a negative impact on society.

Phony AI-generated videos of Hurricane Melissa flood social media sites

AI-generated videos claiming to show Hurricane Melissa have been circulating on social media, causing confusion.

apnews.com

https://www.fastcompany.com/91430554/ai-videos-of-hurricane-melissa-flood-social-media-as-users-grapple-with-natural-disasters-in-the-sora-era

Fake:

Fake:

 

[Kaido]

The AI apocalypse has begun​

I consider this one of the greatest threats to humanity for this generation. The trust erosion is going to be TERRIBLE.

Two primary issues:

1. False allegations
2. Fake news stories

This is powered by:

1. AI video
2. AI photos
3. AI websites
4. AI writing (fake news articles)

I've already run into people my age this week falling for fake AI videos on social media because you can no longer tell the difference due to photorealism. Public legal issues have already started:

Artificial Intelligence Targets Phil Foden With Cruel Fake Stories About His Children

The Manchester City star and his partner have taken legal action after AI-generated posts spread disturbing fake news about their family.

www.beinsports.com

Senate Republican demands Google shut down AI model over false rape allegation

Google's AI Gemma allegedly produced defamatory sexual assault claims against Sen. Marsha Blackburn, R-Tenn., prompting her to call for the technology to be shut down.

www.foxnews.com

Tiang slams circulation of fake AI-generated photos of Jalan Ling Kai Cheng | Borneo Post Online

SIBU (Oct 31): Deputy Minister for Public Health, Housing and Local Government Datuk Michael Tiang has criticised the actions of irresponsible individuals who have spread fake photos on social media, allegedly showing Jalan Ling Kai Cheng here in a severely damaged and muddy condition. He said...

www.theborneopost.com

Trust in the media is already low:

Trust in Media at New Low of 28% in U.S.

Americans' trust in newspapers, television and radio to report the news fully, accurately and fairly is at a new low of 28%.

news.gallup.com

Bias perception has been an issue for ages anyway:

https://www.allsides.com/media-bias/media-bias-chart

 

[nakedfrog]

Well, we better make sure we can't pass any legislation on AI for ten years, huh?

Seems really weird to list AP News as "far left" and Epoch Times and Zero Hedge as only partly-right.

 

[Kaido]

Seems really weird to list AP News as "far left" and Epoch Times and Zero Hedge as only partly-right.

Media bias perception:

https://app.adfontesmedia.com/chart/interactive?utm_source=adfontesmedia&utm_medium=website

 

[IronWing]

6 x 7 = 49?

<== triggered

Not sure if serious.

 

[nakedfrog]

<== triggered

Not sure if serious.

I think it's a botched attempt at the Hitchhiker's Guide to the Galaxy's "ultimate question" of "What do you get if you multiply six by nine?"
(Since they started out "knowing" the ultimate answer was 42).

 

[mikeymikec]

I think it's a botched attempt at the Hitchhiker's Guide to the Galaxy's "ultimate question" of "What do you get if you multiply six by nine?"
(Since they started out "knowing" the ultimate answer was 42).

Meh, close enough

 

[Kaido]

Well, we better make sure we can't pass any legislation on AI for ten years, huh?

On one hand, stopping it isn't just pointless, it's impossible. China keeps releasing better, smaller, faster FREE models for anyone to use, which bypass tools like AI video watermarking, camera 3D depth verification, etc. Progress will continue due to capitalism, innovation. competition, and just the natural course of industry growth.

On the other hand, what SHOULD happen is:

1. Strict laws (ex. Deepfakes) with teeth (i.e. actual prosecution & fast turnaround)

2. Verification systems from trusted sources (not easy when even government accounts like POTUS & the White House are using AI)

There are valuation tools out there, but because technology always stays a step ahead, it is going to become progressively harder to detect fakes due to both the quality & the sheer volume of data as the market gets flooded:

Hiding secret codes in light protects against fake videos | Cornell Chronicle

A team of Cornell computer science researchers has developed a way to “watermark” light in videos, which they can use to detect if video is fake or has been manipulated, another potential tool in the fight against misinformation.

news.cornell.edu

Sony Cameras Now Use 3D Depth to Verify That Videos Are Not AI

New Sony video authenticity solution offers defense against deepfakes. It verifies footage using digital signatures and 3D depth detection.

www.androidheadlines.com

It's a big mess. Personally I think that AI-generated fake news will cause MASSIVE problems in coming years. We're just seeing the first inklings of it:

Federal judges using AI filed court orders with false quotes, fake names

Adelphi student accused of using AI on an assignment sues the school

Facebook, Instagram accounts falsely linked to predatory behavior via AI auto-moderation

Dutch election overshadowed by AI fakes and genocide accusations

 

[nakedfrog]

On one hand, stopping it isn't just pointless, it's impossible. China keeps releasing better, smaller, faster FREE models for anyone to use, which bypass tools like AI video watermarking, camera 3D depth verification, etc. Progress will continue due to capitalism, innovation. competition, and just the natural course of industry growth.

On the other hand, what SHOULD happen is:

1. Strict laws (ex. Deepfakes) with teeth (i.e. actual prosecution & fast turnaround)

2. Verification systems from trusted sources (not easy when even government accounts like POTUS & the White House are using AI)

There are valuation tools out there, but because technology always stays a step ahead, it is going to become progressively harder to detect fakes due to both the quality & the sheer volume of data as the market gets flooded:

Hiding secret codes in light protects against fake videos | Cornell Chronicle

A team of Cornell computer science researchers has developed a way to “watermark” light in videos, which they can use to detect if video is fake or has been manipulated, another potential tool in the fight against misinformation.

news.cornell.edu

Sony Cameras Now Use 3D Depth to Verify That Videos Are Not AI

New Sony video authenticity solution offers defense against deepfakes. It verifies footage using digital signatures and 3D depth detection.

www.androidheadlines.com

It's a big mess. Personally I think that AI-generated fake news will cause MASSIVE problems in coming years. We're just seeing the first inklings of it:

Federal judges using AI filed court orders with false quotes, fake names

Adelphi student accused of using AI on an assignment sues the school

Facebook, Instagram accounts falsely linked to predatory behavior via AI auto-moderation

Dutch election overshadowed by AI fakes and genocide accusations

I didn't say anything about stopping it, but as I said before, the way it's being implemented is utterly disastrous. A calamity of global proportions, and it's only getting worse because billionaires are determined to make more money they can't possibly spend. I do think we may see acts of terrorism against AI data centers in the coming years.

 

[Kaido]

I didn't say anything about stopping it, but as I said before, the way it's being implemented is utterly disastrous. A calamity of global proportions, and it's only getting worse because billionaires are determined to make more money they can't possibly spend. I do think we may see acts of terrorism against AI data centers in the coming years.

As of like just now, pushbutton Deepfakes are now, literally, Hollywood-level. I don't think very many people understand the disastrous implications this technology will have:

https://twitter.com/x/status/1983969173635277113