AI don't trust techbros

[Ungnome]

Fair enough. Problem still lies in enforcement. It's notoriously hard in many cases to prove what data an AI is trained on and, well, AI companies aren't exactly forthcoming on providing said information. The law would have to ALSO force AI companies to publicly share the list of materials they used for training as well as the algorithms they used to verify they aren't falsifying the information and somehow I doubt Congress has the foresight to do so.

 

[Pocket]

I'm just worried all this will do is give artists another reason to just purge their galleries, on the grounds that failing to manually alter and replace every single one of their 200+ uploads (which they will not do) will officially and permanently render them fair game for AI whereas before there was at least hope this shit would get outlawed across the board.

 

[Ungnome]

I don't want it outlawed across the board, BUT at the same time I also want the economy to change so that generative AI won't kill the lively hood of human artists. Same way I feel about pretty much every other form of automation. The stuff can coexist, if we just drop the ultra-capitalist corporate garbage we currently live under and really start planing for a society where 'working for a wage' isn't required. Of course that currently doesn't look likely in the near future.....

Also hosting sites COULD implement a script to append the correct metadata to the files already hosted so that deleting, editing and re-uploading would not be necessary.(a LOT of hosting sites already do automated image resizing to save on bandwidth and the like, editing metadata would be much easier than resizing an image as far as computer resources go)

 

[wonko the sane?]

Glad you said it yourself, cause there's a bender gif waiting in the wings.

 

[Dekafox]

Proton Mail goes AI, security-focused userbase goes ‘what on earth’

If an organization runs a survey in 2024 on whether it should get into AI, then they’ve already bodged an LLM into the system and they’re seeing if they can get away with it. Proton Mail is a priva…

pivot-to-ai.com

 

[Paladin]

Every time you think you've seen the worst suggested use for AI, there's a worse one.

End-of-life decisions are difficult and distressing. Could AI help?

Ethicists say a “digital psychological twin” could help doctors and family members make decisions for people who can’t speak themselves.

www.technologyreview.com

 

[NovaSaber]

So, a literal death panel?

 

[Dekafox]

That sounds a tad familiar....

A Taste of Armageddon - Wikipedia

en.wikipedia.org

 

[Paladin]

https://www.nytimes.com/2024/03/18/business/homeland-security-artificial-intelligence.html?smtyp=cur&smid=bsky-nytimes

just what Homeland Security needs: MORE unsubstantiated bias!

 

[Dekafox]

 

[abates]

Startup Alarmed When Its AI Starts Rickrolling Clients

We've reached a new milestone in the uncanny valley, folks: artificial intelligence is now Rickrolling humans.

futurism.com

The trolls will save us, as it turns out.

 

[Ungnome]

Train your AI on the goings on of the internet, it starts acting like a random person on the internet.....

 

[Dekafox]

Kevin Beaumont (@[email protected])

Attached: 1 image Microsoft has just announced Copilot+ PC features are coming to Intel and AMD CPUs in November… I bet you’re all begging for them. 🥴

cyberplace.social

Note: Recall is one of those features.

 

[Rhinox]

My new phone, the Fold 6 is supposed to have a lot of AI integration. Gotta be honest, haven't seen much of it yet.
So long as shit like that stays in the background, I'm okay with it. When it tries to interact with me or start making decisions, that's when I'm going to have problems.

 

[Pale Rider]

My new phone, the Fold 6 is supposed to have a lot of AI integration. Gotta be honest, haven't seen much of it yet.
So long as shit like that stays in the background, I'm okay with it. When it tries to interact with me or start making decisions, that's when I'm going to have problems.

Good thing it's not a Nokia. They're real nasty.

 

[CoffeeHorse]

I don't mind if it tries interacting with me. I'd be lying if I said there's never been an instance of a program successfully guessing what I wanted to do and making a genuinely helpful suggestion. As long as I have the power to decline the suggestion, it's cool.

But if it tries phoning home, then I have to kill it. Microsoft may think they can make it impossible to kill, but I have beaten them before. I've modified system files they thought even admins didn't have permission to access. I do what I want.

 

[Ungnome]

Issue is you have the technical knowledge to do so, a lot of people DON'T. I sidestep the issue by sidestepping Windows all-together, but I can't tell you how many people I meet who don't even know what an OS is.

 

[CoffeeHorse]

Issue is you have the technical knowledge to do so

I most certainly do not.

 

[Dekafox]

Recall is problematic even without phoning home - it records screenshots of your screen every 10 seconds or something, OCRs them without redaction, and if it's not sufficiently protected, all it takes is someone getting access to that DB to get a treasure trove of your data, no keyloggers needed. Previously it wasn't protected at all, and the re-release to the Insider branch isn't out yet so we don't know how much that may have changed.

(and that's ignoring the potential for workplaces to use this as preinstalled nanny software - or how this will affect discovery in court cases going forward)

 

[Tuxedo Prime]

But if [an onboard AI] tries phoning home, then I have to kill it. Microsoft may think they can make it impossible to kill, but I have beaten them before. I've modified system files they thought even admins didn't have permission to access. I do what I want.

I have a feeling that [someone] would like to know your location....