AI don't trust techbros

[wonko the sane?]

ChatGPT finally understands the wiki article on greek philosophy.

 

[abates]

Nobody can explain why? It's simply a case of gigo. Garbage in, garbage out. The thing learns from the internet, it probably just discovered 4chan or the discussion sections in Steam.

Yes, pretty much this. Current AI is so susceptible to sabotage just by feeding it off-the-wall bollocks that I'm surprised it doesn't happen more often.

 

[abates]

AI making it dead simple to scam the public:

https://twitter.com/i/web/status/1762235022851948668

Actual promotional image:

Love the enigemic sounds and ukxepcted twits.

 

[Pocket]

Dead simple to scam people who lack the basic capacity to read, maybe. Though maybe that's the idea—like the gibberish in a typical spam email, the idea is to filter out anyone with an ounce of thinking capacity so you can turn around and shame them for being dumb enough to fall for it in the first place.

 

[NovaSaber]

China court says AI broke copyright law in apparent world first | Semafor

A Chinese court ruled that an AI bot improperly used protected material in generating images in what appears to be the first ruling of its kind.

www.semafor.com

And ChatGPT is now almost capable of "proving" it's not a robot...

 

[Tuxedo Prime]

I sort of figured that was coming....

Spoiler: How did I know....?

 

[Ungnome]

China court says AI broke copyright law in apparent world first | Semafor

A Chinese court ruled that an AI bot improperly used protected material in generating images in what appears to be the first ruling of its kind.

www.semafor.com

And ChatGPT is now almost capable of "proving" it's not a robot...

Well, considering the secondary(or possibly PRIMARY GOAL, depending on your point of view) of captcha was to train AI image recognition algorithms, can you be surprised?

 

[Dekafox]

Greg Wilson (@[email protected])

Docusign just admitted that they use customer data (i.e., all those contracts, affidavits, and other confidential documents we send them) to train AI: https://support.docusign.com/s/document-item?language=en_US&bundleId=fzd1707173174972&topicId=uss1707173279973.html They state that customers...

mastodon.social

Docusign just admitted that they use customer data (i.e., all those contracts, affidavits, and other confidential documents we send them) to train AI:
https://support.docusign.com/s/docu...zd1707173174972&topicId=uss1707173279973.html
They state that customers "contractually consent" to such use, but good luck finding it in their Terms of Service. There also doesn't appear to be a way to withdraw consent, but I may have missed that.

 

[Pocket]

I was reminded again today of a comment I saw on Reddit last year about how AI might be on track to turn the internet into the digital equivalent of a grey goo scenario. The OP was a link to a news article about how a bunch of websites are getting effectively DDOS'd by AI scraper tools, leading to this comment thread:

AI is gonna destroy the open internet. This is just one more reason why.

The other main one being what the output of those glorified chatbots is doing to things like search results.

That's the big one I'm worried about -- and incidentally, it's also the reason why I'm not worried about AI taking anyone's job. AI training on its own output is gonna turn it into an increasingly incestuous, self-referential morass of bad data, falsehoods, spam, scams, and other assorted bullshit that's of little use to users. Any attempt by companies to replace human workers with AI is gonna end with either those human workers quickly being rehired to constantly babysit the AI and stop it from constantly making amateurish mistakes, or those companies going out of business, just like the media companies in the mid-2010s who got seduced by the overhyped "pivot to video" and lost their shirts because Facebook lied to make its business model look good. "Garbage in, garbage out" is one of the most fundamental maxims of computer science, a field that you'd think Silicon Valley would understand -- but then again, most of these big companies aren't really tech companies anymore, but advertising companies and vehicles for raising VC cash.

Too bad it's also gonna turn the whole internet into a digital Chernobyl in the process. Toxic and worthless for human users who can no longer find anything of value buried beneath the bullshit that AI can churn out at industrial scale. Once advertisers start asking just how much engagement they're getting is from people with any money to spend, or even if any actual human eyes are seeing their ads, they're gonna abandon online marketing altogether, and once that happens, most of the "Big Data" online platforms and services that get their revenue from ad money are gonna fall into a death spiral of enshittification just trying to keep their servers online. It'll probably take down AI in the process as well, bereft of the data it needs to produce its seemingly miraculous functions. Without data to train AI systems on, we can probably expect another AI winter.

And let's not even get into the viruses and malware that hackers, criminals, intelligence agencies, and other shady actors will use AI to create.
The actual jobs that AI is gonna destroy -- not replace, destroy -- are any that depend on the existence of the internet. Doctors, lawyers, scientists, artists, journalists? They're probably safe. Their jobs have been around for centuries, and if they lose the internet, they'll have to readjust to a 1993 internet but will still come out in one piece. (If anything, killing social media will probably lead to a comeback for the local newspapers and magazines that social media wiped out over the last twenty years.) But influencers, bloggers, streamers, online marketers, coders, most of the gig economy, anything with the word "digital" or "mobile" in the job description? They're the ones who are gonna take a big hit. Even a lot of computer hardware and phone manufacturers would probably be in for tough times. We'll still have the same advanced computers, solid state drives, GPUs, and software we have now, even if it's installed via DVD or flash drive instead of downloaded, but without the internet, what's the point of a smartphone or a 5G network when a feature phone and a 3G network can do all the same stuff? How do video game manufacturers react to the death of online gaming, killed by AI assists that can outplay any human while skirting detection? What's the point of a Chromebook at all?

All the "learn to code" memes that went around in the last decade are starting to look really foolish right now from my perspective.

 

[wonko the sane?]

The internet as a delivery vector (think a tap and water.) will always remain, pushing code without physical media is just too useful. The internet as a SOCIAL medium? Yeah, I agree that AI will have a heavy hand in ultimately killing it (either by burying it in reams of garbage Wall*E style, or simply pushing out humanities ability to make money on that aspect of the internet.) absolutely.

And I don't hate it. AI will derp its way into liberating us from the already toxic and lethal morass that is social media.

 

[CoffeeHorse]

I'm telling you, forums are going to get the last laugh.

 

[Tuxedo Prime]

I'm telling you, forums are going to get the last laugh.

It certainly puts Google's recent decision to close posting access to Usenet into perspective.... arguably where all of this all began back in the fateful aforementioned year.

 

[Pocket]

I'm telling you, forums are going to get the last laugh.

Will we? The forums that there are maybe ten of left on the entire internet, with a combined userbase of maybe a couple hundred people, and which keep getting hacked and wiped with no backups because even the few people who still have the time, patience and money to keep one running don't actually have enough of any of those things to prevent that from happening?

That's who's going to be the last ones standing at the end of it all?

 

[CoffeeHorse]

Yeah. I didn't say it will be pretty.

 

[Ungnome]

Eh, with any luck NEW forums will come into existence as the social media behemoths implode. Frankly a lot was lost when we went from smaller community and personal web sites to the over-concentrated social media mega-sites we have now. Granted, for every friendly little fandom forum you're gonna get something hideous like 'Stormfront' but that's the way it goes. It's not like the big sites have done much to keep said groups from from spreading hate on their platforms anyway....

 

[Dekafox]

The Fediverse is promising too, in that it doesn't push interaction via algorithm and the Internet Hate Machine, and because it's broken up among a bunch of small sites instead of one massive behemoth it's manageable by individual people rather than AI "moderation" *stares at tumblr and Bluesky*

it's also customizable in that you can interact with it via a multitude of different ways - for example Lemmy if you want a Reddit-like format, or Mastodon if you want a scrolling timeline like The Site Formerly Known As Twitter and Bluesky. There's also a art-oriented Fediverse interface similar to tumblr et al. But since it's the fediverse, they all can talk to each other - for instance you can "subscribe" to a Lemmy forum/subforum on Mastodon, and it'll put their posts right into your Mastodon timeline.

Of course BECAUSE it doesn't have the addiciting hate engagement algorithm, and isn't a easy-to-choose single site, the masses tend to bounce off it a lot from what I've heard, or sign up and drift off. Mastodon at least has a few other problems it's still working out how to handle as it grows and develops as well. But I could see that surviving the AI invasion too.

Based on the way people ARE bouncing off something that isn't any more complicated than choosing a websiteserver you like and making an account there though(sound familiar?) I fully expect the masses in general will continue to flock to megasites for their dopamine fix, whether it's human or AI generated. Eternal September may have finally ended, just not in the way one would hope, as we nerds return to the old ways and those would would have been AOLers back then lemmming their way off the cliff of human/AI interactions.

 

[wonko the sane?]

The old ways still excluded the bulk of the human race; because they couldn't grasp the technical aspect of use, or simply didn't want to learn. Ease of use is... sadly... critical to push use to the masses, especially in north america and western europe.

I don't honestly blame the megalith social media sites for operating the way they do: they functionally STARTED the concept, and why change what works?

Oh... because it coopts the dopamine process and literally addicts people to hate mongering and information overload. Oh, and it's driving bullying, mental health collapse, suicides... Yeah, sounds like it's time for change.

Frankly: the death of social media (As it currently stands.) is not a bad thing. But... maybe we work on advanced captchas?

 

[Dekafox]

VessOnSecurity (@[email protected])

Attached: 1 image Somebody managed to coax the Gab AI chatbot to reveal its prompt:

infosec.exchange

And yeah, others in that same thread confirmed it:

Thibault D. (@[email protected])

Attached: 1 image @[email protected] LOL this is real

mastodon.social

Anthony, of course (@[email protected])

Attached: 2 images @[email protected] @[email protected] do it yourself at gab.ai. I just tested it myself and it's very real

bitbang.social

Wonder how many other bots this works on too...

 

[wonko the sane?]

We made a chat bot!

...And then you made it racist?

Oh great white christian god, we made it SO racist.

 

[Pocket]

The really wild thing is that AI has reached the point where the way to encode its prime directive is to give it instructions in plain English just like any other prompt.