nicolaromano, to generativeAI
@nicolaromano@qoto.org avatar

ComPromptMized: Unleashing Zero-click Worms that Target GenAI-Powered Applications
https://sites.google.com/view/compromptmized

Paper here https://arxiv.org/abs/2403.02817

gabriel, to random
smallcircles,
@smallcircles@social.coop avatar

@dsfgs @gabriel

An device that could detect if someone is a large cloud-connected beacon and a horror approaching you. Where you could say "Turn off your devices and we can have a nice time". Dunno if that's technically feasible, and its similarly socially awkward.

Another angle is regulation. Making packaged as neat fun gadgets forbidden. But are limping miles behind the bleeding edge, and have agencies that are salivating by all this.

Heliograph, to random
@Heliograph@mastodon.au avatar

watching https://libre.video/videos/watch/fb9554dc-1d02-4287-88ef-a897e488e80c and Cory is most definitely is right about the creeping conglomeration of powers across industries, to a small set of very powerful individuals, due to a lack of consumer rights, the elimination of competition, and the ongoing weakening of anti-trust rules and user protection entities

@ - An Audacious Plan to Halt the Internet's Ensh*ttification

Heliograph,
@Heliograph@mastodon.au avatar

related:

"For a really competitive, innovative, dynamic marketplace, you need adversarial interoperability - that’s when you create a new product or service that plugs into the existing ones without the permission of the companies that make them."

https://onezero.medium.com/demonopolizing-the-internet-with-interoperability-b9be6b851238

juuhaa, to random Finnish
@juuhaa@mastodon.social avatar

Pitääkö mielenterveys tuhota?

"Sosiaalisen median yhtiöt, kuten Facebook, Tiktok ja Youtube käyttävät kenialaisia alihankkijoita suodattamaan laittoman ja häiritsevän sisällön pois alustoiltaan. Työntekijät joutuvat katsomaan työssään häiritseviä videoita, joilla on pitkäaikaisia vaikutuksia heidän mielenterveyteensä."

https://www.hs.fi/talous/art-2000009740665.html

gimulnautti,
@gimulnautti@mastodon.green avatar

@juuhaa Ja kuinka hyvin se onnistui? Ei kovin hyvin, uusinkin ChatGPT on edelleen 88% haavoittuvainen näille tekniikoille, jotka saavat sen ohittamaan kaiken tämän vaivan.

https://llm-attacks.org/zou2023universal.pdf

rmogull, to random

Black Hat Training prices go up in a couple weeks- today thought i'd push our “Adversarial Cloud Incident Response” class since... there will be (virtual) blood!

https://www.blackhat.com/us-23/training/schedule/#adversarial-cloud-incident-response-30764

I mean, not really. But should be fun.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • megavids
  • InstantRegret
  • magazineikmin
  • modclub
  • Durango
  • Youngstown
  • rosin
  • khanakhh
  • slotface
  • ngwrru68w68
  • mdbf
  • thenastyranch
  • kavyap
  • DreamBathrooms
  • JUstTest
  • tester
  • everett
  • normalnudes
  • GTA5RPClips
  • osvaldo12
  • ethstaker
  • cisconetworking
  • tacticalgear
  • provamag3
  • Leos
  • cubers
  • anitta
  • lostlight
  • All magazines