Futurology

3393 readers

5 users here now

founded 2 years ago

MODERATORS

While many are hyping 2025 as the year of AI Agents, some think they have fatal vulnerabilities to malicious attacks. (futurism.com)

submitted 10 months ago by Lugh to c/futurology

4 comments fedilink hide all child comments

all 5 comments

sorted by: hot top controversial new old

[–] Lugh 5 points 10 months ago* (last edited 10 months ago)

OpenAI is on a treadmill. It has vast amounts of investor billions pouring into it and needs to show results. Meanwhile, open source AI is snapping at its heels in every direction. If it is true that it is holding back on AI agents out of caution, I'm pretty sure that won't last long.

[–] CarbonIceDragon@pawb.social 3 points 10 months ago (1 children)

Beyond obviously malicious stuff like this, if people are going to have AI autonomously purchase things, I wonder if we might see stuff like ads designed not to influence human purchasing decisions, but to make someone's AI assistant think that a given product is a better match for what it was asked to buy than it really is

[–] Philosofuel 3 points 10 months ago

Great point, people can think of so many ways to manipulate these agents.

[–] just_another_person@lemmy.world 0 points 10 months ago

They're just as prone as anything else software-based to attacks. They will absolutely be compromised.

The sooner people wake up to hype words about this, the sooner we can all move on from another bubble.

FYI: Siri isn't giving me robot handjobs right now, and that was supposed to be possible 10y ago. Wake the fuck up.