#Tag · bonfire.cafe

🆕 blog! “Agentic AI is brilliant because I loath my family”

At a recent unconference on AI, someone introduced me to the story of a guy who'd tasked an LLM with writing a bedtime story for his daughter. It personalised the tale to include her favourite stuffed toy, whichever cartoon she was obsessing over, and a range of not-too-scary baddies.

And all I could…

Terence Eden

@Edent@mastodon.social · last month

🆕 blog! “Agentic AI is brilliant because I loath my family”

And all I could…

🫧 socialcoding..

@smallcircles@social.coop · last month

🫧 socialcoding..

@smallcircles@social.coop · last month

⚠️ Alert on #Youtube's bombardment by #AI surrogate #influencers. Likely 1,000's of #misinformation channels run by who knows what malign actors.

For instance your popular #geopolitical analist cloned by #LLM #deepfake, with nothing that indicates this, or sometimes with an obfuscated disclaimer ("we're fans of").

Some vids are REALLY hard to discern from the real deal, esp. if you do not know the speaker upfront. Only a slight #UncannyValley tells.

Some channels have millions of subscribers.

Welcome to the #AI #Agentic Web

https://social.coop/@smallcircles/115728232241966073

Meme. Scene from Southpark showing the game troll who slaughtered everyone in Warcraft by sheer time investment. Top text is an AI prompt that reads "Launch troll farm, steal election". The bottom text reads "The Agentic Web" with a big Blobaww emoji preceding it.

Hacker News

@h4ckernews@mastodon.social · last month

Donating the Model Context Protocol and Establishing the Agentic AI Foundation

https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation

#HackerNews #Donating #the #Model #Context #Protocol #Agentic #AIFoundation #AIResearch #TechForGood

Karl Matthias 🇺🇦

@relistan@mstdn.social · last month

My talk on Agentic Coding with Claude from two weeks ago at the Augmented Software Engineering meetup here in Dublin:

https://youtu.be/y8om7rc2kmg?si=eakJ8kzlfepB4ouL

#AiAgent #Agentic #techtalk

Ruth — of systems & em dashes

@RuthMalan@mastodon.social · 3 months ago

“Any AI that is working in an adversarial environment—and by this I mean that it may encounter untrusted training data or input—is vulnerable to prompt injection. It's an existential problem that, near as I can tell, most people developing these technologies are just pretending isn't there.”

— Bruce Schneier

quoted in

https://martinfowler.com/articles/agentic-ai-security.html

Greg Lloyd

@Roundtrip@federate.social replied · 3 months ago

@RuthMalan 🧵 #Agentic #AI #Security

“The fundamental security weakness of LLMs is that there is no rigorous way to separate instructions from data... This leads to the “Lethal Trifecta”: sensitive data, untrusted content, and external communication - the risk that the LLM will read hidden instructions that leak sensitive data to attackers. We need to take explicit steps to mitigate this risk by minimizing access to each of these three elements.” — Martin Fowler