JariwalaH

Wow they didn’t even ask tenants for permission just “assumed” you’d be okay with it…

Agreed, but every lab has almost trained their own flavor of LLMs/agents being good and “generalized” at certain tasks rather than personalized for the user. (Like Anthropic models are great at design, GPT good at infra code, etc)

Also exactly “rules, values, and goals change day-by-day” is why learning needs to happen but across the apps/agents you’re using to personalize to you… not having to reconfigure every time for new agent or app that’s good at a certain task..

Yeah the idea is to have the human-AI Agent User interaction Experience designed that way really to allow you to manage multiple personal agent apps, AI apps, etc.

Gotcha, yeah data privacy is a big concern for some folks and since we’re local you can set rules to keep data local and not give big tech AI access to it without your permissions. (Zero Trust for AI)

Even if they’re pinging the connectors behind the scene in a cloud agent setup you could possibly get notified or alerted and maybe stop them with Emu(like stop pinging that 24/7 hrs PAT that I didn’t want you too..)

I bet you Muse is constantly data exfiltrating Google Calendar APIs in their cloud agent to be “proactive” this way… not true alignment

Well what if we were to put an on-device SLM with our custom RL algorithms, Jev-like capabilities for certain risk thresholds of user’s individual preferences of allowing, denying, and asking user when an agent should and shouldn’t act…(some deterministic enforcement where needed)

Like Muse, Grok bot’s own agents can govern themselves it’ll be kind of bias if the agents knows its own rules and can try to circumvent it or rule hack around if its being trained on RLHF or even if the RLVR optimize for calling tools/data exfiltrating constantly