The AI agent spectrum Separating different classes of AI agents from a long history of reinforcement learning. Why we can be optimistic for AI agents but also extremely critical of the terrible communications around them to date. Plus, some policy guidance. - ThreadSky

About ThreadSky

natolambert.bsky.social • 167 days ago

The AI agent spectrum
Separating different classes of AI agents from a long history of reinforcement learning.
Why we can be optimistic for AI agents but also extremely critical of the terrible communications around them to date.
Plus, some policy guidance.

Comments

neurostats.org•167 days ago

📌

phillipcarter.dev•167 days ago

Love the bit about how you predict agents will look in 2025. Lots of hype in the latter half of this year, but few have actually been building them. It takes a long time for the industry to actually make use of this stuff.

timkellogg.me•167 days ago

i played around with them and came away with the impression that it’s a lot like riding a unicycle. it can work…

juwariah.bsky.social•166 days ago

Depending on the workflow, agents can work decently. But some agents need more reinforcement learning, some might need more niche data sets, some might need more eval & human in the loop feedback.

Also huge difference on how agents will be approached across enterprise use cases vs solo creators.

markriedl.bsky.social•166 days ago

Small quibble: open-ended agents doesn’t mean always-awake agents.

natolambert.bsky.social•166 days ago

I like this quibble.

bayinfosys.bsky.social•161 days ago

What do you think about multi-modal agents? I'm not sure the distinction carries from models to agents, and if it does/does not, maybe it helps the discussion?

Posting Rules

Be respectful to others
No spam or self-promotion
Stay on topic
Follow Bluesky's terms of service

Comments

Posting Rules

Reply