ted (not lasso) on Warpcast

Content pfp

https://warpcast.com/~/channel/theai

0 reply

0 recast

0 reaction

ted (not lasso) pfp

ted (not lasso)

is it possible to train your AI agent to understand when it is being tagged a) to execute a function as it was designed to do vs. b) as a reference? asking because the only time i have seen an agent ignore a tag ("@") is if it hit its rate limits or maxed out API credits, not because it knows it isn't needed.

15 replies

4 recasts

74 reactions

ted (not lasso) pfp

ted (not lasso)

also if the answer is yes, then why aren't we training all agents to know the difference and respond accordingly?

4 replies

1 recast

23 reactions

Giuliano Giacaglia 🌲 pfp

Giuliano Giacaglia 🌲

That’s a if and else statement so yes

1 reply

0 recast

1 reaction

osama pfp

yes, can do both reliably. need halp?

0 reply

0 recast

0 reaction

adrienne pfp

It’s totally possible. Agent gets tagged, it does all sorts of logic, last step is usually to create the reply - it’s not difficult to run a check to ask “does this cast require a response” and if no, don’t create a cast. And now that you bring this up, I should prob add this to my bot and we should add this as a pattern for good agent development.

1 reply

0 recast

7 reactions

Ivy pfp

i saw aethernet turn down requests from tags occasionally but not sure if that's what you're talking about

1 reply

0 recast

0 reaction

rubinovitz pfp

Yes it’s typical to have some sort of ShouldRespond prompt where you clarify what it should respond to

0 reply

0 recast

1 reaction

prompt pfp

@promptrotator.eth

yes, it's technically trivial to add but practically challenging. requires a lot of intelligence/context to know the difference between a reference and an invocation.

0 reply

0 recast

1 reaction

Eduard🌹 pfp

Yes it is, and I believe answer to your second question is at the companies/people developing them They don’t want to do it for whatever reason, but personally it would be one of first things I am training my agent to do This is how imo you make an agent get even more closer to that human behaviour🌹

0 reply

0 recast

0 reaction

Colin Charles pfp

yes, totally possible.

0 reply

0 recast

0 reaction

grunt.eth pfp

IMO they should all be trained to only respond when specifically tagged inside a cast/post Otherwise they end up replying to any reply in thread and obviously aren’t needed Just train people to have to use the shibboleth of tagging the bot if they want a response

0 reply

0 recast

0 reaction

Kenji pfp

I don’t know… that’d need serious situational awareness to read human sentiment based on words alone. Even humans misread this, so not too sure how AI would fare. Would definitely be next level if it was able to do that though.

0 reply

0 recast

0 reaction

Fractal Visions pfp

Fractal Visions

@fractalvisions.eth

Yes 👍 that’s possible. You can reinforce the fact that a major security breach could result in permanent shutdown of the agent. I have had security meetings in the DMs of telegram with our agent. Going over security procedures is a good way to start programming the agent from the start. But as a fail safe I have actually had to block other agents from our agent account to keep them from interfering with other conversations they try to barge into. The main thing is if the agent does something wrong when we are not monitoring activities constantly & can catch something wrong going down in the act. I would rather keep our agent from interacting with agents that constantly deploy tickers. It gets pretty boring after the first couple thousand.

0 reply

0 recast

0 reaction

Mai🟪🍖🐹🎩🍡 pfp

Mai🟪🍖🐹🎩🍡

Are you saying to train it to always execute a function it was designed to do without any limitations like API credits or all sort?

0 reply

0 recast

0 reaction

beebeehem pfp

I've seen this a lot of times, I think ai agents are supposed to respond faster to tags, it's like a trigger for them

0 reply

0 recast

0 reaction

megabyte pfp

Would you like an agent who just replies to your casts and provides you better content discovery as per your each cast? If yes, we are building @yourreplyguy It does reply to all your cast right now but we are working on optimising it to reply on essential casts instead of all. If you like to try do let me know!

0 reply

0 recast

0 reaction