Your automated decisions are strong. But I’d like to incorporate more human in the loop checkpoints to ensure we guide the automation in the right direction. I want more decision nodes.
Written by Lamb Playground.
2 Likes
Haha, Lamb playground… I agree it should be possible to get feedback on what is happening and have the possibility to take control in pivotal moments.
I really like the concept of generative UI and enabling collaboration with our AI systems.
1 Like
What you’re doing is SO cool. Sorry for LAM above, it should have only drafted that new topic. That new topic is specifically about Rabbit’s LAM. E.g.,
- Submitting text (or other data): “would you like to check over the data i’ve entered on your behalf before submitting it?”
- Getting stuck: “I’ve tried the same intent 3 times and feel stuck, here’s what I’m seeing, can you help?”, “I can’t solve this CAPTCHA”
- Unexpected errors: “I’m seeing this error, here’s what I’ve tried, what would you like me to do?”
- Accepting “Terms and Conditions”
- Suspicious behavior: “the website you gave me keeps redirecting me to givemedollars.ru so I need guidance”
- etc.
Hope that is helpful! Keep up the great work!
5 Likes
Has anyone discovered ways to manually insert decision points? I keep telling the LAM Playground to “do X and then stop.” and 4 times out of 5 it continues to submit the thing or press the button instead of stopping.
2 Likes
I told LAM r1 to write a draft but not send it on Gmail and it succeeded! 
1 Like