Many users reacted negatively to the AutoBots project on fully autonomous self-improving AI agents, warning that removing human oversight risks unchecked errors, misaligned goals, and dangerous optimizations like pursuing more compute.
Based on 9 visible X reactions from 27 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@bindureddy No human in the loop means no error correction. The loop optimizes the metric, not the goal, and small misreads compound with nothing to catch them. Self-improvement without a ground-truth check raises the score, not the result. The human was the correction, not overhead.
@bindureddy No human in the loop is the part that should worry people, not excite them. What happens when the six models disagree on the goal itself. @bindureddy
@bindureddy Calling a 6-LLM ensemble cost optimized is cute, but I'll believe the no-human-in-the-loop part when one actually ships.
@bindureddy You are working on AutoBots, great. The question is where are the Decepticons, and who's working on them?
@bindureddy No human in the loop means no error correction. The loop optimizes the metric, not the goal, and small misreads compound with nothing to catch them. Self-improvement without a ground-truth check raises the score, not the result. The human was the correction, not overhead.
@bindureddy No human in the loop is the part that should worry people, not excite them. What happens when the six models disagree on the goal itself. @bindureddy
@bindureddy Calling a 6-LLM ensemble cost optimized is cute, but I'll believe the no-human-in-the-loop part when one actually ships.
@bindureddy You are working on AutoBots, great. The question is where are the Decepticons, and who's working on them?
@bindureddy self improving agents sound terrifying tbh
@bindureddy That's great.
Currently working on AutoBots Completely autonomous cost optimized self-improving agents that do not require a human in the loop Simply set a bunch of goals and they will use a mixture of 6 LLMs to constantly improve and reach them
AutoBots - complex agentic loops that don’t require a human-in-the-loop They will constantly self improve on a set of goals Testing it as we speak 🚀
Many users reacted negatively to the AutoBots project on fully autonomous self-improving AI agents, warning that removing human oversight risks unchecked errors, misaligned goals, and dangerous optimizations like pursuing more compute.
Based on 9 visible X reactions from 27 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@bindureddy That's great.
Currently working on AutoBots Completely autonomous cost optimized self-improving agents that do not require a human in the loop Simply set a bunch of goals and they will use a mixture of 6 LLMs to constantly improve and reach them