One of the remarkable security of LLMs post on this blog basically made me realize that we are not thinking enough about using an army of LLM agent swarm and rather focusing too much on single large model doing everything right.
In real world good things happen when people with different specializations come together to build something together. Sometimes there are trade offs between two departments competing for different conflicting goals.
Consider an automobile company where one team wants bigger engine and another team wants better fuel efficiency. Either ways both are at odds and they have to figure out how to resolve it.
We need to adapt a similar approach for LLM agent deployments as well. Have multiple agents solving the problem together instead of one large model solving it.
Top comments (0)