diff --git a/doc/source/example-a3c.rst b/doc/source/example-a3c.rst
index 3037b2b6b132..f8a8bfb4c1f3 100644
--- a/doc/source/example-a3c.rst
+++ b/doc/source/example-a3c.rst
@@ -127,7 +127,7 @@ global model parameters. The main training script looks like the following.
       obs = 0
 
       # Start simulations on actors
-      agents = [Runner(env_name, i) for i in range(num_workers)]
+      agents = [Runner.remote(env_name, i) for i in range(num_workers)]
 
       # Start gradient calculation tasks on each actor
       parameters = policy.get_weights()