Artificial intelligence (AI) has seen drastic advances given recently powerful models with astonishing individual capacities, while the biological evolutionary strategy focuses more on collective intelligence. We seek to bridge biological collective intelligence with artificial intelligence, by studying collective motion of agents, inspired by the biological ants collectively solving tasks while using chemical pheromone for communication. We train agent in a single setting to acquire chemotaxis and duplicate the trained agents to form a population. We observe several interesting dynamics where collective intelligence is realized in our AI models, and expect to further analyze the impact of communication on collective dynamics.