AI in video games. Development approaches, tools, cheats

What the talk is about

Oleksandr Pindyk and Oleksandra Korobenkova explain how game AI is approached in World of Tanks. Bots are used to onboard new players, run events, prototype features and test maps. The key idea is that game AI doesn't need to be truly smart; it must support the intended gameplay and look convincing, since obviously silly mistakes break immersion. A neural network isn't always a fit because requirements are incomplete and change, and designers want authorial control. The speakers go through bot cheating: acceptable in single-player and tutorials, but not in PvP, which calls for "fair" AI playing by the same rules as a human. It relies on pre-built map data (navmesh, cover, zones), sensors and limited knowledge of enemies. Difficulty is tuned through vehicle presets and bot skills. Useful for game designers and AI developers.