GPT-4’s Gaming Abilities: Playing Doom with a Violent Streak


A Microsoft scientist and a University of York researcher tested the GPT-4 AI’s ability to play Doom without prior training. The AI could open doors, fight enemies, and fire weapons with some degree of effectiveness, but it lacked object permanence and awareness beyond its narrow scope.

The researchers expressed concern about the ease with which the model could be trained to shoot and urged caution when deploying such models. The article also highlights the importance of considering ethical implications and the potential misuse of AI technology.

Read more: GPT-4 can play Doom, badly — doesn’t hesitate to shoot humans and demons

Check Out The New TalkDev Podcast. For more such updates follow us on Google News TalkDev News.