AI Struggles to Play Classic Game ’Doom’
8
3
A new research project introduced VideoGameBench, an AI benchmark designed to test whether state-of-the-art vision-language models can play 20 popular video games. Despite the advancements in AI, models like GPT-4o, Claude Sonnet 3.7, and Gemini 2.5 Pro struggle to play the classic first-person shooter Doom due to high inference latency. When an agent takes a screenshot and queries the model about what action to take, the game state has already changed by the time the response comes back.
By:
Previous
BIS Crypto Isolation Push Is ‘Dangerous,’ Warns Blockchain Investment Firm Chief
Next
BRICS Alliance Announces New Payment System Aiming to Challenge US Dollar Dominance
