AI at Play
Because games make great AI benchmarks
Hello!!
POP QUIZ! COCO (Common Objects in Context) Image Pairs
Shout the right text subtitle - you have 1 second!
Games are DENSE
- There is loads of stuff
- And it’s not just showing it, it’s what it means
POP QUIZ! What do we do here?
Shout the right move - you have 1 second!
The possibility space
- You can do loads of stuff
- And often there is no right answer
ai-at-play.online (it is very silly)
But also maybe interesting?
But what about GPT-5 / OSS?
Some models are violent
Others are chatty