#computing #ai https://arxiv.org/pdf/2106.01345.pdf We can solve [[Reinforcement Learning]] using [[Transformer]]s, i.e. basically all it takes to hack online games is to decrypt packets and throw into a transformer then send back to server... Okay, sure, there is probably more steps in practice, you could build datasets of players playing, sniffing their computer network communications and fine-tune the transformer accordingly to build a sort of [[Imitation learning]] model. ```py packet_to_send = openai.complete("PACKET_THAT_SAY_I_ATTACK_YOU ->") # Here packet to send would be something that is usually # sent after receiving "i attack you", typically, idk, I defend myself... bot.send(packet_to_send) ```