
OpenAI ライブストリーム
OpenAIがライブストリーム配信イベントを開催します。放送中に具体的な発表、新製品発表、またはデモンストレーションが明かされる予定です。
OpenAIが前回予告なしのライブストリームをやった時、GPT-4 Turboをドロップして、一晩で価格設定を完全に変えました
![I Trained an AI to Beat Final Fight… Here’s What Happened [p]](https://external-preview.redd.it/dWZSKa_lMUvycB0q8xwsIkTgDpHLe-W2-Q_S7RwWucQ.jpeg?width=320&crop=smart&auto=webp&s=6a0cfa02507091d2949c8f6b2fcd59a254a23929)
Hey everyone, I’ve been experimenting with Behavior Cloning on a classic arcade game (Final Fight), and I wanted to share the results and get some feedback from the community. The setup is fairly simple: I trained an agent purely from demonstrations (no reward shaping initially), then evaluated how far it could go in the first stage. I also plan to extend this with GAIL + PPO to see how much performance improves beyond imitation. A couple of interesting challenges came up: Action space rem