文章を生成させない場面でLLMを使うのは、そろそろ割に合わない気がしています。OpenAIのDecisions APIが公開ベータ入り。返すのは文章ではなく、型付きの確率・選択肢・スコアで、従来のResponses APIより約10倍高速だそうです。生成と判定を分離する、この設計がついに大手APIにも届いた意味は大きい。分類やルーティングをわざわざLLMに投げている処理は、まずここへ置き換える価値がありそうです。https://code-rebuild.com/ai/typesafe-jev-usage-java/ #判定モデル #LLM #OpenAI #Jev
Related
BREAKING: Nvidia's Jensen Huang faces scrutiny over alleged China tiesCongress is set to push Huang for a briefing over ...
BREAKING: Nvidia's Jensen Huang faces scrutiny over alleged China tiesCongress is set to push Huang for a briefing over his links to a Megaspeed executive US officials have called ...
Apple's macOS still has Script Editor to write simple tools and automations in AppleScript and JavaScript.I wonder why i...
Apple's macOS still has Script Editor to write simple tools and automations in AppleScript and JavaScript.I wonder why it hasn't been enhanced to use LLMs and tools. It would look ...
💻🇬🇧 JUST IN: UK chip startup Fractile raises $720M to build purpose-made AI inference chips#AI #Tech
💻🇬🇧 JUST IN: UK chip startup Fractile raises $720M to build purpose-made AI inference chips#AI #Tech