‧ 22 則回應 ‧ 無 ‧ 表情差分,背景完全不變的情況 ‧ 例如AI不能幫小孩寫作業 ‧ 溝通 ‧ 沒想過 ‧ 跨部門整合,資源協調,人力管理 ‧ 可以看出目前AI的效果都不理想,破綻很多 ‧ 1.更精準的文獻查找 ‧ 2.AI繪圖的人設統一 ‧ 3.AI繪圖的商業化 ‧ 邏輯、創意 ‧ 針對片段資訊或時間/空間跨度較大的關聯資訊,常有文不對題或鬼打牆的對應 ‧ 上題已經回答...重複的問題 ‧ 他的能力有限,因為我使用的是免費的 ‧ 真正的做商業簡報,而不是花拳繡腿那種 ‧ 還要判斷AI的成果否正確 ‧ 企業商用 ‧ No ‧ 講出特定元素並介紹它(ai無法正確的解釋) ‧ 人跟人的感情 ‧ 英文人名會常被識別為錯字 ‧ Ai製作動畫成效不好 ‧ 8 則回應 ‧ Cannot be able to have face-to-face interaction ‧ When research, is not always accurate ‧ creativity ‧ adaptability ‧ relatable sources ‧ Creativity ‧ Anything that cannot be AI-splained within its limited token context memory, that you'd need to "puzzle together yourself". I feel like I have totally random "islands of knowledge" around the CLIP model and what awesome stuff can be done with it (ViT deep dream, GradCAM, gradient ascent image-to-text)... But what else is there? What am I missing? What other cool stuff can I do with this? And how does it all fit together? That's hidden in a paper of crazy-ass mathematics, and it's too broad for AI's token memory. :-) ‧ My assignment