影片進階EN8,603 次觀看
Why LLM Recommenders Will Be AI's Biggest Consumer App — Devansh Tandon, Meta
來源 AI Engineer
看影片(在新分頁開啟原站)連到 AI Engineer
摘要
Devansh Tandon 指出推薦系統遵循與大型語言模型相似的縮放曲線,並提出用語義 ID 結合預訓練與後訓練構建 LLM 推薦器。這種方式能透過解碼現有內容指標而非生成新內容,使資訊流比聊天應用程式便宜 100 倍,有望成為 AI 最大的消費端應用。
Devansh Tandon argues that LLM-based recommendation systems follow scaling laws similar to LLMs and can be built using semantic IDs to create cheaper, steerable consumer feeds.
摘要、重點與章節標題由語言模型整理,細節(誰說的、數字、先後)可能有誤;要引用請以原始內容為準。
重點
- 推薦系統遵循與大型語言模型相似的資料、運算與模型規模縮放曲線。
- 用語義 ID 壓縮內容並結合預訓練與後訓練,可構建可解釋的 LLM 推薦器。
- 資訊流透過解碼現有內容指標,比生成式聊天應用程式便宜 100 倍。
章節
依話題轉折切分,標題由 AI 產生
- 00:00Intro: two big arguments
- 00:45About Devansh
- 01:30Semantic IDs and generative retrieval go mainstream
- 02:15Scaling laws, from LLMs to recommenders
- 03:45Real scaling curves at Meta
- 04:15Instagram Reels: 30% more watch time
- 04:55The tokens in, engagement out flywheel
- 05:40Four S-curves of recommendation
- 06:35LLM-native recommenders
- 06:50Agentic recommenders
- 08:05The three-step recipe for an LLM recommender
- 08:35The five-layer cake
- 09:35Semantic IDs: a Reel in 10 tokens
- 10:45Pre-training on English and semantic IDs
- 11:25Post-training: an LLM re-ranker with chain of thought
- 12:25Steerable feeds: talking to your Instagram algorithm
- 13:15Explaining recommendations
- 13:50Prompted playlists, custom feeds and Ask DoorDash
- 14:05Feeds vs. chat apps: the same flywheel
- 14:50Why feeds are 100x cheaper per hour of engagement
- 16:25Why LLM recsys is AI's biggest consumer app
- 17:40Wrap-up
適合誰看
從事推薦系統研發、內容平台產品設計或關注 AI 在消費端應用趨勢的技術人員與管理者。
摘要依據
- 講者
- Devansh Tandon
- 依據
- 自動字幕
為什麼排在這裡
- 人氣
- 0.74
- 新鮮
- 0.96
在主題頁與搜尋結果裡,名次由相關、人氣、新鮮三個分數決定;這一頁沒有搜尋的關鍵字,所以沒有相關分數。排序怎麼算
相關內容
- Training an LLM-RecSys Hybrid for Steerable Recs with Semantic IDs文章 ・ Eugene Yan(部落格)
- How LLMs Are Reshaping Recommendation SystemsPodcast ・ Software Engineering Daily ・ 48 分鐘
- AI Engineer 2025 - Improving RecSys & Search with LLM techniques文章 ・ Eugene Yan(部落格)
- Augment or Not? A Comparative Study of Pure and Augmented Large Language Model Recommenders影片 ・ 陳縕儂 Vivian NTU MiuLab ・ 17 分鐘(在新分頁開啟原站)
- Distill the LLM, Don't Serve It: Search & Personalization at DoorDash — Raghav Saboo, DoorDash影片 ・ AI Engineer ・ 22 分鐘
- What messages you send to the LLM actually look like影片 ・ Matt Pocock ・ 2 分鐘
摘要由 AI 根據原文產生,可能有誤;完整內容請看原站。看影片(在新分頁開啟原站)
