在如今拥有海量影视内容的时代，我们常陷入「刷剧」的困境，漫无目的地浏览却难以找到真正想看的。本期节目探讨了影视推荐机制的复杂性，以及 AI 是否能改善这一状况。虽然 AI 能利用庞大的数据进行分析，建立跨作品的联系，但最终推荐仍是个人问题，因为我们的喜好和观影选择都太过复杂，无法完全被 AI 理解。节目还介绍了一些利用 AI 工具加快找到想看内容的方法。

🤔 AI 能够利用庞大的数据分析，建立跨作品的联系，例如将不同作品的剧情、评论、人物、主题等进行关联，从而发现传统推荐算法难以发现的联系，更精准地推荐内容。

🤯 AI 模型可以像人类一样理解一部电影的整体内容，通过分析整部电影的剧情、画面、音乐等元素，进行更深层的理解，例如识别电影的风格、主题、情绪等，从而更好地进行推荐。

😔 尽管 AI 技术不断发展，但最终推荐仍是个人问题，因为我们每个人的喜好和观影选择都太过复杂，无法完全被 AI 理解。例如，我们可能喜欢特定类型的电影，但又不喜欢特定的演员，或者我们可能喜欢某部电影，但不喜欢该电影的导演的其他作品。

💡 节目还介绍了一些利用 AI 工具加快找到想看内容的方法，例如使用 Movievanders 和 Reelgood 等工具，通过 AI 帮助你快速找到想看的电影和剧集。

🚀 未来，AI 技术可能会进一步发展，帮助我们更好地理解和推荐电影和剧集，但最终还是要依靠我们自己的判断和选择，才能找到真正想看的作品。

It’s one of the most common low-stakes annoyances in modern life: you flop down on the couch at the end of the day, finally with a few minutes to watch one of the dozens of incredible shows or movies you have access to thanks to the peak TV era and the advent of streaming, and you start scrolling. Instead of actually watching anything, you spend an interminable evening opening apps, aimlessly scrolling through endless rows of same-looking tiles. You eventually give up and watch The Office again.

On this episode of The Vergecast, we look at why TV and movie recommendations are so complicated, and whether AI might be able to make them better. If Spotify can build infinite playlists of music you’ll like, and YouTube and TikTok always seem to have the perfect thing ready to go, why can’t Netflix or Hulu or Max seem to get it right?

AI, it turns out, can help at least a little. Because models from OpenAI, Google, and others have ingested so much information about movies and shows — not just their title and genre, but all the synopses, reviews, recaps, and more from all over the web — they can synthesize that information and find connections between titles that were previously hard to find. And as context windows get larger, these models can actually ingest and understand an entire film at once, which opens up entirely new ways of understanding them.

Ultimately, though, recommendations are a human problem. Because we’re all human. What you want to watch, and why you like what you like, are far more complicated — and vary far more widely — than even the best model can understand. As a result, the idea of sitting down, opening Netflix, and having the exact right title appear immediately, isn’t coming true anytime soon. So instead of hoping for the best, we investigate the ways to use AI tools right now to get to your content at least a little faster. Because watching movies great; scrolling through too many of them is seriously overrated.

If you want to know more about everything we discuss in this episode, here are a few links to get you started:

MovievandersReelgoodThe internet is a constant recommendations machine — but it needs you to make it workNetflix’s Greg Peters on a new culture memo and where ads, AI, and games fit inFrom Scientific America: How Recommendation Algorithms Work—And Why They May Miss the MarkFrom Google: Multimodal prompting with a 44-minute movie