在游戏中学习 AI

What's next?

What's moving fast, the unresolved debates, and how to stay critical over time.

该内容暂无中文翻译,当前以英文显示。 阅读中文完整指南 →
约 9 分钟 8 个步骤 Beginner

完成后你将能够

课程安排 — 8 个步骤
  1. 1AI is moving fast, but nobody knows exactly where it's headed. Even the experts don't all agree.
  2. 2Context windows (how much a model can read at once) keep getting longer.
  3. 3True or false?
  4. 4Sort each statement by its current status
  5. 5What is a "multimodal" model?
  6. 6Tap the passage that describes a misconception
  7. 7Try a multimodal feature (image, audio, or video) of an AI.
  8. 8What's next quiz

建议预修 : AI and jobs

你的进度只保存在本设备上 — 无需账号,无需邮箱。

本页目录

1. What's next?

  1. A path splitting into two branches.

    AI is moving fast, but nobody knows exactly where it's headed. Even the experts don't all agree.

  2. A timeline with three milestones.

    Context windows (how much a model can read at once) keep getting longer.

    Models are becoming increasingly multimodal: a single model understands text, image, audio, and video.

    The cost of getting an answer (inference cost) tends to drop over time.

  3. There is currently a clear scientific consensus on when AGI might arrive.

    AlphaFold is a concrete, verified example of AI applied to a major scientific problem.

    Once you've properly understood AI, there's no need to update your knowledge anymore.

    A multimodal model only understands one type of content at a time.

    A company's flashy announcement alone is enough to judge real progress.

    True or false?

  4. Sort each statement by its current status

    AlphaFold solved a major part of protein structure prediction (recognized by a Nobel Prize)

    A general AI (AGI) will definitely arrive before a specific date

    Model context windows keep getting longer over time

    Every benchmark announced by companies always reflects real general improvement

    Models are becoming increasingly multimodal

    Field experts agree on AGI's exact definition

  5. What is a "multimodal" model?

  6. Tap the passage that describes a misconception

    An impressive announcement of a new model always replaces independent evaluation. Some benchmark results can be optimized specifically for the test without reflecting real general progress. Before judging real progress, it's best to look for independent evaluations.

  7. Try a multimodal feature (image, audio, or video) of an AI.

  8. A path splitting into two branches.

    What's next quiz

    What breakthrough earned Hassabis and Jumper the 2024 Nobel Prize in Chemistry?

    Why isn't a company's flashy announcement enough to judge real progress?

    Model context windows tend to...

    The cost of getting an answer from a given model tends to...

Sort the Trends

60 seconds: sort each statement by its current status.