드디어 WebGPU 혁명이 왔다💥 장단점. 사용법 설명해드림!

Introduction to Web GPU

In this section, we will learn about the new graphics API called Web GPU and how it is a game-changer for web applications.

Evolution of Graphics APIs

• Browsers used to rely on Flash for drawing images on the user's screen.

• WebGL was created as a solution to get native graphics on the web without using third-party plugins.

• HTML5 canvas element was groundbreaking but had limited features and was not performant enough for high-performance games.

• WebGL brought 3D rendering to the web but was based on an old API, OpenGL, which is no longer in active development.

Enter Web GPU

• Web GPU is a new browser API that allows websites to talk to modern graphics APIs like Vulcan, Direct 3D, and Metal.

• It unlocks a new platform for gaming and enables desktop software that uses GPUs like video and photo editing programs to become truly web-based.

• Web GPU can be huge for AI since GPUs are heavily used in training machine learning models. Using Web GPU, users can run machine learning models from their browsers without uploading any private data anywhere.

How Web GPU Works

In this section, we will learn how Web GPU works and how it allows websites to have more control over the user's GPU.

Talking to the User's GPU

• The browser takes the WebGPU commands and talks directly with the user's machine using its native APIs (Vulcan, Direct 3D or Metal).

• Websites now have more control over the user's GPU, unlocking a new platform for gaming and enabling desktop software that uses GPUs to become truly web-based.

• Web GPU can be used to run machine learning models from the browser itself without uploading any private data anywhere.

Conclusion

In this section, we will summarize the key takeaways from the transcript.

Key Takeaways

• Web GPU is a new graphics API that allows websites to talk directly with modern graphics APIs like Vulcan, Direct 3D, and Metal.

• It unlocks a new platform for gaming and enables desktop software that uses GPUs to become truly web-based.

• Web GPU can be huge for AI since it allows users to run machine learning models from their browsers without uploading any private data anywhere.

Using Web GPU

In this section, the speaker discusses how most people will use web GPU indirectly through libraries like 3js, Babylon.js, or TensorFlow.js. They also mention that web GPU is not limited to JavaScript and can be used with languages like Rust.

Libraries for Web GPU

• Most people will use web GPU indirectly through libraries like 3js, Babylon.js, or TensorFlow.js.

• These libraries support web GPU and make it easier to use.

Other Languages for Web GPU

• Web GPU is not limited to JavaScript only. You can also use languages like Rust with libraries like wgpu to write web GPU code.

• You can run that code on the browser of the user by compiling it to WebAssembly.

Risks of Using Web GPU

• Websites can now talk to your GPU which means that in theory a website could use this new feature and use your GPU to mine a cryptocurrency every time you visit it draining your resources and battery life or it could potentially be used for anonymizing and tracking users across the web.

• We will have to wait and see how browsers implement web GPU maybe the user will first have to give permission just like with geolocation which would be a good thing to do.

Conclusion

In this section, the speaker asks for feedback from viewers about their thoughts on using web GPUs and if they feel comfortable allowing websites access to their GPUs.

Feedback from Viewers

• The speaker asks viewers what they think about using web GPUs and if they feel comfortable allowing websites access to their GPUs.

• Viewers are encouraged to leave comments with their thoughts.

Free Courses

• The speaker reminds viewers that they can learn things like JavaScript, Python, React, React Native, Go, Dart, Flutter and many others for free by clicking the link below to join any of the many free courses available.

Closing Remarks

• The speaker thanks viewers for watching and encourages them to subscribe to the channel if they found the video useful.

• They sign off by saying "see you on the next one."