TL;DR 336 | The Google Developer News Show
0:00 - Introduction 0:10 - Wonderous nominated for Webby Award → https://goo.gle/40LIliK 0:28 - How WebAssembly is accelerating new web functionality →https://goo.gle/3ZR0L04 0:54 - Scaling vision transformers to 22 billion parameters →https://goo.gle/3mjc7fG 1:29 - Customizing Compose Pager with fun indicators and transitions →https://goo.gle/43dVOS9 1:48 - Don’t forget to like, comment, and subscribe!
Here to bring you the latest developer news from across Google is Rody, a Developer Advocate. Tune in every week for a new episode, and let us know what you think of the latest announcements in the comments below.
Follow Google Developers on Instagram → https://goo.gle/googledevs
Watch more #DevShow → https://goo.gle/GDevShow
Subscribe to Google Developers → https://goo.gle/developers
#Google #Developers
0:00 · [AUDIO LOGO] RODY DAVIS: Hi, I’m Rody for The Developer Show.
0:06 · And this is your weekly update on the coolest developer news from Google.
0:10 · In case you missed it, the Flutter app, Wonderous, just got nominated for a Webby in the best user interface category.
0:16 · You can vote for the app on the Webby Awards website.
0:19 · This app was made in partnership with the gskinner team and since launch has had 30,000 users.
0:24 · To learn more about Wonderous, check out the Flutter blog on Medium.
0:28 · Next we can head over to Chrome to talk about all the exciting ways WebAssembly is accelerating the web.
0:33 · Some powerful real-world examples include FFmpeg on Wasm, background blurring on Google Meet, SQLite on Wasm, which replaces Web SQL, and the universal scene descriptor in Safari.
0:45 · WebAssembly brings faster iteration speed, immediate support across browsers, and improved security.
0:51 · To learn more about WebAssembly, check out the Chromium blog.
0:55 · If you’re building AI vision models, then you’ll be excited to learn about how to scale vision transformers to 22 billion parameters.
1:01 · LLMs, or Large Language Models like PaLM or GPT-3 show that scaling parameters improves performance but also unlocks emergent abilities.
1:10 · But the largest dense models have only reached around 4 billion parameters.
1:15 · Thanks to VIT-22B, we can incorporate ideas from scaling models like PaLM to a new modified architecture to scale to 5.5 times larger.
1:24 · To learn more about this new technique, check out the Google Research blog.
1:28 · Finally, we can talk about the March 2023 release for JetPack Compose.
1:33 · There are new composables introduced— horizontal and vertical pager.
1:37 · There are lots of fun transition effects and behaviors that can be applied when swiping through content.
1:42 · You can also add page indicators, change the viewport offset fraction in the current page and the page state.
1:48 · To learn more about all this week’s stories, make sure to check out the description box below for all the links.
1:53 · Please remember to like, subscribe, share, and stay safe.
1:56 · I’m Rody for The Developer Show.
1:58 · Thanks for watching, and we’ll see you next week.
2:00 · [MUSIC PLAYING]