My Notes from Google I/O 2022
XR, Ambient Computing, Spatial Computing, AR…
My Notes from Google I/O 2022
XR, Ambient Computing, Spatial Computing, AR…
I loved it looking at how every announcement throughout the I/O was connected to Google’s mission of organizing the world’s information and make universally accessible and useful. We could see how google is pushing its services and products to the ambient world.
- Monolingual translation — A new model to help translate languages that have less digitized data, solving the limitations of the current models which are dependent on these data.
https://ai.googleblog.com/2022/05/24-new-languages-google-translate.html - A deep learning model to determine the footprints of buildings from high-resolution satellite imagery. https://ai.googleblog.com/2021/07/mapping-africas-buildings-with.html
https://sites.research.google/open-buildings/ - Immersive View — A 3D immersive map generated using photos helps you to get immersed in the place. Guess this would play a vital role in maps stepping into the spatial computing world. 🤞https://blog.google/products/maps/three-maps-updates-io-2022/
Immersive Map of London demonstrated in Google I/O
- Summarization — We are generating quintillion bytes of data but for sure our human ability is limited that is where the human-machine partnership comes into play. These new set of features from google on transcription, translation, and summarization would also play a vital role in the new spatial world.
- Project Starline — This has been in the works since the last I/O but the impact it would create in immersiveness of the spatial world would be very high. Some progress highlighted this year is auto light and Realtone to meet calls.
https://blog.google/technology/research/project-starline/
Another key highlight was Monk Tone scale, which could help the new digital world to be inclusive and give the right representation in the virtual world.
Skin Tone Research @ GoogleIntroducing the Monk Skin Tone (MST) Scaleskintone.google
- Multisearch — Content doesn't come in a single form it can be a written word, spoken lines, or an image. With the Multitask Unified Model or MUM Google is building a search that works for the ambient world. This also enables natural conversation with machines similar to how we interact with other human beings using multiple senses.
https://blog.google/products/search/multisearch/
https://www.blog.google/products/search/introducing-MUM/
In action,
Look and talk, Quick Phrases
- Scene Exploration — An new feature on google lens to explore your surrounding, the machine recognizes the objects and helps you with relevant information on search queries. This would be vital in the spatial computing space, already google lens as an engine could be a augment engine that helps the machine to recognize the world and be contextual this is another feather to the cap.
https://blog.google/products/search/search-io22/
- AI Test Kitchen, LaMDA 2, and PaLM
Through its Language Model for Dialogue Applications and Pathways Language Model Google is trying to make the machines communicate with humans more naturally.
In action,
Imagine It, List It and Talk About It
https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html?m=1
https://aitestkitchen.withgoogle.com/
https://blog.google/technology/ai/lamda/
- Augmented Live Caption
There were lot more announcements which helps google to be more accessible, eco friendly. These are updates announced in the main keynote and will add more as I explore this year's I/O.