Google I/O Developer Keynote [video] | Hacker News Reader

Google I/O Developer Keynote [video]

events.google.com

98 pointsLaurensLang21 comments

Some of my highlights:

- The big move to on-device ML for things like speech recognition. This was especially demonstrated with an accessibility angle, showing global subtitles available for any application. And an assistant that allowed you to answer and reply to phone calls using TTS and back.

- Android Q big focus on privacy, promoting privacy to a top level menu item in settings, as well as adjustments in apps to provide ready-access to privacy features.

- Large focus on security, including calling out Gartner reports many times, having the most secure operating system in a number of tests, and most secure device (Pixel).

- Google Home products rebranded as Google Nest. Launch of Google Nest Hub Max ($229), with Camera allowing for Duo calling, and a nice hand gesture to silence the device when loud/noisy (no more shouting)

- Pixel 3a (XL) devices launched, at a lower price point of $399 ($479) with a decent set of features. No longer Verizon exclusive.

- Google wants to be "Helpful"

They showed a ton of advanced ML-based features in the demo. I can barely remember all the names.

Obviously Google is throwing a ton of effort into these features, but I am really curious what the productization plan is behind them.

It doesn’t look like they are differentiating their hardware on those features (or maybe I missed it), and it doesn’t look like those features are products on their own.

I’m curious what the long term business plan is behind “Helpful”

> I’m curious what the long term business plan is behind “Helpful”

Product stickiness? For the competitors, these features are nontrivial to copy. If they work well, people get "addicted" to them.

Sounds like exactly what Apple said when they launched the X. All on board AI, secure enclave, no shipping your data off to the cloud for security.

The issue with Google is that they're an advertising company, so they're always going to have an incentive that pushes them to market your data unlike Apple.

I tend to favor Apple over Google, and am more inclined to use Apple hardware than any Google/Pixel hardware, and I have no devices that have always-on mics (aside from my iPhone), but...

Apple News is already less usable and readable because of Apple's included ad layers.

Apple's ad business isn't a sizable fraction of Google's, but Apple's direction is plenty ad-oriented, so this old canard will probably need to drop by the wayside at some point. It's not entirely meaningful.

Then why are they releasing on-device voice recognition to replace their cloud-based voice recognition?

Because it's faster and works with spotty/no reception?

(I'm just guessing because I can't find details on this stuff by wading through the feed.)

Also, just because it's doing local inference doesn't mean it won't be uploading data back to Google periodically. As GP said, they have a strong incentive to do this, so it's reasonable to just assume they will do this until they can prove otherwise.

And doing the inference locally, and just uploading transcripts for the most part is way more cost effective. They don't have to pay for the power on these endpoints.

Well Apple is exerting market pressure, there's been a lot of privacy backlash lately, and they can make customers feel safe while still exfiltrating their data via other services that you'll be encouraged to use by the "secure platform".

Because they calculated that taking a small L here would turn into a bigger W later. Comparatively, Apple is just trying to sell more units, not prop up its Ads business down the road with higher user counts giving more impressions.

Maybe those large (yet still too small) fines and constant criticism finally got through.

You just can't please some people...

Anyway, it's not all on device. There's still data collection. It's collecting some kind of post inception Delta instead of raw data.

That live transcription built into Android Q that works on device and for any app is incredible. I can see so many uses for that myself, I cannot imagine how awesome something like that would be for people who have difficulty hearing.

On device speech recognition is pretty great, but as a whole the keynote was pretty conservative. Probably because they promised too much stuff in previous years that either wasn't released or quickly failed.

As models start fitting on devices, what keeps it a competitive advantage? Are specific models patentable? Copyrightable?

What's stopping Apple or a foreign company from shipping these models themselves?

Creating & improving that model is the competitive advantage. It's like saying what is the competitive advantage of compiled executables that run on customer devices. I estimate it's copywriteable.

But any would be copy-cat could simply push out the new model a day later.

I'm also struggling with whether such a model should fall under copyright. Models would naturally converge.

copyrightable?

I for one would definitely like to pirate such a model for use on desktop applications. This seems better than many existing local systems...

I think they still have the advantage of improving the base models ahead of anyone else. The fitting part I assume is tailoring an existing model to a specific user on device. They can improve the base model and have the competitive advantage by being first to market with ever improved model.

The on-device Machine Learning sounds pretty amazing. Does anyone know if this will be a part of AOSP or be a closed-source extension (like digital wellbeing)?

Reply on news.ycombinator.com