HN Theater @HNTheaterMonth

It may not have been easy to see that another breakthrough is required, but it was very easy to see that most predictions were laughable.

The CEO of Waymo announced[1] in May 2018:

> Phoenix will be the first stop for Waymo's driverless transportation service, which is launching later this year. Soon, everyone will be able to call Waymo, using our app, and a fully self-driving car will pull up--with no one in the driver's seat-- to whisk them away to their destination. And that's just the beginning!

[1] https://youtu.be/ogfYd705cRs?t=5795

It's not really that much easier, but their new AI let's you type certain common responses like "Thank you, I'll look at this" in an instant. They made a presentation about this.

https://www.youtube.com/watch?v=ogfYd705cRs

⬐ bhandziuk

ah yes. I found those just add a lot of cognitive load to no benefit. By the time I've read "thanks!", "Yes", "How about tomorrow?" I could have just typed a response. Maybe it's just me but I basically never replay to an email with such terse responses as the options provided in Gmail.

They seem to have solved the snow problem, at least having a viable solution for it.

https://youtu.be/ogfYd705cRs?t=5936

In the last keynote they mentioned that they have solved some snow problems: https://youtu.be/ogfYd705cRs?t=6177 :)

Yup I believe they announced this at Google I/O Keynote this year actually -- they mentioned that while it might not be enough with just the audio, looking at mouth gestures can give AI enough to know what might be said by whom.

https://www.youtube.com/watch?v=ogfYd705cRs&t=7m0s

⬐ pbhjpbhj

Reflecting on your last phrase: I was watching Esports the other day, the player was talking with a very loose mouth and I wondered if he was avoiding being lip-read.

I imagine a lip reading AI would get wide usage. Managers will be wearing face masks to hide their lips. (It's probably doable now to listen with a spy-mic, but it's obvious in a way that using a normal video camera isn't).

⬐ laszlokorte

Julia Probst is a german deaf blogger [1] who is famous [2] for lipreading tactical commands given by soccer coaches to their team during a match and posting them on twitter. She is even hired by sport channels on TV to provide the commentator with insight information.

[1] https://twitter.com/einaugenschmaus [2] http://www.sueddeutsche.de/sport/lippenlesen-im-fussball-die...

Sorry, they're very likely running A/B testing of some sort, as this link was AV1 (Codecs: avc1.64002a (299) / mp4a.40.2 (140)) for me earlier, but no longer:

https://www.youtube.com/watch?v=ogfYd705cRs

⬐ wolf550e

avc1 is https://en.wikipedia.org/wiki/H.264/MPEG-4_AVC not AV1

The available formats for that video are:

    $ youtube-dl -F https://www.youtube.com/watch?v=ogfYd705cRs
    [youtube] ogfYd705cRs: Downloading webpage
    [youtube] ogfYd705cRs: Downloading video info webpage
    [info] Available formats for ogfYd705cRs:
    format code  extension  resolution note
    249          webm       audio only DASH audio   70k , opus @ 50k, 37.58MiB
    250          webm       audio only DASH audio   88k , opus @ 70k, 45.92MiB
    140          m4a        audio only DASH audio  136k , m4a_dash container, mp4a.40.2@128k, 98.65MiB
    171          webm       audio only DASH audio  148k , vorbis@128k, 70.46MiB
    251          webm       audio only DASH audio  158k , opus @160k, 95.71MiB
    160          mp4        256x144    144p  114k , avc1.4d400c, 30fps, video only, 29.91MiB
    278          webm       256x144    144p  147k , webm container, vp9, 30fps, video only, 61.11MiB
    242          webm       426x240    240p  226k , vp9, 30fps, video only, 75.75MiB
    133          mp4        426x240    240p  252k , avc1.4d4015, 30fps, video only, 54.98MiB
    243          webm       640x360    360p  413k , vp9, 30fps, video only, 140.49MiB
    134          mp4        640x360    360p  644k , avc1.4d401e, 30fps, video only, 182.59MiB
    244          webm       854x480    480p  764k , vp9, 30fps, video only, 227.69MiB
    135          mp4        854x480    480p 1193k , avc1.4d401f, 30fps, video only, 377.25MiB
    247          webm       1280x720   720p 1521k , vp9, 30fps, video only, 423.38MiB
    136          mp4        1280x720   720p 2375k , avc1.4d401f, 30fps, video only, 730.12MiB
    248          webm       1920x1080  1080p 2674k , vp9, 30fps, video only, 810.25MiB
    137          mp4        1920x1080  1080p 4461k , avc1.640028, 30fps, video only, 1.34GiB
    17           3gp        176x144    small , mp4v.20.3, mp4a.40.2@ 24k, 57.43MiB
    36           3gp        320x180    small , mp4v.20.3, mp4a.40.2, 162.63MiB
    18           mp4        640x360    medium , avc1.42001E, mp4a.40.2@ 96k, 332.87MiB
    43           webm       640x360    medium , vp8.0, vorbis@128k, 452.27MiB
    22           mp4        1280x720   hd720 , avc1.64001F, mp4a.40.2@192k (best)

⬐ markdog12

Ah, my brain missed the 'c', thx for that

https://youtu.be/ogfYd705cRs?t=6m54s

Edit: also, a blog post with more examples and a link to the related publication: https://ai.googleblog.com/2018/04/looking-to-listen-audio-vi...

The announcement bit on video https://youtu.be/ogfYd705cRs?t=1h36m

Excerpt from Google IO 2018 keynote - https://www.youtube.com/watch?v=ogfYd705cRs

⬐ ocdtrekkie

There is some dystopian crud in here in just the few minutes I've watched.

- Google will write your emails for you so you don't have to. Is this Gmail user writing me an email or is Google? I no longer know.

- Google Assistant can now impose Google's parenting values on your kids by making them say please. Combine with the mandatory Chromebooks in my school district for full-on indoctrination from childhood.

I am so uncomfortable with this. All the people cheering for it is even worse.

EDIT: HOLY CRUD, they're unleashing their AI to call and harass people on the phone.

⬐ sulam

To be honest Gmail auto-replies have been amazingly on point for me. The reply options they give me in quick, transactional emails are generally _exactly_ the sort of reply I would have typed out, so they save me a lot of time.

⬐ dragandj

Until you train yourself to respond with such limited set of pre-baked answers to everything!

⬐ sulam

We don't know each other (I don't think!), but trust me, no one who knows me would believe you if you said I'm at risk of that!

⬐ dragandj

My comment does not implicate that you as a single person necessary have to succumb to this, at least immediately, or that you are not a good person in any way. By you, I implied every person using that technology over time, including me, you, and everyone else. Short-term, we might control this, but over time it is inevitable. Something that is done over and over again becomes automatic. That's called training and conditioning :)

⬐ Osmium

You can imagine a feedback loop though:

Some people use Gmail auto-replies more > Other people therefore receive emails in the style of these auto-replies > These other people then start manually sending emails according to this style > More people then become happier using auto-reply (since it produces replies in a style they've come to expect)

In this way, an auto-reply feature, even if used by only a subset of the population would start to move language trends.

Maybe there's nothing wrong with this, I don't know. But there is a feedback loop there, which means there's a blurrier line than we'd like between "this AI algorithm does what I'd expect" to "my expectations have become shaped by AI algorithms."

⬐ earthboundkid

This happened to Google Translate. They were scouring the web looking for bilingual corpuses to ingest and they found… texts translated by Google Translate. Things got funky and they had to stop.

⬐ threatofrain

The majority cheers for parental control. Having control over a child's laptop comes as psychologically obvious to most parents.

⬐ reaperducer

Consider the audience. 20-something and childless, with an underdeveloped sense of right and wrong, which is what has led to so many problems in the tech arena today.

⬐ mFixman

Get some context, FFS. These are seem close to "dystopian crud" because they are new technology.

I'm sure that people seeing a hypothetical AT&T I/O 1980 would think that automated telephone attendants were the end to human communication, commercial internet would ruin good letter writing and therefore society, and fax machines would be used as another form of government control. Imaginary dystopian futures never happen, but people keep imagining new ones.

⬐ None

None

⬐ None

None

⬐ anderber

I'm not sure I agree 100% with what you said, but I get the sentiment. If someone is worried about a dystopian crud, why would they purchase a Google Home or use the Google Assistant? Also, the parenting values I thought was nice, and optional to turn it on.

⬐ lainga

If the people are worried about police surveillance, why would they purchase CCTV cameras and data centres?

⬐ djrogers

CCTV cameras and the data they collect are generally under the control of the people who purchase and run them (short of a court order) - Google-connected bits, not so much.

⬐ ocdtrekkie

I can't escape it when Google's implementing their AI to call and message other people. They just demonstrated Google's AI calling and trying to converse with unsuspecting users who aren't even notified they're talking to Google.

⬐ anderber

What's the issue/reason you see with Google Assistant calling to make a reservation for a business? Why do you need to avoid it?

⬐ e1ven

Maybe it'll work out well.. Maybe not.

But right now, people being called have an expectation of talking to a person, and this shifts that expectation without the business opting-in to the change.

That's going to be frustrating to some businesses, and there may well be legal issues in some areas.

⬐ bendecoste

Not saying I agree with the parent but it's not like Google is going to call it quits here and not try to improve the tech to use in more situations.

⬐ macrael

It's creepy because it's pretending to be people. Phone trees suck, and now a phone tree is calling you and we don't expect phone trees to call us. Now you have to wonder when you get a call wether you are talking to a person or a robot. At the minimum they should announce themselves.

⬐ nfoz

Also, they recorded this call and are using it for their internal dataset and also sharing it publicly with hundreds of thousands of people. Should the recipient be informed of this data collection, much less have a say in it?

⬐ ocdtrekkie

Either the calls given in the example were staged, permission was solicited outside the bounds of this call, or they did it outside of California's jurisdiction. Recording calls without both parties' consent in California is illegal.

⬐ baumandm

If you can't tell, does it matter?

The reason people don't like phone trees is not because they are robots. Rather, because they often make it difficult to get the help you need or reach the right department. If a live human read the script of a phone tree to you it would be equally (more?) infuriating.

If I called a companies number and got a conversational bot masquerading as a person instead of a phone tree, that would be a dramatic improvement.

⬐ candiodari

The problem with phone trees is not that they suck at getting you to the right department, it's that the options required mostly aren't there. It's that it makes it soooo cheap for the company implementing the phone tree to waste your time just to improve the metrics of some low-level idiot manager at the company.

Voice will make this a LOT worse, not better.

⬐ tlb

Yes. Because if I'm having a bad day and I'm rude to the caller (not knowing it's a machine), I will feel guilt afterwards. The guilt will be unnecessary, because no human was offended.

That may be a trivial case, but more generally our social contract is that we owe a duty of politeness and consideration to fellow humans. When you're interacting with another human, a big part of your brain is actively engaged in following social norms. By masquerading as human, machines co-opt this part of your brain and extract a duty that they don't deserve.

I think every machine caller should be required to identify itself as such. Restaurants will still be happy to take reservations from them. If one calls me to sell me something, I can tell it where to go without any guilt.

⬐ djrogers

Serious question here - legally, what's the difference between a google-bot calling my place of business and a robocall, which is illegal in the US and California?

⬐ dragonwriter

What is illegal, IIRC, is a pre-recorded call, outside of certain prescribed bounds.

Duplex calls do not appear to be pre-recorded. It's unclear whether those regulating pre-recorded calls would have treated interactive automated transactional calls the same had they anticipated them, so I don't think it makes sense to blindly treat them as equivalent without new legislation.

Whether they should, on the merits, be regulated, and if so whether with the same or a different model, is perhaps an interesting debate.

⬐ killjoywashere

Agency and the intent of the beneficiary strike me as relevant issues. However, one can readily imagine someone using a system like this to hold hostage the schedules of many, many service businesses. Think OpenTable with no intervening agreement between the restraunt and the intermediary.

⬐ None

None

⬐ lsb

The google-bot calling your place of business is tantamount to someone's butler calling your place of business. Often time, people prefer talking to professional appointment-bookers.

Sometimes, not. When HR departments are recruiting, they avoid talking to agencies, with language like "principals only".

⬐ nfoz

Ah, so I suppose there are two dimensions: 1) The caller has a business-reason for initiating the call; if the receiver doesn't want that sort of call, it's a problem (spam calls). 2) The caller is a robot, and the receiver might not want to accept calls from robots, and now they might not even know it's a robot with which to have an informed opinion.

I wonder which one the robo-call laws are trying to cover, if not both. Google's certainly doing both, what's new is that the receiver is being misled about (2) and possibly even (1).

⬐ vthallam

Google is winning with TPU's. Most other competitors even with great machine learning talent would not be able to match the speed that Google does things with because of the TPU's.

⬐ raverbashing

And optimized sw for those TPUs

⬐ nikivi

The Wavenet assisted new voice got me thinking.

I need an Alan Watts voiced Google assistant / Siri.

⬐ pjmlp

I can hardly believe that people cannot control themselves to the point that adding features to disable the mobile use is a sell point.

⬐ slow_donkey

If you're referring to 'Shush' I'm excited for that purely for convenience sake. I see a lot of my friends use that same action (flipping phone over on table) when we're holding a conversation/eating/etc.

⬐ pjmlp

I just put it on silence and leave it on the pocket.

⬐ gd2

Theme that I see is branding that besides engineering smarts, Google is likable, thoughtful, someone you'd want as your friend. Leading with the hamburger emoji to move the image of Google as more fun / regular human. and IMHO think its well done.

⬐ ikeyany

I get the sense that Google isn't completely sincere with its Digital Wellbeing campaign. Rather than tally up your usage statistics on a dashboard, why not confront the addictive technologies (social media apps, clickbait/sensationalist styles of speaking) head on? Surely they have the data on what addicts people on a deeper level.

One example that comes to mind is "Hey, you've used social media X times today. Here are some websites you may like that would increase your attention span."

⬐ taurath

With the google news revamp (to preferring videos to everything, presumably to increase ad revenue) it seems like they want to move the web to more “bad” engagement.

⬐ maherbeg

The cynic in me thinks "What a great way to kill eyeballs to your competitors for advertising dollars"

⬐ mcondit

They actually seemed to have touched on something like that briefly, saying they'd come back to it in their android segment. Still not sure if it's sincere but if it is, it's a good set of steps.

⬐ djrogers

> This video is restricted. Try signing in with a Google Apps account.

Ughh.. I really hate this - why does Google put videos like this behind sign-in walls? Now I gotta go find my g-apps sign-in info, and rmember to purge it from my browser when I'm done again. GRRRR

⬐ magicalist

It opens in incognito/private browsing for me just fine

⬐ btutal

I wanted to download it to watch it offline during long bus trip but none of the downloaders seem to be working for previously live streamed videos on YouTube. Any other working method?

⬐ kshatrea

I am not sure what all this is supposed to enable us as customers or users to do. What I can see is a bunch of ways to avoid human contact including no longer calling people like a real human or not even choosing my own words in an email. But from the perspective of someone who dealt with Google's awful customer service, maybe they could use this in that regard to actually have a voice on the phone even if it isn't human. Automation is a great thing until you see the human cost. At some point Google is going to have to use its vast capacities to not give people more time, instead give them more opportunities. For e.g. the DeepMind paper showing how to save energy - that was cool.

⬐ tyrankh

> What I can see is a bunch of ways to avoid human contact including no longer calling people like a real human or not even choosing my own words in an email.

Auto-complete is hardly "avoiding human contact", is it?

Regardless, these features are all optional - if you feel as though you're avoiding human interaction by using auto-complete, why not just... not use it?

⬐ reaperducer

>Auto-complete is hardly "avoiding human contact", is it?

It is. It removes the color and nuance from typed communication. There's a lot of metadata in typos.

⬐ tyrankh

> It is. It removes the color and nuance from typed communication. There's a lot of metadata in typos.

Oh puh-lease. Are you seriously trying to argue that TYPOS are the anchor upon which genuine human interaction hinges on?

I mean, wow. If you're trolling that's amazing.

⬐ reaperducer

>Are you seriously trying to argue that TYPOS are the anchor upon which genuine human interaction hinges on?

I said no such thing. I stated, correctly, that typos convey information. Removing typos removes information.

⬐ jakebasile

You conveyed more emotional information in the misspelled "puh-lease" than you would've with the correctly spelled "please".

⬐ arijun

the actual keynote starts at 1:20:17. Can someone get the direct link for people coming later? I can't seem to do it on mobile. And can we perhaps change the link in the post?

⬐ Jaepa

This has to my favorite part: https://youtu.be/ogfYd705cRs?t=2536

⬐ markdog12

Stream was using AV1 codec for me, on both Chrome and FF nightly. Unfortunately quality was terrible, for whatever reason

⬐ chmln

> The all-new Google News provides you a variety of diverse perspectives on any given topic to help you get the in-depth, balanced picture

Proceeds to show stories by: CNN, ABC, Huffington Post, Buzzfeed, Vox, and Politico. Yeah, really diverse perspectives. Its not as if they all hold nearly the same point of view on any issue.

⬐ guiomie

What would you have in there?

⬐ chmln

For the sake of diversity they could've at least added some token libertarian or conservative outlets.

Also contrast the complete lack of neutral (at least relatively), established outlets like Associated Press and Reuters with the presence of...BuzzFeed.

⬐ ihsw2

BuzzFeed and HuffPost are on par with The Daily Wire and Breitbart News -- blatant propaganda ad nauseum.

⬐ reaperducer

It would be nice if SV realized there are tens of thousands of these things called LOCAL news sources.

Every piece of news-related tech we get these days just regurgitates CNN, BBC, etc... How about sending me local content from the local newspapers to start -- you know, things that affect people every day.

⬐ mtgx

Looks like the Google I/O keynote has turned into the Google AI keynote.

⬐ ddalex

All your I/O now belong to our AI

⬐ sidcool

No new hardware this year? The rest was quite impressive.

⬐ mrep

I cannot seem to find the time but they did announce a new TPU and it is now even liquid cooled.

Edit: here is an article on it and they are apparently 8x more powerful than the previous generation: https://www.cnbc.com/2018/05/08/google-launches-tpu-3-point-...

⬐ sidcool

I meant consumer electronics hardware, new pixels or chromebooks etc. But that's for a separate event, apparently.

⬐ 52-6F-62

At IO Extended Waterloo. They said they tend to reserve the hardware launches for the fall, and gravitate toward software launches in the spring when asked the same question

⬐ None

None

⬐ sunseb

I love technology. But I don't like what I see anymore. It's scary now. We are heading towards a dystopian future. My smartphone is now a black box in control of my life. I think I will just install a minimal Linux distribution and forget about these big corporations like Google, Apple and so.

⬐ gulperxcx

What brand of tin foil do you use to make your hats?

>My smartphone is now a black box in control of my life.

Only if you let it control your life.

⬐ tyrankh

> I love technology. But I don't like what I see anymore. It's scary now.

Welcome to almost every century in human history heh.

⬐ rubicon33

I closed it after 20 minutes or so.

I gotta hope that if I were at the helm of a multi billion dollar company that had been operating efficiently for years, I would take MUCH bigger risks than Google.

The projects in this keynote were incredibly boring, and some of them borderline useless. Leave that stuff for smaller companies to tackle, Google. Take on big problems.

⬐ urda

Google doesn't like hard wins anymore, they haven't in a long time. If it isn't easy then it's not worth doing for them. Sometimes even when it is easy, they kill the product off within two years.

⬐ eitally

Google has gotten leery of talking about the actual moonshots because of the extremely negative press/dialogue that results.

⬐ mbrumlow

Well I was agreeing, but now the self driving cars have been shown...

⬐ xedeon

Personally, it seems that during the past few years, Google has become so bloated and inefficient that innovation has slowed to a crawl.

It's to the point that they are now essentially just like Microsoft during the Steve Ballmer days. They need a visionary leader that can lead and inspire the hundreds of smart people they have.

Not to mention the fact how everyone always jokes about how long a product will last after being announced before they kill it off. Their strategy seems to be scattershot, rather than having a sense of direction. I'm curious to see how their internal roadmap looks like.

⬐ jacksmith21006

Inefficient and bloated? They are putting out a ton of stuff. Have YouTube TV and an excellent product but just one thing.

They appear to be doing an entirely new OS and UI. Flutter is pretty impressive tech.

Beyond Corp is pretty revolutionary and can see it being the norm.

But so many more. The TPU 3.0. Google Home Max is a great product as well as Google WiFi.

K8s is on fire and growing like crazy. They have cars driving around without a safety driver.

So many other things. What in the world are you talking about?

Did you mean some other company?

⬐ jacksmith21006

I am curious who has done more than Google in the last year?

Thanks in advance!

⬐ datalist

Still watching it but I second your sentiment. To be honest after the conferences in the past three, four years I did not have the highest expectations in the first place, but so far that was - again - amazingly underwhelming.

As you said, Google should announce game changers and not getting woos for an application that reorients a scanned sheet of paper.

⬐ jacksmith21006

Disagree. So many just incredible things coming from Google. The internet is a buzz on the Duplex demo but just one thing.

⬐ TY

I'm sorry but I disagree. What I see here is a massive multibillion dollar company undergoing a tremendous technological transformation in front of our eyes.

Machine Learning is being embedded into most of their products to make those products increasingly capable and useful that is not easy for other companies to replicate.

Those products are becoming personal and learn about you to the point that they will know more about you than you... I realize that this can lead to a scary future, but lots of useful things do.

These capabilities might look underwhelming but they are simply mind blowing given type of problems being solved.

More importantly, this technology and infrastructure are being put in the hands of developers outside of Google, so more fascinating things are yet to come.

⬐ TY

The call with Chinese restaurant was simply jaw dropping. Sundar says that it was real and I have no grounds not to believe him.

This is as close of a demo of technology passing the Turing test as I have ever seen. Sure, it's not a fully free form conversation on any subject, but incredibly impressive nevertheless.

Oh yes, and the voice synthesis is just simply amazing. I don't think I'd be able to tell that I was talking to a bot either.

So far, it's been the most impressive Google I/O keynote that I've ever seen and it's not over yet.

⬐ fragilistical

If you'd like to read up on Google's voice synthesis system do read up on wavenet: https://deepmind.com/blog/wavenet-generative-model-raw-audio... from deepmind.

⬐ VikingCoder

Jeez, this thing is just about ready to Dungeon Master.

Assistant, "Okay, I need you to roll initiative. What did you get?"

Me: "27."

Assistant, "Umm... You're a first level Barbarian. How did you roll 27 on Initiative?"

Me: "I mean 17."

Assistant, "Uh huh."

⬐ MarkMc

I wonder how long it will be before there is a call to book a restaurant where both sides are AI assistants

⬐ herewulf

AI assistant calls could start with a modem style handshake to determine whether they can conduct the transaction much more quickly with an Internet API. I'm sure it could be much shorter these days, but hearing the same noisy handshake today would be amusing.

⬐ chipperyman573

Bringing us alllllll the way back to dial-up

⬐ tyrankh

Yeah, that was pretty much jaw dropping. I'm excited to see it in action.

⬐ VikingCoder

I missed that. Can someone link to that moment?

⬐ timeu

https://ai.googleblog.com/2018/05/duplex-ai-system-for-natur...

here are some more examples

⬐ zbobet2012

https://www.youtube.com/watch?v=-RHG5DFAjp8

⬐ None

None

⬐ dperfect

https://youtu.be/ogfYd705cRs?t=1h57m57s

⬐ nfoz

I'm really trying not to be another "dystopian" commentor, but I can't help but feel sad for the restaurant. The call was awkward. Restaurants are busy. The AI would pause too long and leave her guessing, and respond with the wrong type of emotion (e.g. a dejected sounding "oh, gotcha... thanks..." after she had answered his question with the positive "you can come for four people, ok?". I can't help but think this sort of "corporate lying" (yes it is a lie to robo-call and pretend to be a live human voice) has an external cost of frustration and emotional drainage to be borne by society's ordinary people.

⬐ slow_donkey

Maybe I'm a robot but i feel like I would have answered similarly to Duplex. I expected to be able to make a reservation but got told I have to walk-in instead.

I'm actually quite surprised it asked for the wait time afterwards, I'm not sure I'd remember do such a thing.