Author Topic: GOOGL - Google  (Read 473679 times)

Jurgis

  • Hero Member
  • *****
  • Posts: 4362
    • Porfolio
Re: GOOGL - Google
« Reply #1960 on: February 13, 2019, 06:55:45 AM »
@WayWardCloud:

I think Waymo currently is encountering something that is intersection of safety and performance and irrational human decision making. E.g. the situations like merging into freeway or a congested lane. Self-driving cars can handle that mostly. But there are situations when this is not easy/dangerous. What humans do in such situations, they just do it (muscle in). But from self-driving car perspective such muscling in would be unsafe, so the code is not tuned to be aggressive. So the self driving car just gives up or at best stops which makes the merging even tougher. And this annoys all the humans behind it and possibly precipitates humans muscling on the self-driving car even more, which it then considers even unsafer situation to deal with.

I'm not sure how this is going to be solved. It might be possible to make the car more aggressive in certain situations, but that would likely make it less safe. Humans just don't think that some maneuvers are 100-1000x riskier than others. They just do them anyway.


There are other issues to be resolved like weather (rain, snow) and poor road markings. I think these are easier to solve technologically, since they are pretty much science/engineering problems rather than world/humans/safety intersection problems.

Just to mention, there are also:
 - 'unexpected situations' problems, i.e. where something happens that is not adequately recognized and handled by car.
 - humans trolling the car problems, e.g. someone just walking in front of a car in a street where self driving car cannot go around the human.
 - quality-of-life problems: I and (most) Uber/Lyft drivers drive risky but save time like merging at last moment into congested exit lane. Would humans accept safer car that takes 40 minutes to get somewhere when risky human does it in 20?

For these reasons, I am more in a Tesla'ish-give-self-driving-but-keep-steering-wheel-and-controls-for-human camp rather than Waymo-full-auto-no-steering-wheel-and-controls camp. Although I understand Waymo's argument against the "Tesla'ish" camp.  8)
"Before you can be rich, you must be poor." - Nef Anyo
--------------------------------------------------------------------
"American History X", "Milk", "The Insider", "Dirty Money", "LBJ"


Liberty

  • Lifetime Member
  • Hero Member
  • *****
  • Posts: 10994
  • twitter.com/libertyRPF
    • twitter.com/libertyRPF
Re: GOOGL - Google
« Reply #1961 on: February 13, 2019, 10:35:30 AM »
https://www.blog.google/inside-google/company-announcements/investing-13-billion-2019/

Sundar post about their 2019 investment plans:

Quote
Today we’re announcing over $13 billion in investments throughout 2019 in data centers and offices across the U.S., with major expansions in 14 states
"Most haystacks don't even have a needle." |  I'm on Twitter  | This podcast episode is a must-listen

LightWhale

  • Full Member
  • ***
  • Posts: 127
Re: GOOGL - Google
« Reply #1962 on: February 14, 2019, 01:07:42 AM »
https://www.cnbc.com/2019/02/12/google-cloud-ceo-thomas-kurian-google-will-compete-more-aggressively.html

Thomas Kurian’s first public comments as Google Cloud CEO..
His reputation in the industry is somewhat mixed - would be interesting to see what happens

What kind of feedback are you guys hearing about GCP?

Various SV guys involved with cloud initiatives seem to repeat the same concern, that GCP is not doing so well vs AWS and Azure. Specifically, the technology is progressive and great for software people, but complicated for less sophisticated users and migration is tough because GCP is less costumer oriented & Google engineers too spoiled to do nitty-gritty work. 

Jurgis

  • Hero Member
  • *****
  • Posts: 4362
    • Porfolio
Re: GOOGL - Google
« Reply #1963 on: February 14, 2019, 07:51:42 AM »
Anecdotally, I have tried Google Cloud recently and it's full of obscure magic incantations. OTOH, I think that's mostly true about any cloud or even data center setup. Although TBH I haven't used AWS recently and Azure at all. 8) Likely once you work for some time you know the magic incantations by heart or at least have them written in some personal doc for copy/click/paste.
"Before you can be rich, you must be poor." - Nef Anyo
--------------------------------------------------------------------
"American History X", "Milk", "The Insider", "Dirty Money", "LBJ"

walkie518

  • Sr. Member
  • ****
  • Posts: 384
Re: GOOGL - Google
« Reply #1964 on: February 14, 2019, 08:59:19 AM »
Anecdotally, I have tried Google Cloud recently and it's full of obscure magic incantations. OTOH, I think that's mostly true about any cloud or even data center setup. Although TBH I haven't used AWS recently and Azure at all. 8) Likely once you work for some time you know the magic incantations by heart or at least have them written in some personal doc for copy/click/paste.
I've used Google Cloud.  I've played around w/AWS.  I have not touched Azure. 

Google Cloud is fine. It's a little clumsy as far as implementation goes.  Credentials and keys could be done in a way that's a little smarter and while the dashboard doesn't look cluttered, it's not always easy to find exactly what you're looking for.  Looking up documentation leaves a lot to be desired.  There are docs for very simplistic use cases, but they don't spell everything out as one might expect.  The feature set is richer than expected but doesn't touch AWS.   

LightWhale

  • Full Member
  • ***
  • Posts: 127
Re: GOOGL - Google
« Reply #1965 on: February 14, 2019, 09:12:14 PM »
Anecdotally, I have tried Google Cloud recently and it's full of obscure magic incantations. OTOH, I think that's mostly true about any cloud or even data center setup. Although TBH I haven't used AWS recently and Azure at all. 8) Likely once you work for some time you know the magic incantations by heart or at least have them written in some personal doc for copy/click/paste.
I've used Google Cloud.  I've played around w/AWS.  I have not touched Azure. 

Google Cloud is fine. It's a little clumsy as far as implementation goes.  Credentials and keys could be done in a way that's a little smarter and while the dashboard doesn't look cluttered, it's not always easy to find exactly what you're looking for.  Looking up documentation leaves a lot to be desired.  There are docs for very simplistic use cases, but they don't spell everything out as one might expect.  The feature set is richer than expected but doesn't touch AWS.   

Thanks, good to have some hands-on reviews
« Last Edit: February 14, 2019, 10:00:51 PM by LightWhale »

meiroy

  • Lifetime Member
  • Hero Member
  • *****
  • Posts: 733
Re: GOOGL - Google
« Reply #1966 on: February 16, 2019, 03:46:08 PM »

VAL9000

  • Sr. Member
  • ****
  • Posts: 419
Re: GOOGL - Google
« Reply #1967 on: February 17, 2019, 04:28:59 AM »
Here's a guess on where Waymo is.

First, some recent data points:
 - Waymo has entered into a number of partnerships with a lot of car manufacturers (incl. Nissan/Renault, Jaguar, Chrysler)
 - Waymo recently announced a manufacturing facility in Michigan
 - Safety drivers are still required for even the most forgiving environments
 - Drivers are clamoring for self-driving technology, to the point of giving their lives to act as beta testers

With this in mind, my guess is that the self-driving utopia that Waymo envisions as its mission is a long way off, but Waymo's L4 capability is commercially viable today.  The timeline for utopia is probably decades out, but here's what can be done today:
 - Integrate Waymo/Android/Google technology into passenger vehicles as an option.  Likely Waymo will offer their technology with no expectation to profit on the hardware, instead opting for a service or other platform-driven revenue stream.
 - Deploy L4 tech to owners of this package, essentially making the passenger the safety driver a la Tesla Autopilot
 - Continuously upgrade L4 tech OTA until L5 autonomy is attained, also like Tesla, but with more advanced hardware, deeper data set and better AI

This approach makes sense to me because:
 - Waymo can establish market share sooner if the outcome is that consumers cling to vehicle ownership
 - Waymo can begin collecting substantially more data on driving habits and enhanced/augmented maps
 - There is a ton of value in L4 tech for anybody who even commutes a little every day, and most auto manufacturers are in no position to capture that value

As a consumer my feelings are:
 - I would easily pay $5k for the hardware and $25-50 per month if it meant I could text and drive most of the way to the office and back. 
 - I wouldn't trust Tesla's approach to L4 automation as it stands today.  I'd look at GM / Cruise closely but would still prefer Waymo.
 - I'd have a hard time accepting the hideous LIDAR and associated Waymo hardware package as it stands.

As Waymo my main fear is that on day 1 my competitors are going to buy one of these cars and reverse engineer my hardware, which is probably much harder to do when every vehicle is owned by Waymo and loaded to the gills with vehicle tracking software.  Harder... unless you hire away one of my product leaders and have them divulge the company's secrets (heh).

I could be completely wrong here, but I do think there's a current market opportunity that meets Waymo's ambitions halfway while also accelerating its development.  Personally I'd love for the opportunity to "vote with my wallet" on this one.

Jurgis

  • Hero Member
  • *****
  • Posts: 4362
    • Porfolio
Re: GOOGL - Google
« Reply #1968 on: February 17, 2019, 08:53:42 AM »
Here's a guess on where Waymo is.

First, some recent data points:
 - Waymo has entered into a number of partnerships with a lot of car manufacturers (incl. Nissan/Renault, Jaguar, Chrysler)
 - Waymo recently announced a manufacturing facility in Michigan
 - Safety drivers are still required for even the most forgiving environments
 - Drivers are clamoring for self-driving technology, to the point of giving their lives to act as beta testers

With this in mind, my guess is that the self-driving utopia that Waymo envisions as its mission is a long way off, but Waymo's L4 capability is commercially viable today.  The timeline for utopia is probably decades out, but here's what can be done today:
 - Integrate Waymo/Android/Google technology into passenger vehicles as an option.  Likely Waymo will offer their technology with no expectation to profit on the hardware, instead opting for a service or other platform-driven revenue stream.
 - Deploy L4 tech to owners of this package, essentially making the passenger the safety driver a la Tesla Autopilot
 - Continuously upgrade L4 tech OTA until L5 autonomy is attained, also like Tesla, but with more advanced hardware, deeper data set and better AI

This approach makes sense to me because:
 - Waymo can establish market share sooner if the outcome is that consumers cling to vehicle ownership
 - Waymo can begin collecting substantially more data on driving habits and enhanced/augmented maps
 - There is a ton of value in L4 tech for anybody who even commutes a little every day, and most auto manufacturers are in no position to capture that value

As a consumer my feelings are:
 - I would easily pay $5k for the hardware and $25-50 per month if it meant I could text and drive most of the way to the office and back. 
 - I wouldn't trust Tesla's approach to L4 automation as it stands today.  I'd look at GM / Cruise closely but would still prefer Waymo.
 - I'd have a hard time accepting the hideous LIDAR and associated Waymo hardware package as it stands.

As Waymo my main fear is that on day 1 my competitors are going to buy one of these cars and reverse engineer my hardware, which is probably much harder to do when every vehicle is owned by Waymo and loaded to the gills with vehicle tracking software.  Harder... unless you hire away one of my product leaders and have them divulge the company's secrets (heh).

I could be completely wrong here, but I do think there's a current market opportunity that meets Waymo's ambitions halfway while also accelerating its development.  Personally I'd love for the opportunity to "vote with my wallet" on this one.

I think what you suggest is the way to go.

Unfortunately, Waymo has been somewhat adamant that Tesla-ish way "human-takes-over-when-self-driving-cannot-handle" is unsafe and that they want to go L5 with no steering wheel/controls/etc. for humans. I don't know if they will be smart enough to change that tune and release what you suggest: L4 package that can handle most situations.
"Before you can be rich, you must be poor." - Nef Anyo
--------------------------------------------------------------------
"American History X", "Milk", "The Insider", "Dirty Money", "LBJ"

obtuse_investor

  • Sr. Member
  • ****
  • Posts: 375
Re: GOOGL - Google
« Reply #1969 on: February 17, 2019, 03:46:42 PM »
Looking further out into the near future, I bet even if Waymo wins the battle, they could lose the war.

Imagine a world where Waymo succeeds in launching with l4 or l5 autonomous vehicles. What’s the incentive for intelligent competition to develop  l4 cars? Not much. As there are more and more l4 cars on the road, it would be easier for l3 cars to exist. So why not focus on those cheaper l3 cars?

It’s almost like herd immunity.

Waymo’s investment could be excellent for the world and humanity, but not for the alphabet shareholders.

I assume Waymo and alphabet leadership know of this. And they are planning to make Waymo into a different kind of business.
Value Investor who manages his personal portfolio with a 25-45 year time horizon | @obtuse_investor