Toyota to Invest $1B in AI and Robotics R&D | Hacker News Reader

Toyota to Invest $1B in AI and Robotics R&D

spectrum.ieee.org

101 pointsmcspecter35 comments

Back in the 80s and 90s there was a lot of media and memes (as in ideas, not the funny images) that Japan was the land of robots. Even a Japanese version of the human evolution chart where the last stage was a mecha.

But it seems to me that it has lost the race. And I don't know why.

South Korea robotics is as advanced as Japan's. And Google just bought all the best people/companies that were out there.

The same thing happened with smartphones. Japan had the best smartphones a decade or two ago. Blazing internet speed. TV on your mobile. Connection to the washing machine (IoT anyone?).

Yet it lost it to iPhone and even Android. Sony is unable to compete with Samsung.

This is purely my idea on things. I could be completely wrong as I don't have numbers from some of my conceptions.

Would also love to known the opinion of someone who had lived in Japan on why they can't; well, scale outside of Japan (english language bottleneck?). @patio11 ???

I was there on vacation not too long ago - it is a very interesting place. Most of the "weird" things we associate with Japan as outsiders are usually very practical and sensible solutions to local problems.

I think it's definitely true that Japan is a leading nation in many ways - not just technologically, but also socially and both in the good and bad aspects. But because Japan is a fairly unique place the local solutions don't translate 1:1 to the outside world, so you have to and disentangle their solutions with their particular environment to get a true sense of what's coming.

Smartphones are a good example - everything they do today (and more, TV for example) existed in Japan years before, but it was in local format. I take is that it was easier for Apple and Google to implement similar technologies rather than for Japanese companies to translate their technologies to a different mindset.

> Japan is a leading nation in many ways - not just technologically, but also socially and both in the good and bad aspects

I lived in Japan. Most expats (granted, not engineers) are in agreement that Japan feels like 2050 for technology but 1950 socially.

There is zero PC-sensibility, zilch, doesn't exist in their culture. It's also a man's society, women are expected to give up their careers after marriage. Race is deeply ingrained in their world view in a way that would anger PC-conscious Americans.

There is also zero culture of accepting integration from people who are not Japanese by blood. Go to your Japanese girlfriend's family's get together and listen to them mention how the Japanese are a superior race to Westerners (maybe that was a fluke but I doubt it from swapping stories). Mention that it's relatively accepted that the Japanese descended from the Koreans and watch how people react.

Listen to seemingly progressive 20-something at bars talk about how Chinese people are unclean.

After all, they believed their political leader was divine until the end of WWII and they built an empire based on their hierarchy of races up until the end of WWII, much like NSDAP's 'project' but without receiving the same kind of international backlash (a wave of pacifism yes, but backlash for the racial elements not so much).

They have a culture of friendliness towards foreigners which is awesome, unlike say China, so you don't fully see this side of Japan until you live there.

I really do love Japan, I have zero ill will towards the country or it's people, but it's simply wrong to claim it's socially progressive in almost any way by Western standards.

> Race is deeply ingrained in their world view in a way that would anger PC-conscious Americans.

What's worse is that it's not even only about race. There's Japanese prejudice towards occupations (anything related to death, meat prep, and so on) and even about places where people are born and where they live

"Progressive" is such a stupid word. Who gets to decide what is, and isn't, progress?

> Most of the "weird" things we associate with Japan as outsiders are usually very practical and sensible solutions to local problems.

Care to elaborate?

I don't live in Japan but, on the surface, the problem seems self-evident -- Japanese innovation is dead. The heads of the conglomerates (Sony, Mitsubishi, etc) are all old and Japanese, and anyone that is not old and Japanese is regarded as a second-class citizen.

On top of that they have strong ties and reliance with the government -- this guarantees their survival, so they have no interest in maintaining international growth.

And, if it's any consolation, Sony's semiconductor business is quite successful, so much so that it's been spun off[1].

[1] http://www.theverge.com/2015/10/6/9460463/sony-image-sensor-...

The common factor is lack of respect for software. They did very well when focused on just hardware with simple software. For example the hardware part of robotics and very advanced features phones. But the next stage requires advanced software such as AI and the iPhone. This is an issue not just in Japan, but Asia in general. Their history had a lot of hardware success, but no major software success. In the US there is both Intel and Microsoft, major success in both hardware and software.

This is actually a competitive advantage of software engineering in the US. We have a history and culture of respecting software. This is passed down generation to generation and allowed more success like Google and Facebook.

In Asia, software is a low status job with little pay and status.

More to the point, to have success in software development you need:

- Independently minded people (incompatible with Japanese culture)

- Flatter hierarchy (kind of relates to the previous point, and incompatible with Japanese culture)

For simpler software, this is easier to overcome (Super Mario levels of software complexity, let's say)

Steve Jobs didn't come up with the iPhone. He let people come up with great ideas then filtered them.

In Japan, the structure goes: "What does the boss want?". And if the boss doesn't know, they will go with the most conservative choice possible.

This set up is never going to win prizes.

Can you explain why a flatter hierarchy is necessary for software development?

My background is in hardware engineering, and this strikes me as surprising (although something I haven't ever given much thought to). What about software makes it incompatible with hierarchy? Don't a lot of traditional software companies have a sort of hierarchy?

Don't confuse lack of hierarchy with flatter hierarchy.

> Don't a lot of traditional software companies have a sort of hierarchy?

Yes, see what kind of crap they produce.

Of course, not all software companies, but the ones that really innovate can't survive certain environments

> What about software makes it incompatible with hierarchy?

The sheer level of independence that every part needs.

In a hardware project you pick your parts (you have some requirements, some limitations in size, power, vendor, etc so it might be a choice between a handful of parts), then put it on a PCB, solve placing, routing, EM compatibility, etc. It's complicated, but you don't have too many people doing that. And more importantly, your boss knows more or less about everything and if he doesn't believe it there are hard constraints that serve as guiding (power consumption, EM emission, performance, etc)

Good? Now, you want to build the software for the iPhone. Or for the Macintosh, same thing

Development is separated in many areas (design, base system, user APIs, drivers, etc)

And the overlap between areas is significant. You need information flowing freely so you can come up with the best solutions (this is not 'design by committee')

You also need independence and not justify every tiny decision (and why didn't you go with the 'safe' solution).

This is the complete opposite of what happens in Japanese work culture

http://www.hierarchystructure.com/japanese-business-hierarch...

True and possibly also a lack of focus on product in general. Most Japanese products I know are STUFFED full of features too complicated to use (see: every Sony camera ever). Sometimes the focus should be on reducing the number of features to more comfortably insert yourself into the users workflow. This is probably true for almost every Japanese commercial robotics venture.

Japan has hardly lost the robot race. Why? Well japanese robotics companies held ~61% of the industrial robot marketshare in 2011[0]. Second, they're a big producer of actuators for robots, so they're in a good position to supply the robotics market. I would not discount korea though, they have a number of companies that produce actuators too.

Actuator companies are something to look out for, because robots are practically made of them. [0] http://www.statista.com/statistics/257177/global-industrial-...

Japan had a financial crash in 1989. when their real estate bubble popped. Japan never fully came back from that.[1] Before the crash, Sony had their Aibo robot, and Honda had their Asimo. Sony had a small humanoid toy-sized robot in prototype, but never shipped it. Asimo work stalled; it's still around, and there's been some progress, but not much.

Robotics in the US has been heavily financed by the military. It took about $120 million to get Boston Dynamics to Big Dog.

It's not clear at all what Google is doing. Google bought all those robotics companies, but nothing has come out yet. Boston Dynamics, which Google bought, stopped updating their web site two years ago. They've demoed the old Boston Dynamics Atlas running in woods. That's about it. Google also bought Schaft, the Tokyo University spinoff, which had a very nice humanoid robot. (All-electric with liquid cooling, unlike Boston Dynamics' Atlas, which is hydraulic and far too heavy.) Schaft hasn't been heard from since. The question is whether Google's robotics groups are doing something really good, or getting nowhere, or the smart people have moved into some part of Google that makes money.

[1] https://finance.yahoo.com/echarts?s=^N225+Interactive#{%22ra...

Japanese robotics tend to be preoccupied with an aging population, so they have elderly assistant robots, and robots which greet people at depatos, etc.

But that's the humanoid ones, not the great numbers of ones on the factory floors which assemble things for people.

They focus on things which affect their society.

Lack of diversity.

Thanks for writing this. If I remember correctly many smarty pants at that time were claiming that US government must impose on tariff on cheap electronic goods from Japan because Japan's industry was subsidized by government and eventually japan will take over entire electronic industry in USA.

All those things Japan did benefited USA and smart people could focus on things other than selling Televisions. We all progressed at much faster pace.

Today people show concern that China is manufacturing US Olympic uniforms and we must impose restrictions on Chinese imports.

Whenever a company writes a press release with a large sounding number in it my first question is "Is that a lot?". I have no idea how much it costs to run a robotics lab. As this is over the next five years, what would an average of $200m/year actually buy you? Is that a team of 50 very clever people working on cutting edge stuff, or is it 5 interns and a really expensive patent lawyer?

Headline numbers are designed to appeal to our "Crumbs, I wish I was a billionaire!" instincts, but without the context of what's being invested in and the domain knowledge of how much the industry typically spends, this could be awesome or it could be entirely non-newsworthy. I don't know.

A PhD student gets something like 2500 Euros before taxes per month where I did my PhD. So it costs the institute maybe 40-50k a a year to fund one PhD student's salary. Let's say 100k including office space and some hardware to play with. 200 Million pays for many PhD students.

Yes, but for graduating PhD students 175k-225k/year is roughly market (base) rate for people with proven understanding and ability in the most relevant areas (i.e. multiple publications in top conferences like ICML/NIPS/CVPR/ICCV/ICRA/RSS/etc).

Though, a billion dollars will still pay for quite a few of those.

I think that depends a lot on the location. I know a couple of people who do PhDs in Japan and they are a lot less well paid than German PhD students. I expect that this difference persists even after they graduate.

Very true. These are rates in the SF Bay Area and NYC. IDK what it's like elsewhere, though I've been told it's possible to get offers in this range in London too.

East Europe universities, particularly lithuanian (VU and KTU) pays about 300$ per month for PhD student.

IBM decided to invest the same number in Watson last year [0]. Not sure if that helps.

[0]: http://www.reuters.com/article/2014/01/09/us-ibm-watson-idUS...

Insofar as Watson (the brand) seems to be the next Websphere (the brand) it's really, really hard to tell what's going on there.

To give some context, a new automotive assembly plant could cost around a billion[1]; or an old plant could be modernized with a billion[2] -- It is well within the realm of a normal capital expenditure of an auto company.

Industrial Automation is more expensive than you may imagine. A robot arm may be 100k to 200k or even more (hard to cite as prices are never listed on the internet); you also have to have a controller (PLC) and all of the associated sensors and connections.

We already have heavy duty robots capable of high precision positioning; the area for expansion / improvement is in making these cheaper, and in "Internet of Things" style automation.

Modern factories are heavily automated in the body welding processes and paint application.

The only area of the plant unconquered by robotics is General Assembly, where all the fiddly bits are put on - wire harness, carpet pad, carpet, etc. Robots are famously bad at handling bendy things. [3]

1. http://www.wsj.com/articles/why-auto-makers-are-building-new... 2. http://www.wfaa.com/story/news/local/tarrant-county/2015/07/... 3. http://www.npr.org/sections/money/2015/05/19/407736307/robot...

Disclaimer - I used to work at Arlington Assembly, and still work for GM. Any opinions are my own only.

I guess you can disclaim speaking for your employer, but I don't get why you'd disclaim your credentials.

Suggest a better wording?

I suspect their heavy bet on hydrogen will sink them long before they see any returns from AI and Robotics.

So they can invest $1B in this but not fix the spaghetti code in their cars that causes unintended acceleration? Priorities!

Perhaps they're doing both?

Hope so

Actually inserting that code was the first anti-human action of Toyota's rudimentary Skynet-inspired AIs.

According to DOT the problem was floor mats and driver error, not code

http://www.caranddriver.com/features/its-all-your-fault-the-...

This is great to hear. I wonder if they're pumping some of this money into TTI-C as part of this initiative.

Reply on news.ycombinator.com