Hacker News Reader

Comment by eric4smith | Hacker News Reader

These are one of the few times I can say that I don't care if Boston Dynamics are making money as yet.

Boston Dynamics are one of the few real serious players in the robotics space worldwide doing deep R&D.

Hyundai are basically paying for the BD Labs research. And they are well positioned to reap the products and techniques from the lab with their own manufacturing and engineering prowess and experience over the next decade.

And it won't be walking dogs.

It will be things like a little flexi-joint in a car that you buy without knowing (or caring) it's in there. Or some industrial machine that Hyundai builds that gives them a 5% competitive edge in the manufacturing of it that 99.9999% of us never hear about.

But for that it's worth it.

Very happy to see people put good money into this Lab.

View on HN

The key competence of BD is locomotion on legs. There are plenty of other companies that make stationary industrial robots, and a few who make wheeled robots.

But there are still a number of areas where wheeled robots are at a disadvantage: inspection of complex industrial installations (like complex pipework and other hard-to-reach areas), firefighting and rescue operations, well, even delivery in places with steps and stairways.

Korea is more technophilic than USA, and Hyundai is well-connected. They may find a more ready market for such products.

There are plenty of patents along the way to get to locomotion on legs that have uses that don't necessarily need to be applied in that manner. Hyundai gets all of those too.

I imagine Hyundai Motors has loose informal connections and regular collaborative efforts with other former Hyundai Group companies, including Hyundai Heavy Industries and Hyundai Robotics, so hopefully there'll be some smooth expertise sharing there.

https://en.wikipedia.org/wiki/Hyundai_Group

https://en.wikipedia.org/wiki/Hyundai_Heavy_Industries

https://www.hyundai-robotics.com/english/

Hyundai is one of South Korea's market-dominating cartels, the chaebol: https://en.wikipedia.org/wiki/Chaebol

"regular collaborative efforts" is an understatement.

Robots that build cars seem most likely? Not sure how many staff your average car factory has, but due to sheer product complexity it's still more than other industries like food manufacturing.

Maybe we'll see vast car factories, churning out thousands of cars per day, with only a half dozen staff in years to come?

Robots already build cars, and supply of new cars is often out-stripped by demand. It's not really new ground.

The limitations in car production are always the component supply chain, not the actual manufacture.

Google image search "car factory" and it's super obviously it's quite labour intensive. I'm not sure why an investment in robots has to be "new ground".

The Google Image search isn't reflective of an actual car factory.. There are of course hundreds of workers there but final assembly facilities produce something like 1 car/minute. They're heavily reliant on robots - most of the parts move around the floor on autonomous transports, the stamping is all robotic, all of the frame welding is robotic, paint is of course robotic.

GM is of course tied in with the UAW, so they're going to be as favorable to labor as possible in their promo videos, but take a look at this video of one of their production facilities, this is reflective of the amount of manual input into modern automotive production:

https://www.youtube.com/watch?v=-Ow7gmZTIpo

I've toured this plant. They don't show it in this video, but in the stamping and body areas of the plant, they don't even have the lights on at task levels - it's a dusk-level lighting until a human enters the space. This plant was built circa 2006, and is pretty amazing. Bring steel coils into one side of the factory, get a fully assembled car from the other side.

You're using the wrong terms. Add "advanced manufacturing" to the query to get some idea of the current state of things. I toured a new automotive plant over 10 years ago and up until the final configuration stage, the only manual labor was cabling and quality control. This included sheet metal unloading and delivery (from rail), stamping of the chassis, welding, paint, installation of doors, hood, suspension, etc, powertrain installation, and many others. From my estimate, 2/3 of the factory was very close to being "lights out." The entire welding and paint process was fully automated and required a line stop to enter the area.

You are right, modern car manufacturers still use humans in assembly lines.

But to see how much robots actually do nowadays, compare that to auto assembly from decades ago when robots were new, or further back when they did not exist

Yup! And really, the highly, highly automated processes in car manufacturing are in the stamping, body construction and painting phases. Almost no humans involved in those three major phases. Final Assembly is the most labor intensive, because it involves lots of fiddly little parts going into hard to reach places. Even then, it's usually robot-assisted tasks, so one person is installing a seat, where it would usually take two, or something like that.

Humans do the basic stuff that robots aren't any good at: pulling wires through, putting some pieces in an inconvenient place,etc. All the welding, painting and many others are fully automated.

Robots indeed build cars, but not from start to finish.

Especially customisation and snapping delicate plastic trim pieces in place is still done by hand.

Humans spend 15h to 60h of labour to build a car. After, humans worked on individual parts, shipping, loading etc.

My rule of thumb is that there is about 10% of manufacture cost that are direct or indirect cost connected to payroll.

If I may add another axis of reflexion that I consider fundamental for modern engineers to consider, food for machines for producing a car (44% natural gas when smithing things) is about 15 to 25% of the energy cost to use the car in its lifetime, according to the EU. Or about 2,400 euro of a 8L/100km car.

So in a world where energy gets more expensive, automation needs to definitely evolve, but not solely to be faster or replace more humans.

https://www.mdpi.com/1996-1073/13/9/2396/pdf

Even 60h seems low here. Do you only mean assembly? Tacking on parts manufacturing, sub-component assembly, and software is likely to increase that estimate by a factor of 2 to 5.

The numbers are coming from Toyota, Peugeot and Renault. You can take the financials - manufacturing payroll is its own category.

Maybe. Another possible variant on that idea could be micro-factories [1] that address delivery constraints and allow a factory to be located anywhere there is a small warehouse.

[1] - https://www.youtube.com/watch?v=Z_Qyor9Yc-s

Seems that would be a supply chain nightmare, plus it would multiply capital costs. I doubt it's worth it.

In a factory setting I would have thought existing robots would suffice.

Looking at the wider group, I would think there's lots of roles for robotics in construction. Delivery of heavy items across uneven terrain or up stairs - may be something like a robotic hod carrier.

They are pretty great at welding and painting, they not as good at attaching bumpers or trunk lids, and terrible at running wire looms through doors or installing headliners.

Also from my last visit at an assembly plant ~5 years ago, most plants are still on large conveyors vs. driving individual vehicles around the plant, and materials handling is still in the 90's with little stripe-following tugs slowly moving about the place.

That's a good assumption to make. They for sure see 800M+ value in it.

Great for Hyundai. Not such a good ROI for the taxpayers who kept BD afloat with decades of DARPA funding.

I think people are missing the point of this comment. The US invested heavily in BD, and now BD is being incorporated into a South Korean company rather than a US company.

I presume this is mostly because US manufacturing industry is relatively dead so capex for potential automation gains don't make sense for US firms.

The perception that US manufacturing is dead is mostly just perception, if you look at the actual numbers.

Before the big dip caused by covid, the value of US manufacturing output had never been higher. [1]

Even in inflation adjusted/real value terms, US manufacturing output is 50% higher today than it was 30 years ago.

My guess is that the perception is cause by a few factors:

- fewer people are employed in manufacturing than were in the past because of improved manufacturing technology/productivity

- other sectors of the US economy have grown at a faster rate than manufacturing (tech for example) which means manufacturing represents a smaller share of a bigger pie

- the US is no longer number 1 in the world as China surpassed the US about 10 years ago. But the US still manufactures more than Germany, Japan and South Korea combined, for example.

[1] https://www.nam.org/state-manufacturing-data/2020-united-sta...

> US manufacturing industry is relatively dead

Here's where we need to be spending our tax money. We should be buying overseas firms and onshoring manufacture.

The legal landscape in the U.S. may be harder to navigate, labor laws more strict, and salaries necessarily higher. This all makes it hard for U.S. manufacturing to compete on production of commodities.

Like many of the EU industries, viable U.S. industries tends to produce and export unique things with relatively small physical volume: complex chips, complex chemical compounds (like medicines), precision mechanics, high-performance agricultural equipment, aircraft, spacecraft, etc. Of course the U.S. does produce higher-volume things, too, from cars to foods to fuels, but these are more for domestic consumption, and often face competition from imported goods even domestically.

Not to mention land prices mean salaries need to be higher for the workers to afford rent/mortgages, facilities more expensive, higher property taxes, even ancillary purchases are more expensive.

What if we opened up immigration and built factories in middle America?

Provide path to citizenship and education opportunities for children of immigrants, but host all manufacturing domestically at rates comparable to overseas facilities. Beat them on opportunity and subsidize workplace safety and vacation.

Provide tax incentives or transferable/sellable credits. We could even federally purchase the land and provide no-cost ground lease.

Dollars to dollars, this would make way more sense than extending Amtrak. It'd also stop population decline and create a next generation of American consumer.

Assess the DARPA funding as a portfolio.

“DARPA Awards Moderna Therapeutics a Grant for up to $25 Million to Develop Messenger RNA Therapeutics” (2013)

https://investors.modernatx.com/news-releases/news-release-d...

I believe DARPA is a national treasure, and has driven a ton of otherwise neglected innovation, but let's not overstate their contribution to Moderna. For exsmple, AstraZeneca did a deal with Moderna including a $240MM upfront to another $180MM for potential milestones six months before the DARPA money. That is all on top of a pile of venture money and other corporate deals over those few years. The value of the DARPA $25MM wasn't the money per se, it was the incentive for Moderna to keep a toe in infectious disease, which was a hobby for them (and the industry) at best before COVID. I believe the platform would have been established with or without the DARPA money.

https://investors.modernatx.com/news-releases/news-release-d...

>For exsmple, AstraZeneca did a deal with Moderna including a $240MM upfront to another $180MM for potential milestones six months before the DARPA money.

This is called "biobucks" in the industry. There's an incentive to close big deals for management and business types in biotech, so they love to craft these large headlines to drop it on their resume and grift at another big pharma.

I'd be surprised if they ever received the $180MM, although $240MM upfront is really impressive.

I think "grift" is a bit cynical. I agree the 'biobucks' concept largely exits, but it also good deal-making. A huge upfront is 100% risk on the buyer. Paying out incrementally for accomplishments is common and good business sense risk-sharing and should happen in almost any industry.

The larger point is that Moderna was flush with cash for years before COVID, and the DARPA money wasn't that 'crazy' of an investment, or essential to Moderna's progress.

Oh this one deeply paid off. Great find.

Agreed - one of the benefits of DARPA is they fund long-shots which once proven viable (or not) are then commercially developed.

This is a good outcome, the money wasn't 'wasted' since it spurred development in a particular area of interest to DARPA - advanced robotics.

Is that even possible? I imagine lots of, if not the majority of what DARPA does is classified.

Not everything DARPA does is going to be a hit. Don't forget they're responsible for virtually all of the funding that helped build the myriad technologies that comprise the Internet.

DARPA funding is actually supposed to fail at a reasonably high rate. If they succeed a lot it means they are not taking big enough bets.

I don't know where to even look for this, but it would be interesting to see if the DARPA investment bought them a perpetual, irrevocable license to technologies they funded.

I think their investment in DARPA has a positive externality for society. If we account for this externality, perhaps taxpayers enjoy a positive ROI?

It's just too bad that the DARPA couldn't flip it like Softbank did in just a few years.

They first sold it to a US company without a vision (Google).

If Google can't figure out how to get a messaging app right, there's no way they can make sense of this purchase.

DARPA basically started the self driving industry

Hyundai has it's own Robotic labs too. They have plenty of reasons to invest in Boston Dynamics, apart from just car parts.

In addition, Robots manufacturing cars or car parts, improving their own robotic products could be another reason behind this.

Automatic plugs for electric cars would be a welcome update, like getting your gas pumped in Jersey

Reply on news.ycombinator.com