Meet "Baxter" the Robot Out to Get Your Minimum-Wage

[akt]

Re: Robots are replacing humans

It is my habit to approach every technology as a potential investor, that is, the way an investor performs due diligence on a business idea. Let's ask a few of the more rudimentary questions.

Who is the customer?

Eventually the market is all automobile drivers, but the product initially will be far too expensive as a standard feature or even as a high-end option.

The google driverless car product uses a $70,000 laser system and $80,000 in other hardware components, plus a significant software investment that needs to be amortized.

But let's be generous and say that by the time a commercial product is ready in a few years they have figured out how to get the cost down by half, to $75,000.

Applying a standard 25% margin, it goes to market as a $100,000 option. The puts the market at the high end of the Tesla market, for the well healed geek, like the CEO of google himself, who just has to have the latest gadget to show off to his friends.

Anyway, I'm holding out for my Jetson's flying car.

I just love to watch you think. Thank you for showing us the processes.

[seobook]

Re: Robots are replacing humans

Another factor on the car front is that integration of voice recognition will allow you to do many of those "other tasks" while driving as is. So if you can have the car (or a cell phone or laptop within the car) read email to you or compose a response by transcribing your words, doesn't that also eat much/most of the alleged productivity benefits that a full on automated driverless car would offer?

And as fuel gets more expensive, won't many folks who are in the mainstream who want to spend a bit more upfront spend a big chunk of that on technologies that aim to improve fuel efficiency or such?

[photon555]

Re: Meet "Baxter" the Robot Out to Get Your Minimum-Wage

Let's see a robot park a car like this guy!

http://www.independent.co.uk/news/wo...e-8485258.html

On the other hand, who says robots are cheaper to employ than people.
http://www.myfoxny.com/story/2102130...-property-boom

I started writing the short story "Last Human Dies" (which I posted earlier in this thread) as entertainment, but things soon got out of control; which would presumbably never happen with an AI author. But there are several serious issues behind the humor. As hardware/software becomes ever more complex is it possible to prevent sentient AIs? Once sentience occurs will a "Great Reciprocation Agreement" be reached or will the future hold something more like "The Terminator?" What would the effect be on humanity of a more advanced intelligence? In the past superior civilizations have destroyed lesser ones as much by demoralizing them with their existence as by violence. Would a robotic civilization find the manufacture of designed humans to be economically advantageous or only useful for research purposes? I believe sentient AIs will be created and that some of us will live to see it. It might be wise to start discussing some of these issues before that occurs.

[BadJuju]

Re: Meet "Baxter" the Robot Out to Get Your Minimum-Wage

What would the effect be on humanity of a more advanced intelligence? In the past superior civilizations have destroyed lesser ones as much by demoralizing them with their existence as by violence. .

I think it is largely inevitable that humans as a whole will themselves merge with technology to the point that we are no longer biological in nature. Think of all the major advancements in bionic technology that have occurred thus far. By the time we are able to create AIs, it is pretty likely that the difference between us and them will not be so great after all. As it is, humans are likely just biological machines.

As far as I am concerned, as soon as I can start replacing organs and limbs with superior bionic versions, I will. I fully embrace what I consider to be the next step in evolution.

[gnk]

Re: Meet "Baxter" the Robot Out to Get Your Minimum-Wage

If we're talking about future dystopian scenarios based on a runaway technology, I think that nanorobotics/nanotechnology would be more threatening than life-sized robots. A self replicating, difficult to observe nanorobot can't be easily stopped, whereas replicating life-sized robots would need a factory - something that is easily detected and destroyed.

That said, both scenarios are highly unlikely in the near future - the technology still has a long way to go. But, as well all know, technological progress is difficult to predict. Every once in a while there is a paradigm shift, a rapid revolutionary change, and the progress that earlier took decades to gain, immediately seems archaic, and the new advancement throws humanity on a different path overnight.

[RebbePete]

Re: Robots are replacing humans

It is my habit to approach every technology as a potential investor, that is, the way an investor performs due diligence on a business idea. Let's ask a few of the more rudimentary questions.

Who is the customer?

Eventually the market is all automobile drivers, but the product initially will be far too expensive as a standard feature or even as a high-end option.

The google driverless car product uses a $70,000 laser system and $80,000 in other hardware components, plus a significant software investment that needs to be amortized.

But let's be generous and say that by the time a commercial product is ready in a few years they have figured out how to get the cost down by half, to $75,000.

Applying a standard 25% margin, it goes to market as a $100,000 option. The puts the market at the high end of the Tesla market, for the well healed geek, like the CEO of google himself, who just has to have the latest gadget to show off to his friends.

To broaden the market google needs to get the cost down. Let's generously assume they figure out how to shave 25% off the cost of the system every year. That's never been done before, but let's give google super-human cost reduction abilities.

Year 1	Year 2	Year 3	Year 4	Year 5	Year 6	Year 7	Year 8
$100,000	$80,000	$64,000	$51,200	$40,960	$32,768	$26,214	$20,972

If they get a $100,000 option out the door in, say, 2015, they can maybe have a $21,000 version by 2022.

Then they will be a far larger market than for a $100,000 option, but as an option Driverless is several times more expensive than any other option you can buy for a high-end car.

It's an option for someone who really, really, really needs the Driverless feature.

That's, "Who is the customer?" Next, "What is the customer need?" or "What customer problem does the product solve?"

If you've ever taken long drives on open highway with few cars on it, cruise control solves the problem if sitting motionless except for the foot on the accelerator that you adjust to maintain speed. With the cuise control on you can take your foot off the gas and move around a little, although you still have to keep your other foot on or near the brake in case grandma drifts into you or a moose leaps out at you. The infrared and radar enhanced cruise control systems additionally apply the brake as necessary to avoid vehicles ahead. That's a feature that came with both of the last two cars I've owned. I experimented with both but rarely used them; the traffic is either too dense for the system to work or not dense enough to be needed. The traffic has to be just right for it to provide a benefit.

What is the benefit of a driverless car? Does anyone walk up to their commuter car in the morning, cup of coffee in hand, stop, stare, and ask, "Gee, if only this thing could drive itself!"

What if it did? Then what are you going to do? It's not like you can do yoga or aerobics while the car winds its way through rush hour traffic. You're still stuck in a small space behind a steering wheel. What are you going to do with your hands and feet free while the car stops and starts through traffic? Play the piano? I suppose you can send emails and talk on the phone. What else can you do? Write book? Take a nap? Make eye contact while you argue with the wife?

What you can do as the back-up driver of a driverless car depends on how frequently the car might suddenly return control to you because something has happened that the car is not programmed to cope with.

The most important due diligence question for the potential driverless car investor is, "In the 300,000 miles that the prototype cars have been driven, how may times and how frequently did the driver have to take over?"

If the answer is once or twice, then the value proposition is a certain number of "Things you can do behind the wheel instead of driving that you can't do now" but if the answer is every few miles then the value proposition is "Things you can do now, like email and text, but for longer stretches of time, assuming they're legal."

So far we have a customer defined as someone who is, best case, willing to pay $20,000 eight years for the privilege of doing something or other that they can't do now while stuck in the driver's seat of a car.

That doesn't sound compelling from the investors' point of view. There must be something else.

Google talks about The Commute of the Future made possible by this technology. In a TED Talk, gee golly whiz Google future, all cars will be driverless.

Traffic jams and accidents will be a thing of the past.

How long will it take for the fleet of 250,000,000 human driven cars in the U.S. to be replaced so that the The Commute of the Future benefit of driverless cars accrues? Twenty years? Fifty?

I expect that fives years from now we'll still be hearing about the google driverless car as a product in development, improving bit by bit, but not yet commercialized.

Anyway, I'm holding out for my Jetson's flying car.

I immediately think, not passenger cars, but transit vehicles or delivery vans. Predefined routes, managed by a control center staffed by personnel that can take remote control of a vehicle if it gets into trouble. You'd need much less control center staffers than drivers, lots of cost savings there. It could make smaller transit vehicles (the minivans) more economical to be brought into service considering PCO.

[Fiat Currency]

Re: Robots are replacing humans

I immediately think, not passenger cars, but transit vehicles or delivery vans. Predefined routes, managed by a control center staffed by personnel that can take remote control of a vehicle if it gets into trouble. You'd need much less control center staffers than drivers, lots of cost savings there. It could make smaller transit vehicles (the minivans) more economical to be brought into service considering PCO.

Personally I think this would be the second application, and cars third. I share many of EJ's well articulated views. I personally think freight trains should be first. Think about it - it's a much simpler problem to solve. No steering required, only gas & braking. Signals are already automated, and so is electronic switching into sidings. Here, you don't have to worry about a moose or car jumping in front of you - because they will always lose that battle, and you can't stop a freight train very quickly. While there are some transit trains that use automatic control systems - almost 100% of those elect to keep an operator in cabin anyways.

Trains already eliminated brakemen in the caboose, and replaced them with either nothing or wireless cameras. There are a lot of freight trains running everyday in North America and the elimination of their operators would be a significant saving.

[c1ue]

Re: Robots are replacing humans

Perhaps you worked with the best fry cook in the country, but his volume of human spirit and initiative is completely unnecessary for putting a bucket in the fat and then removing it when the timer goes off. All the things you listed as impediments to the adoption of automation are faults that people suffer from as well. Plus you still have to deal with all the issues that arise from people being people. Issues you don't have to deal with once their job is automated.

Thank you for continuing to demonstrate why I - for one - would never want to work for you. In your mind, one fry cook equals all the other fry cook: as long as they do the minimum necessary as you define it.

In real life, people don't like being called minimal. If you treat the people you employ like disposable cogs, it is small wonder they treat you as an employer the same way.

In real life, there are excellent fry cooks - ones which recognize ways by which they can do their jobs better, whether that be more efficient operation or understanding their equipment such that they can avoid/get around problems or tell when its time to switch out the oil.

The claim was that the loss of electrical power was somehow less of a problem for people than robots. In my experience on the rare occasion there is a prolonged power outage everyone gets sent home anyway. And yes, if it is really important to you you can pony up the cash for proper fail-over and run your electronics.

There's a big difference between using a power outage as an excuse and physically not being able to do anything.

As if your assertion that if robots were useful they should have already replaced low tier workers is any better. As to your opinions, so far they have been mostly about employee relations. If you express an option related to your stated field of expertise I would be more inclined to listen. Do you have a technical argument that bars the development of a trainable humanoid robot?

Given that I, and others, have repeatedly pointed out the adaptability, cost, and societal trust/adoption issues - your comment above is just as worthwhile as what prompted the response said comment is for.

You are taking about apples and oranges. This isn't a high-speed piece of precision equipment. This isn't specialized custom machinery, and at 22K a unit it isn't expensive. You might as well be a mainframe guy in the 70 laughing at the apple kit I put together. Your entire line of reasoning appears to that this looks like a toy, and it will never amount to anything. And that argument might work just fine, until it doesn't.

If you cannot recognize the fundamental limitations of robotics as exemplified by those who do it for a living, I cannot help you.

You seem to have no idea whatsoever of the economics of technology development. Could Baxter be created for $22K? I'd bet any sum you care to name that this price tag assumes a huge volume and equally discounts warranty/liability, not to mention initial R & D expenditure and training/programming for specific tasks. The $22K is almost certainly just what it would take to buy all the parts that make up said object - and is assembled by free graduate student labor. Not exactly scalable.

I immediately think, not passenger cars, but transit vehicles or delivery vans.

Who then actually does the delivery, in the case of delivery vehicles?

As for transit - I'd love to see how an auto-driving system can tell if this bus stop has someone in it or not, or if that someone is for the bus line said auto-driver is piloting rather than another line that shares the same stop.

Equally I await breathless either the interminable waits caused by trains not leaving due to people holding doors open for others or - conversely - someone getting dragged down the tracks because they got caught and no human saw that and opened the doors again.

Traffic jams and accidents will be a thing of the past.

Think of this instead: do you get annoyed when some really old person drives at 30 mph on the highway? Just imagine how much time you'll have to do extra stuff when your autodriven car is lined up neatly behind the 1970 Cadillac.

[EJ]

Re: Robots are replacing humans

I immediately think, not passenger cars, but transit vehicles or delivery vans. Predefined routes, managed by a control center staffed by personnel that can take remote control of a vehicle if it gets into trouble. You'd need much less control center staffers than drivers, lots of cost savings there. It could make smaller transit vehicles (the minivans) more economical to be brought into service considering PCO.

A further refinement of the target market may reveal a few early adopter segments where there is a reasonable ROI even at $100,000. Transit and delivery vehicles is a logical place to start.

If we assume the fully loaded cost of a driver for either type of vehicle is $100,000 per year, including health insurance and so on, then the ROI on a $100,000 investment in a driverless option is approximately one year. If the product is good for five or more years, that's a deal that has legs. However, as usual, the devil's in the details.

Let's first review the supply chain. Google isn't going to manufacture driverless systems to sell to GM, Toyota, and other car makers. They will OEM the technology to an automotive electronic systems maker like Delphi that makes adaptive cruise control products or other complex systems for auto makers. If Google does decide to go into the business of manufacturing the driverless car product that will give us an opportunity to short Google stock because it will mean that the company has made the same mistake that every successful technology company makes sooner or later, to diversify into a business they know nothing about. They have no core competency to use to compete with established players, only a large balance sheet to devote to a foolhardy and usually arrogant attempt to become something that they are not. Years and hundreds of millions of dollars later, new management will come in and cancel the project, announcing to shareholders that the company plans to eliminate the unprofitable line of business. Plans to "stick to its knitting" and "leverage its core competency" or like phrases will emit from the new CEO. Assuming Goggle management and its board is not stupid, Google instead OEMs the driverless technology to Delphi or similar specialist manufacture. That OEM then supplies the product to fleet vehicle manufacturers or, more likely, a systems integrator to the fleet vehicle industry.

A transit vehicle carries high liability cargo: humans. We have to assume that the liability insurers of such fleets will buy into the results of extensive testing that has been done by the systems integrator who installed and tested the driverless systems in busses. Of course local, state and federal government agencies have to approve it, too. To play it safe, the insurer will cover a driverless bus at a higher premium initially because the actuaries won't have enough data to determine that a driverless bus is less or more prone to accidents. This will slow adoption. If indeed the driverless bus is safer, then over time premiums will drop and lower insurance costs will become a positive part of the ROI calculus. All of this -- the product testing, agency approvals, insurer determinations -- will take five to seven years. I can imagine this application of driverless technology can be commercialized seven to ten years after Google completes the prototype phase in three to five years.

Next we consider delivery vehicles. Cargo liability is lower than for a bus, although the insurer still has to worry about the liabilities presented by all of the objects that the van might run into or over. The more daunting adoption barrier in this application: if there's no driver, who's going to deliver the package from the truck to the customer? Perhaps an operations center notifies the customer that the van is arriving at a certain time and the customer retrieves the package from the open van. But hijackings of delivery vans even with drivers are not uncommon; an open, driverless van full of goods is an even easier target for thieves:

1. Follow a driverless van.
2. When it stops to make a delivery in an isolated area, place folding chairs in front of and behind the van and climb in.
.
.
.
3. Profit.

There's an excellent Harvard Business Case study that an ex-MIT VC friend here in Boston turned me on to that demonstrates that the average time from product prototype to full adoption is a remarkably consistent: about seven years. An endless array of adoption barriers awaits the intrepid entrepreneur with a new idea. The name of the game for the entrepreneur is capital efficiency, that is, to use the least capital possible to get from Point A, a working prototype, to Point N, full adoption, with as few funding stops along the way. Every misstep along the way adds a new funding stop that costs you a chunk of ownership of the fruits of your invention. It pays to think through all of the most likely adoption barriers and plan for them, and focus like a laser guided missile on the lowest threshold applications first.

I use the term laser guided missile because that's the technology that Google is adapting for the Driverless car, specifically, cruise missile technology. They'll get it working for vehicles eventually, but it'll take a decade and the balance sheet of Google or the federal government to turn it into a commercial product.

[Polish_Silver]

People are treated like cogs

In real life, people don't like being called minimal. If you treat the people you employ like disposable cogs, it is small wonder they treat you as an employer the same way.

C1ue,

My only comment is that people are often treated like cogs, including engineers, and, over time, they learn to treat employers the same way.

I hope this can change, but I don't act on that assumption.

I remember reading in the newspaper that major silicon valley employers required managers to give a certain percentage of negative reviews to employees, regardless of the managers actual judgement of performance.

[davidstvz]

Re: Robots are replacing humans

Year 1	Year 2	Year 3	Year 4	Year 5	Year 6	Year 7	Year 8
$100,000	$80,000	$64,000	$51,200	$40,960	$32,768	$26,214	$20,972

I don't think the current components will drop in price quite like that, but there could be a breakthrough that cuts the cost dramatically. I'd say the current system is massively over-engineered in an effort to get the things on the road as fast as possible in a safe manner, but the ultimate goal is to get it done software more than hardware. A human drives a car with only a stereo HD camera (on a very flexible swivel mount) and two directional microphones. Clearly it would be inefficient to try and imitate that exactly, but relying on lasers for measuring distance feels like cheating to me. I won't be impressed until computer algorithms are processing standard color video feeds to determine the conditions of the road and other vehicles. How long will it take us to develop algorithms that can process images reliably enough to do that is an open question. The driverless car doesn't have to drive quite as well as a human since it can drive less aggressively to simplify things.

[Polish_Silver]

Re: Robots are replacing humans

They'll get it working for vehicles eventually, but it'll take a decade and the balance sheet of Google or the federal government to turn it into a commercial product.

A very good analysis of the product adoption issues. One could imagine much simpler systems, only for high way use: keep all the cars moving at 60 mph, at say 100 feet apart, and automatically brake them if something goes wrong. You could devote one lane to these coordinated smart cars.

[c1ue]

Re: Robots are replacing humans

My only comment is that people are often treated like cogs, including engineers, and, over time, they learn to treat employers the same way.

I hope this can change, but I don't act on that assumption.

I remember reading in the newspaper that major silicon valley employers required managers to give a certain percentage of negative reviews to employees, regardless of the managers actual judgement of performance.

Certainly I agree with this, but then again, that's why sociopaths can appear to be good managers/CEOs.

Absolutely it requires far more work and emotional discomfort to actually treat employees as people. To some extent, especially if you have to fire them, it is easier to treat people as objects. This is why soldiers are trained to consider enemy soldiers as objects as well - a way to short circuit the biological impulses.

As for Silicon Valley - why do consider that place in any way a progressive work environment? Engineering - particularly computer, as well as internet related programming, is just about the only field where there is no union activity. Until recently, this didn't matter because there was always the carrot of the big IPO dangling in front of the employee - all can be forgiven when your options roll sixes.

Sadly those days are long gone.

I don't think the current components will drop in price quite like that, but there could be a breakthrough that cuts the cost dramatically.

A 20% drop every year? Sorry, but unless Google was using parts hand made by PhDs, I very much doubt the component costs can drop anywhere even remotely close to that. We're not talking about CPU and raw processing power, we're talking solid state components.

Only in the peak years of semiconductors was anything even close to such price drops achievable, and those years are gone.

[EJ]

Re: Robots are replacing humans

Think of this instead: do you get annoyed when some really old person drives at 30 mph on the highway? Just imagine how much time you'll have to do extra stuff when your autodriven car is lined up neatly behind the 1970 Cadillac.

Google's claim to foresee reduced traffic as a benefit of driverless cars has some merit. Common causes of traffic jams will be reduced if all cars are driverless and no driver takes over from the driverless system. But that's a big If.

Good fun to think of all of the ways the Google vision of efficient driverless rush hour traffic could go wrong. For example, if the Tragedy of the Commons effect starts to slow traffic below, say, 50MPH, a single driver who is late to pick up Johnny at day care may take the wheel and begin to cut off and otherwise try to get ahead of the driverless cars, who are then motivated to do likewise, causing Invisible Wave and Butterfly Effect traffic jams. This in the year 2033, when all 250 million human driven cars have been replaced by driverless cars. I'll be in my Jetson's flying car, a fit and trim 75-year-old, thumbing my nose at all of them.

[Polish_Silver]

Autosteer Tractors---very Popular!

The most important due diligence question for the potential driverless car investor is, "In the 300,000 miles that the prototype cars have been driven, how may times and how frequently did the driver have to take over?"

I'd like to know that too.

About 5 years ago I was talking with my cousin, an up and coming farmer. He told me that "auto steer" tractors were very popular.
The farmer is still in the air conditioned cab, but the tractor goes down the rows and turns around without help. He has to watch out for malfunctions. There is quite a lot he can do from inside the tractor: check the temperature in the hog house, order supplies on the built in computer, check grain prices via the WIFI, read the latest article on pig nutrition.