Hacker News Reader

Comment by ghshephard | Hacker News Reader

It's going to take about 40-50 years to reproduce what humans can do in a restaurant today with robots. More likely, the food preparation process will be altered to allow for easier automation. With the exception of a supervisor/engineer (that position won't be automated for at least another 100 years) - I think we'll see a fully automated fast food restaurant in about 20-25 years - though some elements of the restaurant business (Cleaning, French Fries, Order taking) - will likely be automated sooner - there will be demonstration restaurants with 50% automation in about 10 years.

[note: And yes, I do realize that the above predictions will require radical advancements in automation that are not easy to predict - call me an optimist. :-) ]

View on HN

I think that you're wrong and I will illustrate why. When I graduated from college I spent an evening with friends and we mused what we'd buy with our new paychecks. I wanted a brand new Mustang and indeed I purchased one. But one friend, the only computer major, said he wanted a computer.

This was pre-pc era and the one he wanted cost as much as a house. We teased him all night about how speedily he could balance his checkbook or play checkers. Finally he'd had enough, pounded the table and said with a red face in ten years you will all have a computer that sits on your desk.

Our only exposure at that time was to the IBM 360 mainframe the school had and that only made us laugh harder. Truth is he was right, every last one of us had a pc in ten years.

The computer industry moves faster than you think. A few years later I went to a local store to see one of the first laser printers for sale. This one was $10,000 and as a group of us looked at it with the proud sales guy. I said as soon as I could get one for under $1500 I'll buy it. Everyone looked at me like I was crazy and the salesmen said maybe in ten years it will happen. Actually I got an HP Laserjet in around three years for $1499 and still have it - somewhere.

I'd wager that most of us on HN will have a robot in our house in well under ten years. I can't wait;<).

Robots are harder than computers or printers.

Particularly considering that, before computers and printers came along, you couldn't buy the equivalent for minimum wage.

It depends on the robot.

A Roomba can be had for the price of a few weeks' groceries.

Industrial assembly robots run about $60k, though deployment costs are over 3x that. Curiously, reprogramming a robot is considerably more expensive than the robot itself, according to one source.

If you have a well-defined, and reasonably attainable task, costs are likely predictable and possibly finite. If you've got a poorly defined and hard-to-reach goal, expenses will likely rise to meet resources.

One thing I'll comment on this is that I have a Roomba, love it, but it does require quite a bit of human care. About every three clean cycles (I have a one bedroom apartment) I have to take it apart, clean all the hair that gets stuck around the brushes + gears) and assemble it back together.

FWIW, you'd have to perform similar maintenance on a standard vacuum.

As well as, you know, vacuum with the damned thing.

>50% market penetration for Roombas isn't a hard bet.

The equivalent of "Moore's law" for robotics hasn't shown similar growth. Many things - memory, storage, screens, transistor density, networking (wired / wireless) have growth curves that far outstrip what we've been able to do with robotics - my predictions are optimistic and suggest some advances in the art that aren't an extension of current trends.

But robotics hasn't really gotten its consumer spotlight yet.

If we ignore robotic vacuum cleaners and toys. But there are some incredibly cool robotics applications.

What needs to happen is a "killer application" of a robot in workplace that will be transferable to home. Off the bat I believe that robotic cars (or autopilots as they will be known) will be first major application of robots that can be transplanted from corporate world into home use.

And I can easily see that in 10 years a lot of new cars will be 1. electric and 2. will have an autopilot option.

Depends on how you define robot—do my Roombas count? Does my bread machine?

If they do, I wonder, do most of us already?

"It's going to take about 40-50 years to reproduce what humans can do in a restaurant today with robots."

I'm not so sure. Take a look at the machinery that produces packaged food now-- it's fairly impressive that every step is automated and human intervention is seldom involved. I don't see a technical reason an entire McDonald's couldn't be completely automated now. If you can analyze and quantify each step in the process, you can build a machine to perform each step.

Customers expect to give special orders (hold the pickles, hold the lettuce), but you can provide those options in a software interface. Customers expect handling of exceptional circumstances (the pie is too dry, please replace it), but McDonald's-style preparatory science can help prevent such things (e.g. instead of cooking 12 pies at once, the equipment could cook a pie every few minutes.) I think the biggest problem is customers expecting to give orders in spoken language, but we're having trouble cracking that nut with software (noise, dialects, etc.) Also, there's political backlash for having few (if any) employees.

Why would the robots need to recognize voice at all? Simply have a text menu which could be translated into all major languages and let people pick from that using either their smartphones or wall mounted touch screens.

From a technical aspect, they don't need to. But that's what people expect. John Q Public wants what he wants regardless of what the techies think he wants.

I've never understood the fascination with voice control. It's clumsy, it's inaccurate, and honestly it's a bit embarrassing to use alone or in public. Yet, "everyone" seems to want it, despite the fact that spatial-sensor-based (e.g. motion sensors, Kinect, etc. -- note that I'm not talking about gestures, which have the same problems with ambiguity as voice) and tactile controls are much more efficient.

Reading this made me picture trying to drive a car by voice: "Turn left. Left. Further left. More! AHH! NOT SO MUCH!" Here voice control seems like a desire to avoid having to clearly specify your request. "Didn't the car KNOW where I wanted to go? Doesn't this spreadsheet KNOW what numbers I want? Doesn't this fast food menu KNOW what food I want? Maybe if I could just tell it..." I suspect that voice recognition will be popular in the future, but it will seem a lot less sexy if users are still required to clearly specify what they want.

You've got it exactly -- the more important thing, IMO, is systems "knowing" what you want, predicting your behavior, in a way that is completely effortless, transparent, and expected. I'm doing that now with my startup, using pre-scripted actions, but hope to see user-adaptive, learning systems (e.g. the Nest thermostat) find their way into everything.

Not so sure, people are used to buying allot of things without needing to use their voices. When amazon launched I doubt that the fact people couldn't specify which books they wanted to buy via voice control was a major hurdle.

I think the next generation will view buying most things via some text or graphics based menu as nothing unusual at all.

You don't walk into book stores to ask employees for the book you want. You browse, look, read, and finally select a book. No interaction with humans required. (There's human interaction at the point-of-sale, but you still don't have to speak to checkout.) Of course, if they didn't have the book you wanted, you might have to speak to a human to get it ordered ... if they didn't have computer kiosks for you to use instead.

Moving this type of sale onto the internet was not a change. Moving the food service industry, where human interaction is expected, into an environment without human interaction goes against people's expectations and habits. Is such a change technically feasible? Yes. Is it practical? Yes. Profitable? Sure, if you can change people's expectations.

I'm not saying it's a bad idea. I don't even think it'll be a difficult change. I personally would prefer it. But the first chain of restaurants to offer automated food service (and I do believe it'll need to be a chain) has many psychological barriers to overcome, in its investors and its customers.

Simply have a text menu which could be translated into all major languages

My understanding is the cash registers at a lot of fast food chains currently use pictures/icons rather than text (and a lot of cashiers can't do simple math for making change if the machine doesn't calculate it for them or there was an error or you would like to change something after the drawer is open). I don't see why this wouldn't work for ordering, and no need to translate into "all major languages".

What I think will happen is that we won't reproduce how humans work in an McDonalds, we'll design new packaging/delivery mechanisms so that robots can deliver a Big Mac without having to do things like lift the burger from the freezer to the counter to the heating ovens to the table and finishing in the microwave.

Also - I would far, far prefer to order my food from a decent touch screen - I suspect the accuracy would greatly increase. (One of the best inventions ever in the takeout was the video screen showing you what the order taker thought you had ordered)

What parts of what humans do in restaurants would take that long for a computer to be able to do?

Of course I'm not talking about proper restaurants here, I'm talking about somebody who creates a fast food joint based on what can be automated as easily as possible. If something requires a human to make then it doesn't go on the menu.

After all Mcdonalds was basically designed to emulate a factory from the era as much as possible and factory production processes have moved on substantially.

The supervisor role would basically be replaced by a technician, but since humans need allot more supervision than robots, if there was enough reliability and redundancy built in then instead of a full time manager these chains could simply rotate engineers interchangeably between sites (since they are all using the same equipment which does not have human quirks like loyalty).

20-25 years is a pretty short time frame, given there is nothing in robotics technology today that suggests we're close to doing this. I've seen sandwich making robots for six years - they never seen to get much better than the OP video.

From that video allot of it seems focused on using human tools to take food (such as toasters, forks etc).

It would probably be better to replace the single robot with things like a toaster robot and a lifting robot etc.

I think you're overestimating how long it'll take until we have automated restaurants. I think we'll have deployment of several fast food places with automated drive thrus within ten years.

Though I expect it to be a touch screen coated with thick plastic that horribly obscures the actual menu. Like a shoddy ATM at a corner store.

Then again, I'm perpetually disappointed with the slow pace of advancement. I often expect too much.

I guarantee you that there will not be a fully automated fast food restaurant in 10 years, possible exception being a massively inefficient MIT Science project.

There is just no way to make it economical versus hiring humans for $10/hour in that time frame.

I think it will be more like the "self-checkout" stations we see already at grocery stores. Instead of 3-4 people cooking, and 2-3 people taking orders at the counter/drive through, it'll be a bunch of "robots" cooking the food, and a self-service ordering station, with one person that oversees the robots, and one person overseeing the registers. Unfortunately, you usually need a human being to handle exceptional situations whether it's the human beings making mistakes, or the food preparation going wrong.

Reply on news.ycombinator.com