Hacker News Reader

Comment by perl4ever | Hacker News Reader

"a couple more years of improvements and you could have a machine able to reliably do the work of 2- 3 employees"

I see frozen uncooked pizza in stores. I feel like it existed 20-30 years ago too although I'm not sure exactly. Surely these are not hand made.

Whenever I read about impending automation, I wonder which millennium I'm in.

View on HN

Exactly. In the 1980s people in America mostly got coffee from machines, in their home or office. Few humans were employed making coffee ("barista" was not really a job title in the 1980s and if you told people it would be in the future, they would probably think you were crazy).

I know what you mean, not sure why others are brining in spurious arguments. Coffee shops in the 1980s were donut shops, mostly with truckers sitting around.

Well it's gone full circle now, as Starbucks (and a lot of other chains) use machines to make the espresso part of the drink at the touch of a button.

The barrister is still responsible for steaming milk, but I only assume that is because the sounds and smells of that are what make the coffee shop atmosphere.

Starbucks is selling an experience and it's central to this experience that there be a human putting a little TLC into what happens, whether it's a friendly greeting, a product recommendation, remembering your name and writing it on the cup, etc. The key is literally for there to be a human involved, because we get emotional satisfaction from a positive interaction with other humans.

There's nothing new in this discussion -- we've been automating away manual labor since the Industrial Revolution. We end up creating these human touch jobs with the resulting productivity gains (this is the transition from a manufacturing economy to a service economy, the value isn't in the muscle anymore, it's in the relationship).

In fact, thanks in no small part to capitalism's ruthless efficiency, loneliness and isolation are at record highs. The demand for a little TLC has never been higher.

We'll have fewer pizza makers but more people whose job is essentially to make you feel good.

From my experiences with coffee machines that produce cappacinos, the act of frothing milk seems to be one of those tasks that is much more difficult to automate than I would have thought. I've had a few that were passable, but none that are remotely as good as a decent human made one. The actual coffee on the other hand seems to be easeir to automate.

There's a fair amount of skill involved in pulling a good espresso shot.

Fair enough, though a well adjusted Super-Automatic Espresso machine can reliably and reproducibly get you 90% of the way there hundreds of times a day. The majority of customers wouldn't be able to distinguish that last 10% anyway. Have a barista on hand to adjust the Super-Automatic daily, to account for changing bean freshness, and you're pretty much good to go.

What about variance in tastes of customers or even variance in bean itself (beyond age).

How can a barista accomodate the variance in taste of a customer when pulling an espresso shot? The barista can dial in the pressure profiling, how long to pull the shot, and the grind size. For a given batch of beans, based on the type, level of roast, and age, there likely exists a sweet spot for those parameters that most baristas would agree upon. My understanding is that this is what baristas do anyway - adjust the machine for the beans available today, and then just crank them out all day long. A super automatic (or future versions) that's dialed in daily could achieve pretty much the same thing.

AFAIK, the barista will try to pull the best espresso out of the beans that they have. I'm not sure variance in customer tastes can be accomadated at this level, beyond a ristretto or an americano.

There is which is why I’m finding that the Costa machines here in the UK are consistently better than the humans are.

I never had a coffee from a Costa machine, but purely from the looks of it, and my experience with similar looking machines, I wonder: does this really produce espresso (as in say 16grams of fresh ground beans in, water at proper temperature and pressure of like 6 or more bars, yielding something like 32grams of beverage) or rather something which can best be described as 'strong coffee'?

Your metrics for an espresso seem incredibly trivial for a machine to exactly reproduce, yet wildly open to human error and deviance.

Yes that's pretty much spot on. Making espresso with a manual lever machine isn't too hard too mess up, especially not in comparison with pressing one button or filter coffee. Yet it's apparently not that trivial to make such a machine, and keeping it working for hundreds of cups a day: prices of those are easily well over 10k$.

Sure but espresso varies, change the bean even the age of the bean and the recipe needs to adjust. Not to mention not everyone wants the same experience every single time, and not every person wants the same exact espresso that the last person had.

Good question. I’m not sure I care if I’m honest. It tastes pretty good and keeps me alive :)

It seems like this is a matter of fashion and it could easily swing the other way? Why couldn't a machine make excellent coffee?

They can. People pay up for the experience.

I think the key to this is to hide the robots from public view. Crank out pizzas, but not with the machine in human view. Have someone visibly handing them out. Let buyers assume the pizzas are made by artisans out the back.

Of course it can. One of the best machines for making coffee is the Technivorm Moccamaster (the thermo pot version). It's been in the market for decades, lasts for decades, it's easy to clean and together with a good grinder and fresh good beans, it produces a perfect mug of coffee every morning.

does this follow the trend for (or perhaps presage) an increasing emphasis on “user experience”?

Coffeehouses have been a thing for over 500 years now, so you must be pretty old.

Pedantic. In the 1980s there weren't many in America, now there are Starbucks'es and others throughout.

They were a thing in the 50s, 60s, 70s. Maybe more of an urban or hippie thing.

I know the term "coffeehouse" was a cultural touchstone in the US because it was used in titles of books published before the 80s. Sorry I don't remember specifically, but this "nothing existed until Starbucks" is ridiculous. Starbucks came out of somewhere. Coffeehouses came from Europe to the US, before and after the American revolution.

I'm tempted to just call out the absurdity of arguing this point, but instead I'll just air-drop a factoid: if you Google "https://www.statista.com/statistics/196590/total-number-of-s..., you'll find a graph showing approximately 1,650 specialty coffee shops in the US in 1991, and 31,490 in 2015.

I'd link directly to it but the site forces you to pay to see the graph unless you clicked on it from a Google search results page, which is bad and evil. I can't find a better source.

Starbucks has something like 15,000 locations in the US, half your latter figure. They could have a million stores and it still wouldn't mean they invented coffee.

They were pretty rare.

Even in NYC, a coffee shop was mostly a Greek diner. While it may reveal my Irish working class roots, I never saw espresso made or consumed outside of an ethnic Italian restaurant in Arthur Ave until I was like 17.

Last week, I had a mediocre latte from a Starbucks at an I95 truck stop adjacent to a cotton field in North Carolina.

Pedantic. The world is not America. Actually, the U.S. not America either.

or just American.

"Philadelphia’s first coffeehouse opened in 1703, and by mid-century half a dozen operated within the city limits"

https://philadelphiaencyclopedia.org/archive/coffeehouses/

The major difference here is the economies of scale. When you have a centralized facility, you can afford to invest in expensive, high-throughput machinery because you make up for those investments with volume. However, centralizing necessitates some distribution costs, which in the case of many foods imposes some harsh constraints on ingredients and quality; there’s a reason those pizzas are all frozen, and don’t reaaaaaallly taste all that fresh. If the cost of these small-scale production robots drops low enough (achieved through smarter control rather than high-precision parts), you can remove the ingredient constraints and make a better product while capturing savings of automation.

Fun facts: Average radius (from city centre) of top 20 US cities by population [1]: 15km

Top speed of the latest racing drone [2]: 265 km/h

Top speed allowed under FAA rules [3]: 160 km/h

Average Flight time for a delivery (7.5km @ 160 km/h): 2 min 48 seconds

Assume additional time for loading, take-off, speed ramp, speed slow, landing and unloading: Door-to-door in less than 5 minutes (average).

Sources:

[1] https://en.wikipedia.org/wiki/List_of_United_States_cities_b...

[2] https://thedroneracingleague.com/racerx/

[3] https://www.faa.gov/news/fact_sheets/news_story.cfm?newsId=2...

If you want to model a city as a circle, you should probably instead use a radius where half of the area of the circle lies within that radius. When you halve the radius, you divide the area by four!

For example, the area of a circle with radius 15 km is about 707 km², while the area of a circle with radius 7.5 km is only about 177 km². If the population density by unit land area is evenly distributed, this circle will only enclose a quarter of the city's population, rather than half.

I think an interesting future limiting factor is air traffic from delivery drones. Right now it seems absurd to worry about that because there are no delivery drones flying, but if this method becomes popular, the presence of other drones in the airspace could limit delivery drones' mobility.

You make a good point about circles.

I'd like to add that the pi r^2 circle we know and love isn't the only one. It depends on the "metric" which in that case is Euclidean ("as the crow flies"). If you have have traffic rules, like only flying certain routes you get a different shape circle and a different area as a function of radius. A famous example is the "Manhattan" metric where you drive around on a square grid. In this case the "circle" looks just like a square and has area 4 r^2.

I feel like there's a lot fewer SKUs involved in store frozen pizza than varieties available from a takeaway pizza store, and the volumes are also a lot higher.

So the trick might not be "cheaply and reliably automate making a fifty thousand a day of the same pizza" , but in the detail of "cheaply and reliably automate making a pizza, composed of any combination of the basic ingredients, but only amortised over a hundred an hour for a few hours each day".

That's an interesting point. Does that mean you think a chain like Little Caesars, that offers about 4 types of "Hot N Ready" pizzas could basically automate away most of its workforce? If frozen pizza is possible, it seems really weird that no one has pursued this vigorously already.

> Does that mean you think a chain like Little Caesars, that offers about 4 types of "Hot N Ready" pizzas could basically automate away most of its workforce? If frozen pizza is possible, it seems really weird that no one has pursued this vigorously already.

Little Caesar's entire business model is "the lowest quality pizza at the lowest possible price". I assume they already have automated away most of their workforce; given a minimum wage, that's a key part of low prices.

They patented a pizza robot last year

https://www.zdnet.com/article/little-caesars-just-patented-a...

I've only been to Little Caesars a few times recently, but in none of those experiences did I see more than a single young (hourly) employee running the entire operation.

The article responds some of your concerns:

“Machines have been making frozen pizzas for years, but Picnic’s robot differs in a few respects. It’s small enough to fit in most restaurant kitchens, the recipes can be easily tweaked to suit the whims of the restaurants, and — most importantly — the ingredients are fresh.“

From my extensive experience watching episodes of "How It's Made", frozen foods still often are hand-made. There are enormous assembly lines, heavily automated but with some steps done by hand. Especially those that involve irregularly-shaped, delicate, or sticky objects[0].

They did do an episode on frozen pizzas[1], and it is fully automated. But the techniques are difficult to adapt to making individual pizzas. Sauce, cheese, and toppings are sprayed all over the assembly line; the stuff that falls off is dumped back into the hopper.

They also did an episode on building pizza-making vending machines, and it does seem as if that was technology that was ready years ago.

[0] https://www.youtube.com/watch?v=_htXkd0djzE [1] https://www.youtube.com/watch?v=SqULgulLq3Q

Is there an episode on the frozen, uncooked, rising type?

> I see frozen uncooked pizza in stores. I feel like it existed 20-30 years ago too although I'm not sure exactly. Surely these are not hand made.

When looking into manufacturing of things I'm often amazed how much manual labor there often is involved as humans still are cheaper than construction and maintenance of machines doing complicated tasks. Also humans can be replaced simpler on failure than a machine and can simpler be adopted to varying products.

Specifically on pizza I only found this video: https://youtu.be/OMPFlbGXdFA where humans at least fix the salami slices. It doesn't show how they portion and form the dough, which I assume to be the complicated part.

Reply on news.ycombinator.com