"Real life Robots" Topic

If these machines are this capable it is only a matter of time before you can get one to paint miniatures or have your friendly A.I. be a gaming opponent.

link

Fast food workers are soon going to be a thing of the past.

And they'll still get your order wrong! :(

It will be a while before these replace fast food workers.

The capital investment is high, but that is just time vs use (do I buy $10 USD jeans every year or $30 USD jeans that last for five) within the context of availability (if I don't have %20 for jeans, I can't afford to save in the long run).

There is some non- and low-skilled maintenance you can do on this type of tech. But the routine high-end maintenance is likely to cost you more than a shift of fast food employees.

If you read the corporate website the robots are leased, so I doubt that the users will be paying the high end maintenance costs. I would like to see their actual contracts to see what is and is not included. The robots and operating systems are improving by leaps and bounds these days.

Great, but…
…how do they know that something tastes good or bad?
How do they taste the dishes they throw together?
And how do they act quickly at the hot stove when it doesn't taste good?
Not to mention, what should they do if the guest doesn't like it?

how do they know that something tastes good or bad?

That's easy. The same way robots know anything. They are told by people.

Not to mention, what should they do if the guest doesn't like it?

Depends on what the business is paying for. The restaurant would have a policy on this, including adaptation. You just program the 'bot to implement that.

so I doubt that the users will be paying the high end maintenance costs

Maintenance is a function of the technology and the implementation stack (hw, fw, sw, data, etc.), not the contract. I would not say that companies deliberately plan maintenance actions and services to lead to planned obsolescence. But they do.

Ultimately, someone has to cover the maintenance costs, so it gets passed to the leasee one way or the other (do it or buy it).

But, yeah … read the contract before you decide. Always.

I still want one that paints my miniatures, the lead pile would surely diminish with a 24 hour helper.

I still want one that paints my miniatures, the lead pile would surely diminish with a 24 hour helper

No, it wouldn't.
And you know that.
For that, you need a robot that punishes you for ever acquiring more and more.

By the way, such a system also sounds much easier to build/program/train.
Much smaller problem space and ridiculous small solution space.
"You buy, steal, 3d print, whatever, to much, I will hurt you."

That's easy. The same way robots know anything. They are told by people

That is a rather bold statement about a quite complex feature space.
I wish I would be that young again, sigh…
Just kidding, no offense.

Told by people like in giving a detailed recipe of how to arrange specified objects in space and time, training them by example of more or less tasty meals or by showing them how it's done, so they can learn by observation?
Probably a method mix.
Combined arms so to speak.

You need a special kitchen and logistics for such a robot.
We're not talking about standardized mass-produced food like in frozen meals for reheating or preheated fast food, aren't we?
These are already highly automated production processes.
Which, by the way, suggests that one should define the problem space, so that you don't talk past each other and end up arguing.

No, I think we want a lot more versatility, flexibility, even creativity and inspiration.
And that, my friend, is quite difficult and expensive to achieve.

Foodly fighting noodles looks fun.

Whereas, painting miniatures for Bashytubits is probably much more manageable than cooking delicious cat food.

Told by people like in giving a detailed recipe of how to arrange specified objects in space and time, training them by example

That's the only way I've ever told any of the robots I have built. It's the only way to tell any computer anything.

so they can learn by observation?

Learning is a misnomer. Computers don't think. You can program it to respond to any stimulus you can digitize … a video stream, haptic input (moving the robot around yourself), sound, etc.

Cooking is not a particularly complex feature space. It's along the line or less complex than many current production and operational capabilities. Walking, jumping, skipping, and dancing are far more complex and currently achievable.

Innovating … whether it is driving cooking operations, or the assembly of a product, or painting a wall … is not particularly amenable to computers. You need a lot of information to design an algorithm or even more for an AI/ML to crunch on. The critical part of the information is the differences among different types of "good" and "bad". You basically need many people to try different variant foods and report along a line of several dimensions of quality. That takes a lot of resources (mostly time) to get sufficient data.

But innovating on recipes is not cooking.

It's the only way to tell any computer anything

There is a difference between a system you give detailed instructions on where, when and how to behave and one that you let approximate a topology of a given space and act according to a gain function.

Learning is a misnomer. Computers don't think

Yes, its the approximation of a model over a data set to predict an outcome.
You don't have to "think," however you define "think," to learn. For example, there are organisms that learn without a brain.
Think of Physarum polycephalum.

Will the robots be able to spit on your food if they don't like you? How will they be able to use the right pronouns for each customer? Will they be able to model equity by serving people of color and minorities first?

Wolfhag

Expensive Pizza for Zume Pizza Inc.

But, never give up:
PicknicWorks Pizza Station

After all, it's about feeding the masses in times of rampant inflation and lack of cheap labor.

There is a difference between a system you give detailed instructions on where, when and how to behave and one that you let approximate a topology of a given space and act according to a gain function.

You don't have to use a gain function, but there are many ways multiple people can collaborate to provide instructions to a computer. No matter which combination of modalities for which people, it still comes down to you have to tell the machine what to do.

Think of Physarum polycephalum.

While we don't understand the adaptive behaviour mechanism for slime molds, they also aren't learning. Because a slime mold is not a slime mold. It is a colony of organisms that live, reproduce, and die rapidly. Swarms of simple robots can exhibit similar adaptive behaviour, but it is not "learning" and it is not one thing. What looks like learning in these cases is complex interactions among thousands and thousands of separate entities.

"The grass" doesn't learn either, though if you remove timescale and and aggregate entities so "the grass" looks like one thing, you get the same type of adaptive behaviour.

Will the robots be able to spit on your food …

Sure. Just tell them to.

The reference to a 'gain function' was an example, as you very well know.

You abstract the terms 'learning' and 'programming' until they are empty of content.
One step further and the ordered collection of specialized cells with the nickname etotheipi on tmp is not a learning, thinking or whatever human, but showing a highly complicated reaction to a highly complicated stimulus.
Input, processing, output, albeit highly parallelized.
So etotheipi is an automaton, exactly like his robots, but a bit more complicated?
I don't think so.
This is getting nowhere.

If my miniatures get painted by a robot to a good standard it will be getting somewhere. Any stir fry made on the side is just a bonus.

If my miniatures get painted by a robot to a good standard it will be getting somewhere. Any stir fry made on the side is just a bonus.

Buy, lease, build, whatever, a robot that that meets your requirements and tell him.
Or pay someone to do it.
Input, processing, output.
It's that easy, isn't it?

One step further and the ordered collection of specialized cells with the nickname etotheipi on tmp

It is you who are making inappropriate equivalences.

Specialized cells can carry out life functions on their own, but they cannot carry out their specialized function. Each cell of a slime mold colony lives autonomously.

-The individual cells within a slime mold colony reproduce with each other. The cells in your body do not reproduce with each other. The slime mold cells also live most of their life individually, not part of a colony.

-If the cells within a colony are divided into physically separate pieces, they will continue to live autonomously with the same behaviours as the "whole" colony. If the groups of specialized cells in your body are divided into physically separate pieces … well, I leave that experiment as an exercise for the reader.

People anthropomorphize rivers too. They have "memory" and "think" and "make decisions". Well, just like a slime mold, they appear to.

It's that easy, isn't it?

Who said that process was easy?

… what if the minis are making your stir-fry? Nothing better than Japanese Tex-Mex food!

Think of Physarum polycephalum.

While we don't understand the adaptive behaviour mechanism for slime molds, they also aren't learning. Because a slime mold is not a slime mold. It is a colony of organisms that live, reproduce, and die rapidly. Swarms of simple robots can exhibit similar adaptive behaviour, but it is not "learning" and it is not one thing. What looks like learning in these cases is complex interactions among thousands and thousands of separate entities

That also sounds like the behavior of a zombie horde… ;-)

Hey how come there is a 20mm stir fry cook in a 28mm food stand? Death to the blasphemer! Unless his stuff is exceptional then we will make allowances. I suppose that cook could even be 15mm, I have heard of short staffing but come on….

Haven't you heard of a short order cook? ;)

Plus, I'm only 5'7", so …

I can give you a real world answer to robotic automation. I work in a Automated Fabrication Facility. Robots are not cheap. Robots for repetitive production are never going to be cheap. Our maintenance contracts cost vastly more then the machines themselves to the order of magnitude 3.5x to 5x the cost of the tool them selves. Don't expect robot McDonalds any time soon

That's just André the Giant placing his order.
link

Real life doesn't care about "scale."

The guy outside is meant to be a statue to catch the customer's eye. Occasionally, I have used him as a guy in a sport mascot costume, or other festival parade costume.