> Robots aren't going to be cooking anyone a six course meal or directing the next Star Wars movie any time soon
Of course they won't be directing Star Wars - although a case might be made that that would be preferable - but I don't see any reason almost all cooking will not be done robotically 10 years from now. Hell, try and buy a non-sous-vide steak these days. That's a dramatic step towards automatic cooking right there.
I would not be at all surprised if most of the creative work of filmmaking isn't automated eventually, once you can generate photorealistic virtual actors and sets, and algorithmically generate plots, camera work, dialogue etc. As formulaic as popular movies tend to be nowadays, would anyone notice the difference? Just imagine having the same server cluster somewhere crank out the movie, its sequels, and the video games from the same models. Digital actors never get old, never get arrested, and never need to get paid.
Also there's no reason at all menus and courses can't be automated as well. Everything from the design of the menus to the design of the building, to its construction, and staffing, everything but the eating can be automated somehow. I think people are going to come to a shocking, ego-blowing realization that, yes, creative work can and will be automated as well.
Well, we've already had glimpses of this, with varying degrees of disruptions. CGI generated still pictures don't seem to have any real success except in some very niche communities. Video games are certainly successful with it although still dominated by human level design. Partially generated photo effects and CGI are enormously popular, but suffer much blow back and derision from many.
Current AI algorithms are limited in the sense that they appeal to the common denominator. These algorithms can only observe and measure what humans like and iterate from there. Current AI algorithms couldn't have invented Jazz.
Also, digital actors may never get old, but most people don't want to fuck digital actors. And have you noticed that for some actors, getting old is part of their charm?
And most importantly, algorithmically generating plots is an AI-hard problem - http://en.wikipedia.org/wiki/AI-complete - given our current status quo of machines that can't even understand what we are telling them, well, good luck with that.
I think you're over complicating the problem. Simply put all the recipes and all the movie plots ever published in a database and randomly generate combinations. This is effectively what most "new" food and movies are nowadays.
We eat mostly the same food recipes, but there is a lot of variation in how that food tastes, depending on who makes it, not to mention the mistakes happening in the preparation process. That's also why you can't find in a restaurant the familiar taste of the home-cooked meals that your mom or grandma used to cook while you were a child. Can you automate a fast-food chain, like a McDonalds? Of course, but that's not food.
Plus, one of the reasons for why many of us like to try out restaurants is to get some variety and even though the database of all known food recipes is limited, the combinations possible for sweetness, sourness, saltiness, bitterness, and umami, as received by our tastes buds in various proportions is infinite. Therefore even though we have plenty of food recipes already, there's always room for more.
Your impression that movie plots can be generated with random combinations stems from the recent wave of Hollywood movies filled with special effects and dumb plots. I personally like movies with special effects and dumb plots. It's mindless fun. But sooner or later I'll get tired of it, as will many people. Comedy and drama on the other hand need originality and while you can find traces of inspiration everywhere, every successful book or movie ever made also have original pieces in them.
You're also oversimplifying the technical challenges, as if random combinations of available plots is easy. It's not. This is not like playing chess. For this to work you need (1) a way to generate something coherent, because as much fun as it is to generate gibberish with Markov chains, the step from here to generating something meaningful is quite big and (2) a way to measure the fitness of a combination and drop the combinations that don't work. What are you going to do? Expose people posing as lab-rats to millions of combinations and measure their brain activity?
(yes, I'm the type of engineer that rains on people's dreams, by thinking of the technical challenges involved :-P)
>>Comedy and drama on the other hand need originality
While i believe at some point computers will have those abilities, i believe that crowdsourcing(maybe with a better search engine based on natural language understanding) can go a long way for those tasks, especially joke generation.
>What are you going to do? Expose people posing as lab-rats to millions of combinations and measure their brain activity?
Basically, yes.
Imagine there are no more theaters, because they're huge, expensive money-pits and everyone watches movies on a device anyway. Because everything is distributed on the net, and the net is everywhere people can be exposed to any number of ideas, concepts, and prompts throughout their day. Gestural devices have become, if not ubiquitous, then quite common. Every phone and multimedia device has a camera on it which can read your emotions by measuring your skin's electrical charge or by measuring the subtle shift in your pulse. Everything watches you, notes you. Huge masses of correlated data about you, your friends, your likes and dislikes, even your emotional state over time and space, are being gathered from any number of companies and state agencies. There are attempts to virtually model your behavior and expectations in any number of locations, and these models are constantly refined over time until they begin to become predictive. Most of this already exists in some form or another, if only as prototypes.
So that part at I think is plausible. Or at least concievable. We've been building a vast and intricate Skinner Box out of the internet for years, it just hasn't become that obvious yet. Wait until the real quasi-transhuman augmented reality BS starts to get pushed. Wait until your appliances have their own social media accounts and actually have conversations with your friends. Or until you can have a digital analogue of yourself inserted as a background character into a film and follow it around like a Sim. Or pay to have yourself inserted into an Alternate Reality game, the plot of which is essentially generated on the fly. It just takes a couple of generations for this sort of fine-grain observation and feedback loop to become conditioned as being acceptable and commonplace. It may require redefining terms. I'm reminded of what the Architect said in the Matrix -- that there were "levels of survival" the Machines were willing to accept. What passes as a "blockbuster" or even a "movie" in the future may not bear any resemblance to what we know.
A system more complex than the algorithms currently available will doubtless be necessary, and maybe some fields like comedy will prove elusive, but the addition of the vast amount of data available through the internet and a global, always-on surveillance system may make machine learning more feasible. Your example of McDonald's is apt. They could make better food by hiring more competent cooks, using better equipment, and serving a better menu. But, they nonetheless make an insane amount of money paying people almost nothing to put garbage into styrofoam boxes, and that's only until they can make even more money replacing the people with garbage-into-styrofoam-boxes-putting robots. Quality isn't paramount when your market is billions of consumers who have the attention span of gnats. They can be taught to eat shit and like it on a massive scale, that's already been proven.
Which makes more money? That's the one i'm talking about. My point is that the only difference between replacing the physical labor of the auto factory worker and the intellectual labor of the programmer, designer or artist is just a matter of complexity. Yes, it may be a difficult, apparently "hard AI" problem, but so was textile manufacturing in a way until the Jacquard Loom came about. If one doesn't need human labor to build a car, or weave complex patterns into a rug, one may not need human labor to make a movie, write a book, or compose a song. It doesn't have to be genius, when most commercial art is merely sufficient anyway.
If chefs just have to design a menu once and then have robots actually cook it, the number of chefs in employment is going to fall pretty dramatically: I'd estimate 30-fold at least.
In a restaurant kitchen, generally only one or maybe two people are actually involved in designing the menu, out of a significantly larger number who do the preparation. And unless you're running a very unusual, top-end restaurant, that's a matter of a couple of months' work every year, at absolute maximum - closer to a couple of weeks a year for most normal restaurants, I believe.
Of course, some cooking tasks are going to be harder than others. I wouldn't want to bet on a top-flight pastry chef getting replaced any time soon, for example. But creeping automation in the kitchen is definitely a possibility: sous-vide techniques can already make things like preparing a hollandaise considerably quicker and simpler, for example. And baristas are staring down the barrel of robot barista workers right now.
> Hell, try and buy a non-sous-vide steak these days.
Obviously you do not live in the UK, where finding places that serves steak that has not been chargrilled to to the point where it tastes of lighter fluid and coal is an extreme sport (it is getting slightly better, but you can go into almost any low end steak house in the US and find steaks that are better than most high end UK restaurants still; the UK is the only place in the world where I've had to explain to a waiter what "rare" meant)
Of course they won't be directing Star Wars - although a case might be made that that would be preferable - but I don't see any reason almost all cooking will not be done robotically 10 years from now. Hell, try and buy a non-sous-vide steak these days. That's a dramatic step towards automatic cooking right there.