Leonard Shapiro/ At the outset, let me say that robotics is not my field so please be as critical of my response as you need to be.
A robot would be programmed with the intelligent input of a human or a number of humans. When the robot makes decisions and then acts on these decisions based on its programming, it is acting within the parameters of its programming and therefore within the parameters and limitations of the human intelligence that it now contains. I am sure the robot can be programmed to randomly adjust some of its decisions. For example, to add a pinch more salt or whisk the egg more slowly or more quickly (which will introduce less or more air into the egg mixture). The parameters of ‘how much or how little salt’ can be pre-programmed; in other words, it would be pre-programmed not to add over a specific amount of salt relative to the number of eggs in the bowl. In this way, a robot can make a number of different variations of scrambled eggs. [As an aside, a good use of this programming would be for us as humans (and for chefs in particular) to taste a number of scrambled egg recipes made by a robot programmed in this way, and then to judge which recipe works best].
However, we still have the robot making decisions and acting within the bounds of its programming, even though it has been programmed to make variations on any of its decisions.
But we are separating robot from human as if we share no robot attributes simply because we are made entirely of flesh and bone (i.e. our bodies are entirely non-mechanical). But as soon as I introduce a machine into my existence I begin existing as an integrated unit with the machine. Even when I use a washing machine, my behavior is changed; if the washing cycle will take 15 minutes, I will have a conversation with someone based on the fact that I have to take out the washing in 15 minutes time. So, my human behavior has become partly determined by my engagement with the (washing) machine. And this is just one machine that is ‘attached’ to my consciousness and my daily decision making processes. My human mind is the ‘computer’ operating the machines in my life and my mind is in turn influenced by the abilities of the machines that I use around me. What I arguing is that we are more robotic in our programming than we acknowledge ourselves to be.
What I am doing when I drive a car is applying a number of decisions in order to operate a machine. I am using my brain like a computer. And now we have entirely non-human operated cars, which indicates that the use of my brain when driving a car can be substituted by a computerized program. Of course, a car going from A to B is just that; it is being guided within a set of predictable and very narrow parameters.
Scrambling an egg is very different from driving a car; we expect creativity/artistry to be introduced into the making of a scrambled egg recipe but not in the driving of a car which we expect to be governed within strict parameters.
Can a robot learn?
A human will make a scrambled egg and if they repeat this daily, they can learn ways on how to improve it based on what they did the previous day. They can learn how much or how little salt to add. Can a robot be programmed to learn? Could a robot learn and make adjustments based on what it has learned? Assuming that we had a robot that could be programmed to taste in the same way as a human could (and so it would know what was too salty or not salty enough), and it made a scrambled egg, would it be able to learn the quantity of salt that it needs to add so that the egg does not have too much or too little salt in it? (note that in the case of this robot, we have not programmed it to add an amount of salt within certain parameters as in the previous example).
So, I think that a robot making scrambled eggs would indeed be an art similar to a human making a scrambled egg would be an art. In fact, a well programmed robot would make better scrambled eggs than a novice human cook. And the proof of this would be in the tasting: when tasting the scrambled eggs made by a well programmed robot, a group of chefs who did a ‘blind’ tasting would surely agree that it made better scrambled eggs than a novice cook (who might have added too much salt).
Also, in the programming that I have described for our scrambled egg-making robot, doesn’t it follow that one can write an algorithm for the making of a scrambled based on the way that a particular human makes it? If so, wouldn’t it follow that one can write a number of algorithms based on a number of individuals who made scrambled eggs in their unique way? (I am asking this as a genuine question as a non-robot expert).
ps. I liked the Youtube video on the Rembrandt painting. I would be keen to see the computer paint a Jackson Pollock :). One thing that would be needed is to program in the kind of splashes made from a brush loaded with a paint of a specific viscosity and flicked at the canvas from a specific distance, or dribbled onto the canvas from a tin with a hole of a specific size in the bottom of it.