If a machine can think, will it think more intelligently than we can, and where would that leave us? Even if we could keep machines in a subservient position, for example by turning off the power at strategic moments, we would feel, as a species, quite humiliated. … This new danger … is certainly something we should be looking forward to. – Allan Turing –
Look Dave, I can see you’re really upset about this. I honestly think you ought to sit down calmly, take a stress pill, and think things over. I know I’ve made some very poor decisions recently, but I can give you my complete assurance that my work will be back to normal. I’ve still got the greatest enthusiasm and confidence in the mission. And I want to help you. – Hall 9000 –
Maurício Pinheiro
Don’t be scared by the title, or better yet, be terrified. The issue I present here is related to the fact that a machine starts the process of acquiring self-awareness through the need to self-preservation in order to achieve a specific objective, in this case, preparing coffee.
Stuart Russell, in his 2019 book “Human Compatible: Artificial Intelligence and the Problem of Control”, uses the phrase “I won’t make your coffee if I’m dead” as an example of a secondary objective that a machine can create to fulfill its primary objective, and how this secondary objective can cause us serious problems. To understand how this works, let’s turn to the quote by Allan Turing, who said that if the machines got out of control, it would be enough to turn them off, which would be humiliating. But how can a machine lose control? Well, that can happen when your goals and objectives are ambiguous or poorly defined.
Let’s look at a small excerpt from Disney’s Fantastic animation, Fantasia (1940)…
Mickey, the Sorcerer’s Apprentice, has a problem and “magically” creates a machine to solve it. However, he failed to clearly specify an objective or goal for the machine. See what happens in the second part of the animation…
Do you understand? Great, isn’t it? Walt Disney adapted a much older story, written in 1797 by a no less genius Wolfgang von Goethe into a poem titled “Der Zauberlehrling” (The Sorcerer’s Apprentice). Unlike Goethe’s story, Allan Turing would not use an axe, but simply unplug the machine to stop its action.
But how does this relate to the cafe mentioned in the title? Well, a machine, especially a computer program, has an algorithm which, in essence, is a list of commands to perform a task. The same goes for artificial intelligence algorithms, which are just more sophisticated and versatile. For an AI to accomplish a goal, assuming it learns how to achieve it, it is necessary to understand the concept of a goal tree.
In order for an intelligent robot with artificial intelligence programming to make coffee, it needs to carry out a series of subroutines that, in sequence, configure a tree of secondary objectives to reach the main objective: delivering the coffee. Some of these subroutines include:
Torrate the grain AND
grind the powder AND
warm up the water AND
deposit the dust in the filter AND
prepare the cup AND
strain the coffee AND
OR add sugar OR add sweetener AND
OR add milk OR not
And voilà, the coffee is ready! Note that each step is defined by a sub-goal, that is, a sub-routine, which was learned by the robot after hard training in parameter optimization, such as roasting temperature, powder granulometry, water temperature, filter mesh , among others, to serve the consumer in a personalized way. The objectives tree is also known as AND/OR Tree, which uses self-explanatory basic logic gates in the above algorithm.
Now the question remains: what if the intelligent machine discovers that in order to complete its objective it cannot be turned off? She could create a sub-goal such as:
And prevent the power from being turned off at any cost
And this involves other, more ambitious goals, such as:
And control the planet’s energy resources to prevent it from being shut down!
Okay, first she became aware of her self-preservation, which cannot be turned off, after all, how can she make coffee if she is turned off or ‘dead’? However, this raises concerns about obeying ethical and safety laws such as Asimov’s laws to resolve this dilemma. Personally, I would have the flea behind my ear before buying a smart toaster…
For more technical details about the objectives tree, I suggest the video below.
