User:Manetta/serving-simulations/the-human-simulation-as-interface
serving simulations: the human-simulation as interface
[questions]
How is artificial intelligence mainly an interface?
What are the elements one can describe of this interface?
What could be the functions/purposes of these elements?
[underground questions]
How does the computer interpret human speech?
These days we have Siri's, Cortana's, Google Now's and Echo's produced by Apple, Microsoft, Google and Amazon. It are pieces of software promoted as being the personal assistences of the future, that mostly react on commands that you give it through your voice. "It's a feature all about our voice. (…) It helps you get things done just by asking."[1] They all interact with their users without much physical interaction. The main part of this software is build around one incoming medium of information: the human voice.
human simulation
[what - human simualtion]
Is it really "man's dream since the birth of science" to "create an artificial being", as Steven Spielberg stated in the introduction of his film A.I.? Technology brought us plenty of different machines already, that replaced work done by humans: a porter is replaced by automatic doors, a baker uses a machine to produce many breads at once, and as we're having thermostats we don't need to put coles on the stoves to keep our homes warm. The door opens automatically when one comes close, the bakingmachine has a (simple) set of buttons, and the thermostat follows the number of degrees that the little slider is pointing at.
But where is aimed for when this work is not only been replaced by a machine, but by a machine that asks one to communicate with through a human simulation? How do we interact with software that is designed to be mimicing its user? What is the purpose behind the choise to create an human simulation as interface? Is this a gesture of user-friendliness? In order to let the user interact with the interface as smoothly as possible?
Or is that gesture rather an act of user-un-friendliness?
[camouflage not to hide, but to come close]
If software is being designed in order to be as 'smoothly' as possible, it brings up a similarity to the tranforming abilities a chameleon naturally has. Chameleons "have the ability to change colours" which "functions in social signaling and in reactions to temperature and other conditions, as well as in camouflage."[2]
This highly flexibility to adapt to other systems, increases the safeness of the chameleon. It is a hiding technique for possible enemies. The chameleon and its enemy are two beings that are playing a hunting game, the one hunting the other.
The AI machines don't have a need to hide, do they? Its users are not a danger for them, they rather are their complementary party they need in order to have a use anyway. Their act of camouflaging is rather a rapprochement then an act of escaping. They are rather close then far away. By coming as near as possible to the nature of humans, AI machines try to become invisible. An act of naturalizing.
[non-abstract system]
A keyboard. We all know it as the most common writing tool nowadays. This human to hardware interface offers one a very static interface, where the buttons are either 'on' or 'off'. The keyboard is built according to the size of our hands, and the length of our fingers. The position of the letters on the keyboard is derived from the specific language one chooses to type in: English typers do have the 'Q' as the upper left key, where French typers placed the 'A' on that position. Although the system is designed in a way to make typing on a keyboard as efficient as possible, the system is not mimicing an human hand in order to write. The keyboard offers the typer an abstract system, where one can get used to over time.
An AI interface is not such an abstract system as a keyboard is. It is by nature more flexible, and for the user more naturalized. Communicating through an AI interface doesn't cost much effort and time. And learning how to use the system is minimalised to one instruction: "just ask".
human intelligence & empathy
[functions of an AI simulation: the human-simulation as metaphor]
The AI interfaces are build as simulations. They are designed with an attempt to reach a level of intelligence that comes as close as possible to the complexity of human intelligence. The more 'human' an interface seems to be, the more 'intelligent' we name it. This leads to a marketing dialectic of making something that is 'smart', which actually means 'as human-like as possible'. And hence we now live with a range of such 'intelligent' dogged attempts: the Siri's, the Cortana's and the Echo's.
The denominator 'artificial intelligence' itself is already containing a metaphor. By valuating the accomplishments of a computer system with the word 'intelligent', is already using an identifier that we normally use to valuate human's capacity to have e.g. a good memory, a certain level of learning abilities, and skills for problem solving. By using the metaphor 'intelligent' for a computer system, one is trying to understand a complex process by comparing it to human characteristics.
Aren't metaphors following the intention of the one who uses them?
The metaphor of 'being human' as is present in the AI interfaces, seems to not only simplify the system behind it, but has another purpose as well. By interacting with an interface that feels like it is an human, raises a certain level of empathy.
Wikipedia describes the term as: "Empathy is the capacity to understand what another person is experiencing from within the other person's frame of reference, ie, the capacity to place oneself in another's shoes."[4]
The human-simulation can be seen as a form to understand and describe problems and solutions one runs into while using the software. But because the software is an 'human', it causes reactions of diappointment, impatience, or even anger. But also reactions of surpriseness, or astonishment.
Next to that, the human-simulation interface also knows a low level of intelligibility. As it operates with the human vocabulary, a tool that the user is pretty comfortable with already. There is no need for learning another language anymore.
It's this aspect of comfort that raises a whole set of possibilities according to Apple: it's making the users eyes free, so it can drive, text, call, search, and being a king in multitasking. It makes "everyday tasks less tasking."[3] Because: "All you have to do is ask."[3] And it's always on, which makes it a pretty reliable feature, one who's always there for its user.
It is exactly this comfort, and this smoothness that makes it possible for the interface to be adapted easily by its users. The system is already pretty human-like from out of the box, but it is evolving while using it. It then seems to 'get to know you', which is another metaphor to use an human characteristic. As the AI system manages to answer the expectations of its user, it's also establishing a certain amount of trust (another human metaphor). This all will lead to a smooth attendance of AI software in the user's life.
...
And after a while the user starts to appreciate that the AI software is delivering information automagically[5]. Getting it delivered just before knowing where it is looking for. "Siri is proactive, so it will question you until it finds what you’re looking for."[3]
[1] Apple's Keynote - presenting Siri, 2011
[2] https://en.wikipedia.org/wiki/Chameleon/
[3] https://www.apple.com/ios/siri/
[4] https://en.wikipedia.org/wiki/Empathy
[5] term used by Michael Murtaugh in February 2015 at the PZI, Rotterdam
conclusion
Remembering many tasks at once, finding information about ones current location, or informing about the latest news. These are all properties that a human being is incapable of doing. And it is this lack that makes a human looking AI interface look intelligent. It's faster in looking into Wikipedia to search for simple facts. But isn't that a kind of task where computers are built upon?
By using metaphors, the AI interfaces are camouflaged of their original type of being machines. This camouflaging technique brings them close to their users, so close that they become almost invisible. So close that it is easy to forget about them. How much does one trust on machines? Its user-friendliness expanses, untill the machine is almost not present anymore. The AI's are trying to be part of the human natural surroundings, through speech, 'realtime' information deliveries, and an increases amount of knowledge about one's contacts and behavior.