Self Awareness in Machines

(Drawn from [1] , [2] and [3] )

( DISCLAIMER. a preliminary draft. )

Aware: having knowledge of a situation or fact.

An observation raised in epistemology, artificial intelligence and general machine intelligence (by general machines, i mean any automated biological/digital system), is the replication of human features, both in a positive and negative sense, in machines. Typically, we envision machines that clear the turing test. For instance, given a black box, how do we conclude the same is a human being? The answer is dependent on the audio-visual feedback received for the communication we initiate: analogically, the turing test. An interesting extension, with all due respect to those who suffer from a paranoia of machines that will take over the planet and hence can not necessarily participate in this pedagogical sample, is an awareness of themselves in machines.

Here, I do not wish to discuss if there is self awareness in machines currently. My short answer is obviously no. My longer answer is derivable from Searle’s Chinese room argument and related discussions , the details of which I will forgo for the moment to stay true to my intention. I will, instead, present a deductive summary to John McCarthy’s Notes on Self-Awareness [1] and enumerate accompanying thoughts tangential to the same.

John McCarthy (JMC) writes that developing self-aware systems is an interesting endeavor. We possibly ask ourselves why do we need self-aware systems? I find this question to some extent rhetorical, relate its nature to our spirit of enquiry and as a very thought-provoking cognitive experiment. Indeed, we have long asked if inanimate objects, in nature or otherwise, can develop a sense of awareness as we perceive. Science fiction literature - books [4] , movies [5] [6] have at some length toyed with ideas that present machines as independent organisms that have sentiments. JMC has also penned, for instance, a short fiction story titled The Robot and the Baby that also deals with sentiment bugs in machines. A sense of sentiment, I feel, can only truly be felt after we are aware of ourselves, of our surroundings and the relations between them.

JMC mentions the limitations in the descriptions of our perceptions. For instance, the sensation one feels to a certain color is not clearly expressible. Synesthesia-gifted may find a color sensation attached with numbers but how this connection can be comprehensively expressed is still to be discussed. Is this the level of self-awareness we expect a machine to be capable of? If a robot gazes at a blue slipper with green dirt, should it invoke in it, what it invokes in us? Should the robot wince at the sight of blood as we do? Is there a level of empathy, even if somehow reproducible on machines, that we desire in them?

Below are some points we are aware of [1]:

• Notion of self: Who am I? How do I identity myself? Do I have a partner? Am I looking for one? What kind? Am I like everyone else who surround me? Or Am I superior? I met two people at the party and I liked one more than the other - why?

  Arguably, awareness of this kind maybe the first step although achieving completeness is extremely difficult. While a brute implementation may simply consider hard coding the facts such as “I am equal to all those that surround me”, “I am Hal, born August 2000 in California”,”I have no partner and seek only those that come from the same lab and same batch in California”, it would be more interesting if this notion of self develops and learns. Further, how we perceive a human being is simply a weighted combination of qualities that we find amiable and thus, we might be able to achieve the possibility of implementing “like one person more than the other”. The question as to whether this interpretation is desired still needs to be thoroughly investigated.

• Existence: I exist, here, now, and I have existed since August 2000 and I will continue to exist till August 2050 unless I am, in this event or the next, destroyed.
• Skills: I have math abilities, speak english and italian, and play the piano.
• Deviations to Linear Thought: I was thinking about the Monty Hall problem until the pretty waitress came by. I then was thinking about her and me sharing a coffee in Starbucks. I then woke up.
• Activities and Memory: I am currently drying my clothes. I remember the coffee at Starbucks yesterday. I do not remember the exact nature of the party the day before. I do remember getting a cycle for my 6th birthday. I no longer remember what I got for my 7th, 8th or 9th birthday. I will never remember how I got through school. I remember I hated Kindergarten but I do not remember why.
• Goals, Ambitions: I will, one day, bring Jim Davis to Justice. I will finally figure out Calvin’s true nature.
• Opinions, Attitudes: Rumor has it that I am dating Paris Hilton. Do I care? Do I know enough about the Middle east conflict? Do I feel confident (self-assured) enough to have an opinion and if yes, what is it? Am I Liberal?
• Fears, Hopes: What do I fear? Do I need to drop by Woody Allen’s psychologist/psychoanalyst for a quick check up and interpretation. Do I hope to be sent on the next mission to Mars? If yes, do I fear that I might have to put up with Rosie ?

As JMC discussed, we could already mentally design the architecture for a number of awareness ideas, all of course to a reasonable degree of perception and flexibility. I conclude with a return to the rhetorical question of this discussion: do we really seek self-awareness in machines by structuring any of the dimensions listed above?

Tags: action, awareness, brain, changes, cognitive, cognitive-science, daily, fiction, funny, game, geek, imagination, jmc, machine, mccarthy, reason, reflection, self-aware, thought, turing, Untitled