Friday, July 24, 2015

What does it mean to have a body: AI with minds and bodies

2015 seems to be the year of discussing Artificial Intelligence (AI).  From philosophers like Nick Bostrom to scientists like Steven Hawking, many high-profile people have joined the conversation about the possibilities of AI research.  Lately, it was Facebook’s CEO Mark Zuckerberg who made headlines when he hosted an online Q&A open to the public on his Facebook page, where he talked, among many things, about exploring AI.

Of course, the perils of AI always arise in these conversations, as sci-fi flicks like Terminator illustrate through a terrifying scenario: instead of serving us, intelligent machines enslave humanity.  Interestingly, Arnold Schwarzenegger joined the online Q&A to ask Zuckerberg about exercise, but he also asked if machines would eventually "win."  Zuckerberg lightheartedly replied no.

(Image from Mark Zuckerberg's Facebook page)

Putting aside quibbles about the rise of the machines, a fascinating possibility of AI was raised in Zuckerberg’s Q&A: computers with vision, which could see and describe visual images such as pictures and videos.  That possibility would not only be AI that ‘thinks’ intelligently—by processing information and presenting solutions to problems.  It would also be AI that ‘perceives’ the world—through visual sense.  In other words, this AI would have mental abilities and bodily senses.

(Image from Mark Zuckerberg's Facebook page)

Of course, as philosophers like William James and neuroscientists like Antonio Damasio have argued, mind and body are inseparable.  So mind-body continuity should apply to AI too.  If we want AI with a fully conscious, intelligent mind (what's called Strong AI), then it needs a living body.  Cognitive scientists refer to this idea as embodied mind theory.

The Kismet robot
(Image from Kismet's Official Website)

The Cog humanoid
(Image from Cog's Official Website)

In fact, AI research has already taken insights from embodied mind theory to design robots.  You may recall Cog, a humanoid that could interact with people and learn from environmental stimuli.  Perhaps you saw Kismet, a robobic head with visual and audio systems that could recognize and express emotions (using an approach called affective computing).

Now here’s an interesting question: What makes these AI bodies different from the ‘bodies’ of other technologies.  For example, doesn’t a smart phone have a body around its memory card?  Doesn’t a computer have a body around its silicon chips?  Doesn’t most AI already have a body anyway?

Here’s an answer: There’s a difference between a living body and an outer covering.  Just because smartphones or computers possess outer coverings doesn’t mean they have living bodies.

Consider what living bodies have that outer coverings don’t:
  • Emotion: Bodies generate emotions felt by the brain, and emotions provide feedback to evaluate thoughts.  I read the news, a thought occurs, my body produces an emotion, and my brain feels it and realizes how the news makes me feel—e.g., happy, sad, angry.  (Neuroscientists call emotions somatic markers—from the Greek word soma meaning ‘body’—because they’re how your body ‘marks’ your thoughts with positive or negative valence.)
  • Self-Organization: Unlike phones, computers, or manufactured products, your body self-organizes itself—using genetic material and environmental resources.  (Systems theorists refer to self-organizing bodies as autopoietic—meaning ‘self-creating’—because they organize themselves; phones and computers are heteropoietic—‘other created’—because they're organized by outside intelligence.)
  • Meaning Making: Phones and computers run algorithms without understanding the meaning of data.  Bodies, with their emotional value and self-organizing capacities, create meaning—a point obvious to artists such as James Joyce or D. H. Lawrence, to name two of my favorites.  (Cognitive scientists refer to this meaning-making process as sense-making.)
  • (Image from Amazon)

  • Multidimentionality: A final point, which philosopher Mark Johnson makes in The Meaning of the Body.  The body isn't merely physical.  It’s "multidimensional," with a biological dimension (body schema), an ecological dimension (environmental affordances of the organism), a phenomenological/subjective dimension (body image), a social dimension (body politic/corporate entity), and a cultural dimension (gender, ethnicity, etc.).

Cog and Kismet had some characteristics of living bodies, such as rudimentary (or maybe just quasi-) emotion and meaning-making.  I wonder how much longer till AI incorporates other characteristics that make living bodies with conscious minds.  If recent science fiction in cinema gives us any clue, future AI may be creating not sentient computers like HAL but cyborgs like Terminator, Ava, or Vision.

Terminator in Terminator Genisys
(Image from IMDb)

Eva from Ex Machina
(Image from IMDb)

Vision from Avengers: Age of Ultron
(Image from Wikipedia)