Page 118 - Artificial Intelligence in the Age of Neural Networks and Brain Computing

106 CHAPTER 5 From Synapses to Ephapsis

augmenting existing human intelligence and memory. Given that the field approach
scales easily to larger spatial levels, integrating social networks with such assistants
is almost frictionless.
When two or more sensors are correlated in real time, their power increases
exponentially. There are no commercially available multimodal applications today
but university researchers such as Alex Pentland at MIT, Dana Ballard at University
of Texas, or researchers at MSR Cambridge produced experimental working proto-
types delivering promising results in a laboratory environment. I believe that the
next generation of personal assistants will augment text-based user profiles with
multimodal and motor data from our hands, arms, legs, feet, torso, neck, head, as
well as internal organs such as heart and lungs.
The key to usable personal assistants lies in the embodied approach, grounded in
neural science. It will start with a minimally acceptable amount of body sensors in
multiple modalities, eventually reaching critical mass, resulting in a phase shift and
the emergence of new chaotic attractors augmenting the existing landscape of users’
knowledge. Neurally based personal assistants will thus build this knowledge land-
scape and, at the same time, travel through it in itinerant trajectories formed by the
user experience. There are numerous applications for such technology:

Augmented Memory. Human memory is fallible, and it gets worse as we age.
Anyone over forty was probably in a situation when he would appreciate quick
help in attaching a name to a face, movie or a context to a meeting he attended
some time ago. Over time, the personal assistant will detect user’s intentions
and become his trusted advisor, able to assist without prompting, such as to
remind him to take a medicine at the usual time or suggest a restaurant menu
selection based on his likes and recorded allergies. A personal assistant can
take on a different role as the user ages and experiences greater memory loss.
As the personal assistant learns the user personality better, it can offer more
active guidance for the user throughout the day.
Stress Relief. Stress relief can be accomplished by relaxation techniques and
meditation. Both rely on the entrainment between neural activity and body,
including breathing lungs and beating heart. Embodied assistants can integrate
motor and EEG sensors with these techniques. First, motor and EMG sensors
will be used to provide additional user feedback during meditation, which
should minimize the tendency of EMG artifacts to pollute EEG readings.
Second, controlled motor activities, such as breathing or simple motor exercise
(e.g., Qi-Gong) can drive the user to the desired state directly by combining
EEG and motor feedback into a single solution.
Health and Fitness. While most wearable sensors today are focusing on health
and fitness, they are hardwired to a specific activity such as walking or
swimming. Neurally based assistants capable of capturing the meaning of user
action and his intentions can provide advice and real-time monitoring of any
fitness activity, from a simple exercise to meditation and suggest dietary regime
optimal for users’ current intent and his long-term goals.