Artificial
Intelligence can unlock the door to an improved user experience, every time you
step into the car
By Shadi Mere
Google’s
search engine has become famous for delivering a user experience that is
personal, intuitive, simple and indispensable. In fact, for most consumers, the
term “Googling” has become synonymous with performing a Web search. You may not
realize it, but when you do a Google search, you’re tapping into artificial intelligence
(AI) machine learning, cognitive science research applications and behavioral
psychology – supported by a large investment in big data mining.
Google
is not alone. An increasing number of companies – particularly those that sell
portable consumer devices – realize that if their products don’t incorporate
some form of artificial intelligence, consumers will perceive them as outdated.
This perception has negative implications for their brand image – and auto
manufacturers are not immune.
Leading
research is uncovering the mechanism by which a brain forms a memory or learns a
skill. The brain relies on clusters of neurons that hold bits of information.
With synapses firing to firm the connection between these clusters, the
frequency and the chemical incentive (ex: dopamine for pleasure or adrenaline
for fear) solidifies the connection. Furthermore, we don’t form memories as
photographic impressions, but rather as a loose recollection of memories that we
detail using imagination, logic and our current emotional state, remembering
differently each time.
In artificial intelligence,
neural networks are represented to recognize complex patterns, such as an image
or even mood. These networks are structured with layers of connected nodes that
apply calculations. The nodes and its layers go through learning and
training processes through iterative inputs of patterns to categorize
information and behaviors. Once the network is fully trained, it becomes
capable of instantaneously recognizing patterns and profiles with a large
amount of complex data at a speed not possible by humans. Examples include:
Google (and other Web) searches, spam filtering, speech recognition, robotics, medical
diagnostic systems, and many popular recommendation algorithms (such as those
employed by YouTube, Netflix and Amazon).
The HABIT cockpit concept
The
Visteon HABIT cockpit concept applies this artificial intelligence approach to the
user interface, embodying realistic 3-D graphics and animation to deliver a
futuristic vision of human machine interaction (HMI) in the car. The demonstration
boldly moves the user experience in a novel direction: an AI logic engine with an
evolving learning system that factors short- and long-term memories. This is not
an attempt to create a static procedural cycling of use-case routines; instead it’s
an AI learning system that adapts with each new input.
The goal of HABIT is to deliver an experience that
improves each time the driver uses the ever-aware system. After days, weeks and
years of owning this system it is likely that the driver will be disappointed
in any other system that does not get to know him or her like the HABIT system
does. Additionally, the driver
will be able to transfer his or her personalized system of secured knowledge and
learning anytime he or she rents a car, shares a car, or upgrades to a new car.
Finally, the system will always be relevant,
upgrading organically over time to reflect the latest in state-of-the-art AI
algorithms, and the latest in our habits.
Our approach at Visteon starts with the Consumer Experience Model. This model is
derived from extensive consumer research and is built around the notion of
creating the ideal experience for drivers. As opposed to employing “technology
for technology’s sake,” Visteon takes great care to ensure that any new
technology is vetted against the Consumer
Experience Model before it makes its way into a new product concept.
Successful organizations do not shoehorn technologies and ignore its shortcomings.
This is why our clinical studies are crucial in shaping our concepts. Research
during the testing of HABIT showed that consumers liked the concept and rated
it high; however, they did have a few concerns and dislikes. The main issues were around “Big Brother” and the privacy of
their data. The consumers also did not care much for an Avatar “Infotendant” and
they preferred to hear the infotendant’s voice but not see the image, which
they viewed as distracting.
They also expressed dislike for “forced” habits, in
other words, performing actions without first asking the consumer. This dynamic
was particularly interesting because in general people like to think of
themselves as unique and unpredictable; with changing tastes in music and
habits. Most research contradicts this notion, showing that consumers are
generally much more predictable then they like to believe. This creates an
interesting paradox between what consumers believe to be true and reality.
Perhaps over time, people’s negative perceptions of being predictable will be
outweighed by the convenience of an interface that is much easier to operate.
In light of
our consumer research findings, the HABIT concept changed to reflect a few
important improvements: Examples include:
- Personal data and profiles are
secured and can be locked to the person’s phone, Cloud account or their
personal voice identification print. Alternatively, the system can be
switched to an agonistic identification mode altogether.
- The visual infotendant was
eliminated with interactions limited to voice only.
- The system user interface (UI)
looks similar to a typical UI. Visually it does not behave differently
based on habits; it only shows faint “bread
crumbs” highlighting a path to the default predicted habit. The UI
also prompts the user to engage its recommendations - they are never
forced. Alternatively, the AI can be turned off altogether so that the
system can revert to a more traditional operating mode.
- The approach starts simply and without a learning curve. However, it is possible for the user to customize the system AI settings to a very deep level, satisfying both the passive user and the technophile.
Moving forward, the ongoing challenge for this technology will be creating a system that is dynamically learning, but at the same time does not cross the line into being intrusive or annoying. The AI is dogmatic, yet it must be simple to perceive, while being accurate and robust. Ultimately the goal is to create a system that makes us bond with it as it bonds with our habits.
The Visteon HABIT Cockpit Concept
The Future:
Multiple Intelligence
If an area of the brain that processes sight is totally damaged, can we
see again? What does the above question have to do with the car?
Research indicates
that, contrary to long-held beliefs, areas of the brain are not rigid centers
for specific tasks, but can evolve to take on sensory skills that were thought
to exist strictly in specialized areas. Given enough training and stimulation,
severely damaged areas in the brain that process sight can be delegated to
other areas of the brain – like the parts that process hearing. Similarly, one
day your car’s “brain” might be able to compensate for a damaged sensor by
having a different sensor (with a different purpose) take on a quick AI
learning pattern to process the critical information. There are also broad
implications of how our brains handle multi-tasking, automated tasks,
concentration and distraction. Our brains can adapt to certain tasks,
experiences or even a new vehicle (HMI), but they have difficulty with true multitasking.
(Instead, we have become very good at “task switching.”) AI, on the other hand,
can multitask. For example, in the
future one can imagine scenarios in which cars can talk to one another and
simultaneously apply learning to blind spots, weather conditions, traffic
rules, impaired drivers and complex driving patterns – amounts of information that
humans simply can’t process quickly enough to make the best decision. Eventually,
AI will be able to adapt to perform tasks that (today) we consider to be
uniquely human.
At some
point, artificial intelligence will be ubiquitous, making our lives much more
productive in both expected and unexpected ways. This technology could be seen
in vehicles in as soon as five years from now.
Shadi Mere is an innovation
manager with Visteon’s “Innovation Works” team. He works on advanced innovation,
“disruptive” technology, human-machine interaction,
creative design management, consumer experience research and high-technology
trends. During his 17-year career, Shadi has
worked in engineering design, advanced manufacturing, product development and
strategy, and program management -- with a focus on bringing promising
inventions to life.
