Dell World sure was different this year because it was done
virtually. One of the most interesting parts of Dell World is the session on
the future. In past years they surfaced the coming of the Fourth Industrial
Revolution and the coming wave of robotics.
This year, they spoke on a new branch of engineering that is
solely AI-focused, the blended technology revolution surrounding food
production, how AIs were intentionally corrupted, and how music, math, and the
Internet create new entertainment types.
Let's talk about that this week, and I'll close with a product of the week
that isn't quite a product yet, a technology that will transform your earbuds
into smart earbuds and potentially give you superpowers.
Genevieve Bell: The 3A Institute on AI
Genevieve Bell is one of my favorite people and an Intel Senior Fellow
working out of their research organization. She opened this segment on the
future talking about a new effort she spearheads that pulls together many very
different skills and people to create a new engineering-centric branch solely
focused on artificial intelligence. I say engineering-centric because the folks
involved come from a broad spectrum of backgrounds and skills that include
engineers as well as anthropologists, scientists, policy experts, and even
musicians.
Unlike most other efforts that seem to focus more on the
technology, Genevieve's begins with studies on the history of
cybernetics going
back to the conferences in the 1940s and 1950s that initially defined the
space. For instance, in Australia, where the effort is based, they studied the
aboriginal fish traps which were useful in the early parts of the last century
and successfully performed for centuries before that for guidance.
This course of study focuses more on asking questions than
on problem solving. Often, jumping into problem solving forces people into a
tactical approach without appreciating either the problem's full scope or the related
dependencies. It can conceal potential collateral damage from the resulting solution.
Focusing on questions first helps assure the problem is fully defined, creating
a pathway to a more comprehensive and arguably safer solution.
The six core questions they ask about their AI efforts are:
Autonomy: Or whether the system can perform without
user intervention. Does it automatically do what is intended?
Agency: Whether the output of the system is constrained
to an area of defined performance. Does it just do what is intended, or will it
exceed its intended parameters to do unintended things?
Assurance: This is a quality metric assuring the result
is secure, safe, you can trust it, it complies with the laws that surround it,
it is well regulated, and it does what was intended.
Interfaces: How does the AI communicate and interact
with the world, people, and other systems? Does it interrelate and communicate
both effectively and optimally?
Indicators: How do you monitor the thing? What is the
external mechanism to ensure the system doesn't go rogue?
Intent: What is, was the intent of the designers, what
did they want to accomplish, and does the result comply with the intent of
those that created it?
(It strikes me that if the folks that created Skynet had
followed these elements, the Terminator movies would have been very different.)
This group launched a master's program and currently students from all
over the world have enrolled in it. In the end, it is this effort, and those
like it, that will help assure that the AIs of tomorrow don't make
"Terminator" unfortunately prophetic.
0 Comments