Artificial intelligence. We've been reading and watching
science fiction with walking, talking robots for nearly a century. Researchers
have been tinkering with it for decades. Have we come any closer to android
production factories? Not quite. But the CALO project, under the direction of SRI International, is looking at making
headway in basic intelligence for widely used computer software.
CALO, or Cognitive Assistant that Learns and Organizes, is a very ambitious
collaboration between more than twenty different organizations. "The goal
of the project is to create cognitive software systems, that is, systems that
can reason, learn from experience, be told what to do, explain what they are
doing, reflect on their experience, and respond robustly to surprise,"
states SRI's CALO information page.
CALO brings together many experts from different fields of artificial
intelligence, like machine learning, natural language processing, and Semantic
Web technologies. Groups work on a different piece of CALO, which will be part
of the whole functionality.
The project is being funded by the Defense Advanced Research Projects Agency
(DARPA) under its Perceptive
Assistant that Learns (PAL) program. The PAL program is expected to spawn
innovative ideas that bring new science, fundamental approaches to current
problems, and algorithms and tools and yield technology of significant value to
the military. SRI was awarded the first two phases of a five-year contract to
develop a personalized cognitive assistant.
While it’s not the artificial intelligence made popular by science-fiction
writers like Dick and Asimov, CALO looks to be genuinely helpful to its
targeted end-users, government agencies and possibly business. The PAL project
is aimed at military use, but future packages or derivatives of CALO could be
very helpful to business professionals that are constantly on the move by
helping them schedule meetings and prioritize information.
The package can assist users in this way by analyzing patterns in information
such as e-mail correspondence. Information importance can be learned by CALO so
that the data is pushed to the top of the list judged by which projects and
people it comes from.
The system's speech recognition abilities can also put data prioritization to
use in meetings. The software can prioritize the data it gathers in regards to
the user's projects and create lists and make appointments with involved
One of the strengths of the system is that it can learn the needs of
individuals through their habits and interactions, much like a personal
assistant of the human kind. Rather than offering canned advice and only acting
on human intervention, the software can make assumptions about a user's needs
and plan accordingly. It will even be able to reschedule meetings if
participants become unable to attend. Whether or not it would inform other
attendees was not specified, but imagine if the system was interconnected to
other learning systems on a network, that it could very well inform those
assistants, who could in-turn inform their users of the change.
In an intranet situation, such as for a large business or the given military
agency application, the system would be incredibly beneficial in that rather
than depending on humans, who sometimes aren't at their desks or transpose
numbers in a date or time, the software agents could work together seamlessly
and accurately. Since the system is a learning system, mistakes are not
probably out of the question, but replacing human error and time delay may
outweigh the occasional mis-prioritized e-mail -- which the system could learn
was mis-prioritized, reducing the likelihood of a similar mistake in the
One of the most challenging endeavors for the project is creating a consistent
data system that CALO can use for decision making. Gathered data will likely be
very disjointed and uncertain. To use this data, various members of the project
are working on a probability consistency engine. This engine brings together
two of the traditional approaches to artificial intelligence: logic and
probability. Probability will be good for finding related data in the chaotic
data the software gathers, while logic will better handle the meaning of the
Adam Cheyer, program director of the artificial-intelligence center at SRI says
of the project, "What’s different and has never been done before in this
way is the truly integrated approach of bringing all of these technologies and
all of these capabilities into a single system. It takes a system of this size
to give you something that can understand and organize so much
While the CALO project will probably not be able to respond
to a user's mood, play
them in a game of poker, or drive
a car, the ambitious undertaking promises evolution in the artificial
intelligence field by combining so many different types of systems, methods and
applications. Some of the key features for a true AI are in its ability to
learn from many different sources of data, adapt in adverse situations and
interact with humans on a level that we are comfortable with. While not housed
in an attractive mechanical body, CALO could show us the first steps in unified
systems capable of such performance.