Ben Goertzel wrote:
AGI company A2I2 has released a product for automating call center
functionality, see...
http://www.smartaction.com/index.html
Based on reading the website here is my initial reaction
Certainly, automating a higher and higher percentage of call center
functionality is a worthy goal, and a place one would expect AGI
technology to be able to play a role. Current automated call center
systems either provide extremely crude functionality, or else require
extensive domain customization prior to each deployment; and they
still show serious shortcomings even after such customization, due
largely to their inability to interpret the user's statements in terms
of an appropriate contextual understanding. The promise AGI
technology offers for this domain is the possibility of responding to
user statements with the contextual understanding that only general
intelligence can bring.
The extent to which A2I2 has really solved this very difficult
problem, however, is impossible to assess without actually trying the
product. What they have might be an incremental improvement over
existing technologies, or it might be a quantum leap forward; based on
the information provided, there is no way to tell. For example
http://www.tuvox.com/ is a quite sophisticated competitor and it would
be interesting to see a head to head competition between their system
and A2I2's.
The available materials tell little about the underlying technology.
Claims such as
"
Functionally, it recognizes speech, understands the caller's meaning
and intent, remembers the evolving context of the conversation, and
obtains information in real time from databases and websites.
"
are evocative but could be interpreted in many different ways.
Interpreted most broadly, this would imply that A2I2 has achieved a
human-level AI system; but if this were the case, there would be
better things to do with it than automate call centers. Based on the
available information, it's not clear just how narrowly one must
interpret these assertions to obtain agreement with the truth. What
is clear is that they are taking an adaptive learning based approach
rather than an approach based on extensive hand-coding of linguistic
resources, which is interesting, and vaguely reminiscent of Robert
Hecht-Nielsen's neural net approach to language processing.
Have you asked Peter if he would be willing to share a demonstration
with us, perhaps at AGI-09?
I agree that the marketing rhetoric could be interpreted anywhere from
incremental improvement to quantum leap: but my money is on something
relatively incremental.
Remember: something like OCR, when it was first available, seemed
amazing when it could boast a pickup efficiency of 95%. But I have had
the unenviable task of proofreading an entire (Welsh) dictionary in
which the OCR did 95% of the work and I did the other 5%. It was a
nightmare.
That last 5% is where all the action is.
Richard Loosemore
-------------------------------------------
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244&id_secret=126863270-d7b0b0
Powered by Listbox: http://www.listbox.com