|
Greetings,
Here is what was communicated during the most recent AITC
Information Meeting on December 13,1988.
Speakers: Dennis O'Connor, Ellen Glanz, Steve Gutz
--------------------------------------------------------------------------------
Topic: AITC Director
Dennis will remain the Acting AITC Director (and ISTG director) until
at least the end of the fiscal year.
--------------------------------------------------------------------------------
Topic: Rewards and Recognition Program
The AITC now has a Rewards and Recognition Program in which a $300
award will be made to employees for "significant contribution to short-term
business goals, or peak performance in cost-cutting, creativity, integration
and relationship building, or intensive individual effort". Any employee can
nominate any other employee, but the nomination has to get by management and
AITC staff representatives. 30-40 awards are expected annually.
--------------------------------------------------------------------------------
Topic: The AITC Business Plan
On February 1st, Jack Smith will be given the AITC's business plan for
the next three fiscal years. The plan will be revealed to AITC staff in a
three hour meeting in March 1989. The new plan will be divided into three
segments: External Applications (Fellowship, Select, etc.), "Imbedded
Applications" (a new category including all instances in which AI is imbedded
in conventional products), and Internal Applications (AIAG, etc.).
Also this summer marks 10 years of AI work at DEC and some kind of
event is in the works to commemorate that milestone.
--------------------------------------------------------------------------------
Topic:New Ventures Group
Another layer of management has been created to manage/generate DEC's
external AI customer projects. It is the "New Ventures Group" under Doug Busack,
and it will be headed by Terry Potter.
It consists of an East Coast Branch with Terry Potter as business
manager and Steve Gutz as head, a West Coast Branch with John Walters as
business manager and Jim Fong as head, and a third group to handle the
integration of various DEC groups on very large interdisciplinary projects
(such as Mack Truck).
--------------------------------------------------------------------------------
Topic: AI/CAD Group moves
The AI/CAD Group (14 people) under Bill Blake is transferring over to
the Mid-Range Systems PBU in Boxborough.
Dikk
|
| Greetings,
Some general experiences with our competitors
There are some experiences with competitors which occur often enough to
be worthy of mention. Three competitors pop up wherever I go: Sun, Symbolics,
and IBM.
My experience of Sun is that they have a habit of lowballing, so that
their price/performance is usually impressive. But, apparently, they don't
sell any AI services. So, people who have bought Sun equipment seem to complain
a lot about the lack of handholding and AI service/software support.
Symbolics customers are uniformly afraid of owning soon-to-be-orphaned
machines. Whenever I go into a Symbolics shop, the people I deal with seem to
be on the defensive, as if they have been hearing taunts and snickers from
inside their own companies and are trying to recoup from a bad purchasing
decision. They generally ask how to convert what they have in the Symbolics
onto a Vax, rather than how to create expert systems on a Vax.
IBM does nothing but muddy the water for me. When they get to a shop
first (usually, by offering to train someone in ESE for free) the customer is
always gravely disappointed and virulently anti-AI by the time I meet with
them. If IBM gets there after I do, the company usually looks at ESE, denounces
it for the rubbish it is, and gets on with working with DEC. But there
certainly are a remarkable number of people out there who believe that if
anyone will have AI, it will be IBM - and, as a result, when ESE turns out to
be a pitiful joke product, they toss all of AI out with the bath water. And
then I waltze in to try to repair the damage.
Some overall DEC AI Financial Customer concerns I've noticed
Insurance:
In insurance companies, unsurprisingly, linkage of IBM to DEC is often of
top concern. Everyone who will be talking to an insurance company should be
given some kind of update, along with materials to distribute, that document
our skill at this task. It is not going to go away as an issue.
Also unsurprisingly, underwriting is the top application concern. It lends
itself to an AI approach because it is company specific (can't be pulled out
of a textbook), knowledge-intensive (human processing has always appeared to be
the only way to accomplish the task), resists conventional solutions (you
can't write a Cobol program to automate the task), and expensive.
But document assembly (drafting of policy holder information) is number
two on the list of candidates; and investment/portfolio management number
three (my unscientific survey there).
Brokerage:
Brokerage houses seem more interested in trading/dealing and securities
analysis systems than in any kind of fee-for-service productization of expert
systems. Trading and analysis both require attention to timing, pattern
recognition, and anomaly detection. We should be ready to address those issues
with financial services customers. The major technical problems I've had with
the Houses include 1) the fact that they are often interested in realtime
systems as first projects, 2) want substantially hybridized or integrated
systems, and 3) want to use traders as domain experts (traders have ZERO time
to contribute).
Also, what is interesting to me is that the securities analysis expert
systems that houses think about are always from the technician's (chartist, elf)
perspective. AI can intelligently process technical analysis results, it's true.
But it's forte, I would think, would be the QuaLitative approach because
qualitative analysis is more heuristic-dependant (less algorithmic) in nature.
The emphasis on expert system technical analysis seems to be a hold-over from
conventional programming with its emphasis on formulaic solutions. Houses
might be pleasantly surprised to find we can finally handle qualitative
information.
Banking:
Bankers are also interested in trading applications. Systems that show
real-time (or even delayed feed) analysis and recommendation in any market
(especially forex) would be of great use here. Arbitrage/trading/dealing is
where the money and the AI interest are.
The features of expert systems which seem to appeal most to traders are:
1) The fact that expert systems can be designed for INDIVIDUAL traders (DP
programs are usually for a group and not to be changed once written.
And DP people have taught everyone to think one-program-for-all. It is
often a delightful shock to the people we speak with that expert systems
can be tailored to any particular trader's universe - provided he/she has
the time to tell us what is important to them.).
2) Intelligent consolidation of information. Expert systems become the nexus
for all information given to the trader to support his trading/hedging.
From A Marketing Standpoint (Not Actually My Job): My Biggest Problem here
Non-disclosures alone may account for the fact that DEC is not better known
as the world leader in Expert Systems development.
Non-disclosure agreements forbid me from telling any customer about what I
am doing with any other customer. ALL of my work in Fellowship and Select
is under strict non-disclosure (as is the work of all Select and Fellowship
KEs).
Because of the possibly dire legal implications (breach of contract suits)
I do, in fact, make a point of NOT telling certain DEC people (in sales)
about the projects/companies I am working with. (It's been my experience that
salespeople would stick pins in their own mum's bum to make a sale. So I say
as little as possible to them about my work.)
---------------
Some Humble Opinions On Why Nihon DEC Is So Successful In Selling AI
DEC Japan has had rather phenomenal success working with Japanese AI
customers, especially in financial services. In my lowly engineer's opinion,
the reason for the success of AI in Japan has as much to do with the customers
as the deliverers.
I work with Japanese customers often (almost every day, in fact) and even
understand a little of the language and culture. And I've noticed some
distinct differences between American/European AI customers and Japanese
customers.
1) The Japanese are not afraid of AI.
2) There is a skilled labor shortage in Japan, which encourages automation
of knowledge-intensive tasks.
3) Knowledge, skill, and the ability to teach that skill are highly respected
in Japan (unlike in the US where teachers receive one third the salary of
bomb-designers).
4) Many Japanese companies have a policy of
rotating workers through different jobs on a 3-5 year cycle (to make them
well-rounded managers in 15-20 years), which means expertise in one field
is often lost when the worker rotates (perfect expert system application
area).
5) MITI's (ministry of international trade and industry)
direction for the nation has changed, and that change is in favor
of AI. After the war, the direction was in favor of heavy industry. Then,
in the 60's it changed to electronics. Now it is "Bi, Kan, Yu, So" (beauty,
feeling, play, creativity). Software development (along with travel, fashion,
and the leisure industry) is at the top of the list.
And AI is at the top of the software development list. AI, then, is not
perceived as frivolous or experimental. It is seen as necessary for Japan's
economic survival.
HOWEVER, the reason Nihon DEC is so successful may well be because Nihon
DEC's approach and method of delivery are different from ours (and should
certainly be emulated). Nihon DEC, for example, may not be hampered by
non-disclosure agreements as we are here.
Dikk
|
| From: 21568::ESG_FLASH "ESG MANAGEMENT FLASH 10-Jan-1989 1057"
ESG MANAGEMENT FLASH
SPECIAL ISSUE
*** DIGITAL INTERNAL USE ONLY ***
If you would like to be added to the distribution list, please send an
electronic mail message indicating your name, title, badge number, cost
center, location code, and DTN to CURIE::ESG_FLASH or ESG FLASH @MRO.
********** THIS ISSUE INCLUDES *********
o DECstation 3100 Performance News
****************************************
DECstation 3100 Performance Eclipses Sun
Some Good News
Eight performance benchmarks show the DECstation 3100 (PMAX) to
be from 1.36 to 7.4 faster than the Sun 4/260. When prices
of DECstations and Sun 4/260's are included with the performance
data, the DECstation price/performance advantage over the Sun 4/260
ranges from 3.6 times to 18.3! The DECstation 3100 outperforms
ALL Sun 4 systems, for these industry standard benchmarks, at a
price comparable to the Sun 3/60!
This benchmark information is explained in detail and clearly
graphed in an ESG-produced document titled:
DECstation 3100 Performance Summary - Part 1
Price/Performance Overview.
Some More Good News
The document is available now in both hardcopy and electronic (Postscript)
versions.
Some Even More Good News
These documents can be shown, copied and used with customers.
To get hard copies:
Contact: Melanie O'Leary (CURIE::PROGRAM_TEMP) -
DTN: 297-5527
To get electronic (Postscript print format files):
HARBOR::SYS$PUBLIC:DS3100_PERF_SUMMARY_PART_1A.PS
HARBOR::SYS$PUBLIC.DS3100_PERF_SUMMARY_PART_1B.PS
PART_1A is text
PART_1B is graphs
|