| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 114.1 | Please review carefully... | GYPSC::BADE | | Tue Jun 27 1989 15:58 | 40 |
| This whole effort should be carefully reviewed. It greatly
suffers from oversimplification and neglects country specific
legal constraints - e.g. matching skill profiles to employees
is not allowed in GY.
In order to get a common understanding of the questions&answers
in this note it should be clearly stated what these skill profiles
will be used for.
From a first view to the topics it is already evident that the
such theoretical skill profiles will be hardly matched by a single
knowledge engineer. On the other hand, several skills needed in
practice do not show up at all. Given the fact that AI usually
turns out be 10-20% of a standard application, the integration
aspect seems to be strongly underestimated. What is meant with
WS/GRAPHICS ? In practice, an "intelligent" user interface is half
the way to a well accepted application and should be treated
accordingly. Similarily, "intelligent" database design tends to
limit the need for AI or can make it unnecessary at all.
Another aspect is the ability of a knowledge engineer to understand
an application in order to see, where AI is applicable, and talk
to the experts in a language they understand. Such skills are
much more important in the daily business than knowing one or
the other tool. However, they will be hardly be described by a skills
profile.
We have started to collect feedback on this issue in GY - but we
don't believe the skill profile makes much sense in its present
form.
Hope this helps,
Dirk
skills needed in practice
|
| 114.2 | my (strong) input
| UTRUST::DEHARTOG | | Thu Jun 29 1989 17:47 | 24 |
| I agree strongly with Dirk in .1
To illustrate: have a look in topic 38.* ("about what makes a KE a KE")
and read the booklet "Guide to Knowledge Engineer Selection" which is
part of the "Guide to Expert Systems Program Management". Compared to
those sources of information, the "AI Skills Profile" is far too much
simplyfied to be able to measure somebody. It would be unrealistic,
unfair, misleading and (as Dirk said) illegal in some countries, to
measure or people or even make assumptions about their capabilities.
Also, the "Skill Matrix" that I had to fill in this week (and that is
going to be used as a "Resource Planner") showed again the impossibility
to distinguish people with respect to their capacities, skills,
knowledge and other qualifications by a simple checklist.
I would strongly recommend Jim, that some way or another you try to make
it clear to the people who are doing this, that they will never get
something positive out of this wasting, timeconsuming and demotivating
abuse of company resources.
Hans de Hartog
Technology ACT
Digital Equipment bv
The Netherlands
|
| 114.3 | Pointer to US skills profile | HERON::ROACH | TANSTAAFL ! | Fri Jun 30 1989 11:20 | 21 |
| I have hard copy of the "Guide to Knowledge Engineer Selection" for
those of you who may not have it. Send me mail and I will interoffice
mail it to you. I know that it will be too late to be useful in
responding to this note, but may be of interest to some of you in
general.
As a bit of background, this guide and one for AI Program Manager
selection are the result of a six month research study headed up by
Jeff Clanon, Manager of US AI Training. It involved intensive
interviews with successful and non-successfull KE's and PM's within
Digital. It covers skill sets from a business, technical and
organizational point of view. It also offers guidelines on how to go
about evaluating someone with the tool.
This instrument is used in two ways in the US; 1) as a tool for
evaluating a potential students potential for success prior to
acceptance into the AI Apprenticeship Program and, 2) as a behavioral
objectives profile around which to design and maintain their training
program. If you take a look at the skills profile, and then their
training curriculum, you will see that there are modules in the program
to cover the various skill sets listed in the selection guide.
|
| 114.4 | GY feedback | GYPSC::BADE | | Fri Jun 30 1989 16:03 | 247 |
| My personal comment to the skill profile is already in EURO_SWAS_AI,
not 114.1. We have passed the request in GY and the feedback is
generally very sceptic. Nevertheless we have collected it below.
--------------------------------------------------------------------------------
First some remarks:
If skill profiles make sense at all, it may be in the field service organization
or within engineering. In analogy, we'd see them useful within SWAS for
developpers of applications with AI components. However, SWAS doesn't
offer many positions of that kind at present. Only in some ACTs occasionally
quite a few applications are completely developped inside SWAS. Instead,
we try to leverage sales of standard HW and SW with our consulting effort,
making use of partners wherever possible. For the consulting effort we
need further qualifications that can be hardly measured by a skills matrix:
- problem recognition and analysis
- matching problems to Digital-'s and partner offerings
- (non)verbal communication skills
- goal- and result-oriented work style
- efficient use of implemented internal decision procedures
General Comments to the skill profile
-------------------------------------
- why don't we use the skill profile from the "Guide to Expert
Systems Program Management" , what is this "stripped" version for ?
- nontechnical skills (see above) are missing
- there is no reference to Digital's project methodology
- practical (hands-on) experience mast be rated higher
- application specific knowledge: for at least one industry segment:
the KE should know about
o current problems
o Digital's standard solutions
o possible AI-approaches
o references
KE's should not try to use AI for problems that can be solved more
efficiently with conventional techniques. Therefore
- database and user interface know how must be rated higher
Now for detailed comments (identified by (***)):
---------------------------------------------------------------------------
OVERVIEW
--------------------------------------------------------------------------
Artificial Intelligence:
- Has the theoretical background to understand AI concepts - with
the following priorities:
Natural Language (L)
*** (very low)
Image Recognition (L),
Knowledge Representation (H)
Search Methods (M)
Knowledge Acquisition (H)
*** sounds all quite theoretic - how can this be checked or trained ?
- Is proficient in at least (1) base language or low end hybrid tool.
- Is proficient in at least (1) or more paradigm specific shell.
*** we have only one. how shoud this be trained ?
- Has basic understanding of high end Hybrid Tools and a good knowledge of
one - Epitool.
*** or NEXPERT, OPS
- Has Good knowledge of VMS architecture and tools to facilitate
linking AI to conventional software and data bases.
*** the architecture is not enough here, the KE should know about efficiency
tradeoffs and hooks in standard software to link to.
- Understands limits and strategic fields of AI applications.
*** there is no pure AI-application. There are just customer problems to be
solved with applications containing AI-components.
- Has a good knowledge of Digital partners and how to cooperate
with them.
*** this is very vague. How to be trained ?
- Good knowledge of competitive offerings.
*** how should this be checked or trained ?
---------------------------------------------------------------------------
SKILL PROFILES FOR AI - see definitions of L, M, H, E at end of document
-------------------------------------------------------------------------
Knowledge
Engineer
profile
technology:-
------------
REQUIREMENTS ANALYSIS* H
KNOWLEDGE ACQUISITION E ***
KNOWLEDGE REPRESENTATION H
SEARCH METHODS M
IMAGE RECOGNITION L
NATURAL LANGUAGE L (very low)
PC INTEGRATION L
Workstation standards M ***
Graphics standards and tools M ***
DATA BASE design and application M ***
products:-
---------
VAX VMS H ***
ULTRIX L
LISP L
PROLOG M
NEXPERT M +++
OPS M +++
VAX DECISION EXPERT M +++
EPITOOL H
+++ should be H on one of NEXPERT, OPS or VAX DECISION EXPERT
PC-Tools L ***
KEE/Knowledge Craft... L
Spreadsheets M ***
Business Graphics L ***
VAX FORTRAN L ++
VAX C M ++
VAX PASCAL L ++
++ should be H in one of the FORTRAN, PASCAL or C
DECwindows M ***
Rdb/SQL/DTR M ***
Industries
------------
INSURANCE L +
INVESTMENT BANKING L +
RETAIL BANKING L +
RETAIL L +
MANUFACTURING (PROCESS) L +
MANUFACTURING (DISCRETE) L +
TELECOMS L + ***
PROCESS INDUSTRY L + ***
* In addition to AI technology and product skills, these specialists must
be able to practice some formal method of requirements analysis. Their
primary role is to provide knowledge engineering support in customer
projects. Excellent understanding of the "Guide to Expert Systems Program
Management" is a must.
+ This level of rating in at least (1) item in this category
(note to AI community.. these are the the only industries chosen by
country swas management to be included in the skill profiles at this
time)
---------------------------------------------------------------------------
PROFILE RATING DEFINITIONS:
---------------------------------------------------------------------------
L = LOW
-------
Can describe the basic concepts and/or the main functions and
characteristics. In the case of products, can utilize basic functionalities
of the product.
M = MEDIUM
----------
Can make standard presentations and answer customer questions at an
overview level. Can describe full functionality and internal positioning of
products. In the case of products, can install, demo and perform start-up
activities and utilize full functionalities of the product in medium
complex project situations. In the case of Industries, can describe all
major operations and terminologies.
H = HIGH
--------
Can perform competitive positioning and can answer detailed customer
questions. Can provide full technical support to Sales in making proposals.
Can effectively participate in design and implementation of projects. Can
teach basic courses. IN the case of Industries, can analyze all operational
requirements.
E = EXPERT
----------
Can perform in-depth competitive analysis to support Marketing.
In case of products, can perform complex benchmarks and performance tuning.
Can provide convincing arguments on Digitals strengths in the space. Can
teach stat-of-the-art courses in Technology and advanced courses in
products an/or applications. In the case of Industries, can make
recommendations on operational processes in the Information Technology
context.
|
| 114.5 | YES BUT..... | PRSNRD::PELOFFY | | Fri Jun 30 1989 20:02 | 34 |
|
Complement to Dirk's comments (I agree with).
This is personal(no time to share with others).
Refering to the matrix Skills/profiles from the SWAS RESOURCE AND SKILLS
PLANNING GUIDE V1.0., I would suggest the following add-on.
1.Skills crossing profiles:
Skills
------
Products:
VaxDecisionExpert T03 (decisionnal) M
All AI products M01-4 (Methodology) L-H
Technologies:
AI opportunities T01-11 A01-5 (applics) M01-5 L-M
Methodology:
ESPM (Program/project) M01-4 M-E
It is important to start to define goals for all populations not
only "AI sp�cialists" when talking about AI skills.
2.Profiles crossing skills.
It is important that a minimum understanding of applications be
demonstrated by T08 (KE) (at least 1 )
I worry about the use that can be done of such a simplistic approach.
AS an example,how can consulting skills be differenciated from strictly
expert(technical) skills ? that is to say, how can the system show to
someone not knowing the people ,who is the one expert on AI and able
to understand the impact on the business and the organization ,from the
one very sharp on AI technics but not showing consulting skills!
Rgds.
|
| 114.6 | good inputs - keep them coming | HERON::KANE | | Mon Jul 03 1989 13:21 | 5 |
| Thanks for taking the time everyone - I will wait until the end of the
week, consolidate all your inputs and send them to geneva. I do share
many of the concerns expressed and will make sure they are highlighted.
thanks, jim
|
| 114.7 | Where are the non-tech skills? | LISVAX::VALENTE | Jo�o Valente | Tue Jul 04 1989 10:51 | 33 |
| I agree in general with Dirk and Hans, but I want to add some coments.
The SWAS Resource and skills Development guide, refers to Technical
Skills only, and must be seen as applyable to all SWAS all over
Europe.
There is another document refered in this one that is the "SWAS
Consultant Development Guide" that is supposed to tell about
non-technical skills.
This last document I have never seen it, and most of the issues
spoken here should be considered non-technical and so they should
be on that other document.
In Theory and as Jim says several times a Knowledge Engineer does
not have the need to put the hands on a computer to program, because
that should be done by an AI programmer. In practice . . .
Do those skills try to measure if a person getting out of the
University with AI languages and tools experience can realy become a
Knowledge Engineer? Or do they mean to measure if a person working
for 10 years building Expert Systems 'IS' a Knowledge Engineer or
should change work because does not understand VMS?
I think that the skills metrics can help any SWAS Unit manager to
understand if they have anybody with some potentialities in a certain
area, and help plan the training. Specially in the AI arena
non-technical skills are allways more important that technical ones,
so there should be a strong reference to those in the metrics.
Any way does anybody have the "SWAS Consultant Development Guide"?
Jo�o
|
| 114.8 | context | MORIN::EXTON | Macro-32: The assembler of AI | Wed Jul 05 1989 13:24 | 14 |
| One thing that is not clear from .0 is the type of slot this profile
aims to address. The ES programme management guide focusses on the KE
super-human, which is appropriate in such a company independent guide.
Within Digital, however, there are many r�les which include some
component of AI ranging from marketing, through consultancy to the KE,
AI programmer and conventional systems development staff. Each of these
has reasonably well defined functional models, such as the Consultants
guide. Clearly .0 will need to address how it fits with these models.
I guess what I am saying is that it is difficult to provide the correct
messages in reply to .0 without understanding how the data will be
used.
Malcolm Exton
|