OPAL FILOR - The next generation database

Copyright Dr Alan Solomon (1986-1995)

     The  most  exciting product we saw at COMSPEX was a data base
     management system (DBMS) called OPAL FILOR.  It is one of the
     new   generation  of  artificially  intelligent  DBMSs  (like
     Paradox),  but  takes  the  concept  even  further.   In  the
     traditional DBMS, you have to define a data structure, create
     input screens, decide your file structure and so  on,  before
     you  can  even  think  about typing in data.  With OPAL, this
     process is reversed.  You start off with a blank screen,  and
     simply  begin  to  type  your  data  in.  OPAL soon begins to
     recognise a pattern to your data and starts to define a  data
     structure  for you.  It also "learns" how the data is put in,
     and after a few records it begins to offer a crude data entry
     screen,  which  is gradually refined as data entry continues.
     After a suitable amount of data has  been  entered,  OPAL  is
     able  to  finalize  the  data  structure,  and can define the
     format of the files that the data will be stored in.

     Data retrieval procedes in a similar way.  The first few data
     searches  take  a long time, but after a while, OPAL "learns"
     how best to index the data for fast access, and what would be
     suitable  data  query  menus.   OPAL  can  also  be  told  to
     "remember" procedures, so that they can  be  replayed  later,
     perhaps with different inputs.

     Report writing is made simple.  You simply set up a report on
     the screen in the way you want it to look, and OPAL fills  in
     the  actual  numbers.   Totals and formulae are entered using
     the familiar @SUM syntax used by most spreadsheets.

     OPAL is one of the first of the new artificially  intelligent
     DBMSs,  which will do for DBMS what 123 did for spreadsheets.
     No longer will users have to struggle with arcane  terms,  or
     be  forced  into making choices whose consequences they don't
     understand.  Instead, the  computer  will  use  its  built-in
     expert  DBMS  designing  system to organize the data, and all
     the user has to do is type the data in, and ask the questions
     she needs to ask.

     OPAL  FILOR  is  available  in the UK from Lion House, 01-722
     3333, from 1/4/86 onwards.

     There will be no Tundra, says IBM

     In a surprise  announcement  today,  IBM  revealed  that  the
     top-of-the-line Tundra mainframe would not be marketed, owing
     to lack of orders.  An IBM spokeswoman said  "The  PC  family
     has  been  so  successful in distributing processing power to
     local workstations that  there  were  not  very  many  Tundra
     orders.   But  there  will  be  no  redundancies or closures;
     staff and plant that could have manufactured the Tundra  will
     be turned over to PC manufacture."

     In answer to a question about the future of the rest of IBM's
     mainframe product line, she said "It is not IBM's  policy  to
     speculate  about  future  announcements.  But you can be sure
     that we constantly review the  profit  contribution  made  by
     each  of  our products, and if a line proves unprofitable, it
     is a candidate for rationalization."

     This announcement must cast a shadow over the future  of  the
     mainframe business in general, because if IBM is not going to
     be  making  the  Tundra,  other  companies  must  be   having
     problems.  We spoke to Rolf LaPoi, an expert on the mainframe
     computer market.  "This is  not  completely  unexpected",  he
     said.   "Sales  of  mainframes  have  been very sluggish ever
     since the 80286  micros  became  widely  available.   With  a
     standard  PC-compatible  costing as little as #499, companies
     are finding it increasingly  difficuly  to  justify  spending
     #10000  on a computer, let alone the six or seven figure sums
     that  mainframes   can   cost.    Company   accountants   are
     increasingly   beginning   to   query  the  usefulness  of  a
     $1,000,000 computer that can't even run Lotus 123."

     We   also   asked   Ila  Porlof,  who  runs  an  IBM-watching
     consultancy in Boston.  "This move is  just  the  beginning."
     she  commented.   "There  will  be  a  rush  of manufacturers
     abandoning the mainframe market for the much  more  lucrative
     micro  market.   IBM's  strategic  plan  is  to  build on its
     dominance of the  PC  market  by  introducing  more  powerful
     micros  at  prices  that  other  manufacturers  simply cannot
     match.  If they exit from the dicey  mainframe  market,  they
     will free up valuable management resources and ultra-low-cost
     manufacturing plant, making the success of this  new  venture
     almost certain."

     Olaf Rolip, another industry  pundit,  disagreed.   "This  is
     just  a  ploy",  he said.  "IBM will wait until everyone else
     has followed them out of the mainframe market,  then  pounce.
     They  have  under development now, a mainframe that fits in a
     PC box, and this will be used to seduce PC users back into  a
     mainframe  environment.   By  doing  this,  IBM will have the
     mainframe market to itself, and will enjoy  a  more  dominant
     position than ever."

     Whether these industry experts are right or wrong is a matter
     of opinion.  But  one  thing  is  fact  -  IBM  will  not  be
     marketing the Tundra.


     With the increasing interest in Artificial Intelligence (AI),
     it was probably inevitable  that  an  AI  chip  would  appear
     sooner or later.  Z-Mostel have announced the 16086AI, a true
     16 bit chip with hardware implementation of AI - in fact, the
     first AI coprocessor.

     The advantage of implementing AI on a chip is the same as the
     advantage in doing floating point arithmetic on a  chip  like
     the  8087,  instead  of  in  software.   The  8087  speeds up
     floating point calculations up to 100-fold.  The  16086AI  is
     claimed to speed up AI applications up to 200-fold.

     The  16086AI has 1 megabyte of on-board memory, which is used
     to store its 60000 word dictionary and  thesaurus.   It  uses
     these  to  "understand" what commands mean, instead of merely
     treating them as sequences of keys to be obeyed.  So if in an
     application,  you  type  Remouve  when  the proper command is
     Delete,  the  16086AI  recognizes  that  the  Remouve  is   a
     mis-spelling  for  Remove,  and that the two words Remove and
     Delete are synonyms.   The  correction  and  substitution  is
     made,  and  the  command  is processed.  If the command would
     result in something drastic being done (such as deleting your
     data base), the 16086AI checks back with you, telling you the
     consequences of proceding with the command.

     The 16086AI fits in a chip carrier that can be  plugged  into
     the  8087  socket  on  the IBM PC;  an 80287-style carrier is
     also available.

     AI PROOF/LL is the first product based on the chip, and it is
     distributed  free  with it as a powerful demonstration of the
     chip's capabilities.   It  uses  the  Lexicographical  Lambda
     calculus  to  check  the content of documents for grammar and
     style.  Spelling mistakes are almost impossible, as they  are
     automatically  corrected  by  running the document through AI
     PROOF/LL,  without  any  user  intervention  being  required.
     Grammar  is  also  corrected  automatically,  and  the common
     typing error of repeating a word is caught.  Stylistic errors
     are  treated  as "soft" errors, and are displayed to the user
     along with  a  classification  of  each  (cliche,  convoluted
     sentence,  missing  verb  and  so on) together with suggested

     We understand that the next product will be software that  is
     able  to  understand  speech.   Again, for some time the only
     barrier to proper  speech  recognition  has  been  processing
     speeds.   Programs  that  can understand the spoken word have
     been available for some time, but take a minute  or  more  to
     analyse  each  utterance.   The 16086AI will speed this up to
     the point  where  speech  can  be  analysed  on-the-fly,  and
     understood in real-time.

     The  16086AI  and  AI PROOF/LL is available in pre-production
     versions to software houses and OEMs only.  Volume production
     is expected to start up on the first of April, 1987.