Re: FAA To Finally Ditch Floppy Disks & Win-95

On 6/13/25 11:35 PM, rbowman wrote:

On Fri, 13 Jun 2025 22:38:38 -0400, c186282 wrote:
 

    In any case it's become very clear that a major update is needed for
    the US airport/routing system. Knowing the govt process, the stuff
    will already be obsolete by the time it's installed, but not nearly
    SO obsolete.

 In the early '70s we had a contract to build the controllers for the ALS
system. The heart of the controller was an Eagle Signal electromechanical
stepping switch which was pretty much obsolete. The harnesses had to be
laced since the FAA wasn't sure about those new-fangled nylon cable ties.

   Ooooh ! Nylon ! EVIL !!!   :-)
   But yes, it's VERY typical for govt projects to be
   WAY behind the tech curve. The specs are writ, but
   then there's a HUGE delay in the implementation.
   The lunar-lander computer still used "rope memory"
   because of that - even though PROMS and erasable
   PROMS were widely available by the late 60s.
   Took a tour of an attack sub in the late 80s - a bud
   of mine was crew so I got the better tour. Ever put
   your hand on a torpedo ? The sonar 'room' was highest
   tech ... but that tech involved a cubic-meter CPU box
   with a few of the old removable-stack hard drives ...
   early 70s/late-60s tech. Same reasons as mentioned -
   the vast delay between conception and realization.
   The sonar guy seemed well versed in making best USE
   of that equipment, but STILL !!!
   Some of this is why I'd like people like Musk in the
   loop - chop years, maybe a decade, out of that awful
   process. It's NOT good to be too far behind the curve.

    There are few radiation-hardened chips coming out these days. We're
    still talking 80s tech. Slow - but robust. Big enough transistors so
    cosmic rays and such won't compromise things.

 One project I worked on used TI's TMS9900 microprocessor of TI-99/4 fame.
Or notoriety, take your pick. Its claim to fame was TI produced rad-hard
parts. TI had ties to the defense industry that made them a natural.
 TI was also involved with the 'expert systems' flavor of AI after neural
networks fell on their face in the '80s.

   I've done TMS-9900 programming ... odd but interesting chip.
   Seems to have evolved when there was little diff between
   on-chip cache memory and main memory (not the tech to PUT
   a lot of ram into the CPU either). Also had a sort of
   hardware solution to multi-user/multi-processing which
   was very unique. Still remember "BLWP" - Branch And Load
   Workspace Pointer". The 990 minis weren't so bad.
   Super-weirdly, I sometimes have a dream about finding 9900
   chips at a surplus store - and debate buying them. The
   reality though is that too many SUPPORT chips were also
   needed ... common prob with old hardware. Modern "All
   In One" chips will spoil you ......
   As for neural networks ... yea, they just didn't have
   the Right Stuff back in the day. Good IDEA, but the
   hardware and/or conventional CPU approaches weren't
   nearly there. Alas disposed of my huge NN tome when
   I retired. Good stuff, but too obsolete.
   TODAY though ... might be better than LLMs for 'AI',
   the hardware is catching up. LLMs are getting most
   of the funding, but NNs will, I think, eventually
   out-class them or at least be the real 'brains'
   in a composite system.
   Hmmmmm ... is it 'artificial' intelligence if we're
   basically replicating biological neural nets - or
   simply "by another means" ??? There's probably a
   better, more general, 'intelligence' paradigm out
   there. It's SOMETHING to do with 'mirrors', 'self
   feedback' ... I can feel it. Lizards have more
   personality, self-awareness, than our current
   models. I know, I watch them leap and battle
   along my window-ledge daily - a 300 million year
   old micro-brain solution that's STILL more
   with it than our crap.