Input re Microplanner from Terry Winograd...

[Lyle Bickley]

I've been keeping Terry Winograd (Author of SHRDLU, etc.) appraised on the
PiDP-10 project, Lars work on SHRDLU, latest Microplanner discussion, etc.

Terry would like to help were he can to get SHRDLU "interactive" working again.
Here are files he just sent me re: Microplanner (attached)

Cheers,
Lyle

--
73 NM6Y
Bickley Consulting West
https://bickleywest.com

"Black holes are where God is dividing by zero"

[David Betz]

Didn't Microplanner come from the previous AI wave that kind of collapsed? Is there anyone doing this sort of work anymore? It seems most AI work these days is oriented around neural nets.

[pbi...@gmail.com]

That’s at least two waves ago, possibly three depending on how you count.  It’s from the days when reasoning using explicit knowledge was the principal paradigm.  Now we pretend that “patterns are all that we need” … and the (generally uncurated) “patterns on the web” are good sources of those patterns.

--
You received this message because you are subscribed to the Google Groups "PiDP-10" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-10+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pidp-10/56b3d5f0-6dee-4822-bbe0-3d3f52e2cb88n%40googlegroups.com.

[Bradford Miller]

If neural nets worked, we wouldn’t have Biden v Trump.

All kidding aside, while uPlanner itself is obsolete (in that it would be subject to, say, Schubert’s Steamroller that later KR systems could avoid), symbolic reasoning is not dead (it grunts and mumbles once in a while) and there is ongoing research combining NN with symbolic reasoning methods. My own interest in uPlanner stems from how certain implementation issues were addressed back in the day...

[wjegr...@gmail.com]

OMG, Terry is still interested? I was at MIT in the era. I never used Microplanner, but I'm willing to learn. I was just depressed today playing the Greeenblatt chess prog on the 10, and losing. But, so much sw from the era has been the basis of, well, everything now. I ended up adapting the minimax algorithm for a telecom routing app that was widely used by just about every operator in the western world. One of my fondest memories was from the AI lab, those midnight Mazewar games on the Imlacs. :)

On Saturday, June 15, 2024 at 3:08:34 PM UTC-4 lbic...@bickleywest.com wrote:

[wjegr...@gmail.com]

BTW, re neural nets, Minksy mathematically proving that neural nets (Perceptrons) were useless except in limited ways killed AI research for decades. Turned out of course that the 'limited ways' were exactly what was needed for the limited real-world. Not sure he should get so much blame, but then again. I still have the original book he published. Oh, and his daughter, Margaret, was a classmate. Sigh, nostalgia flashback.

[Heinz-Bernd Eggenstein]

If I "remember" (from reading :-) ) correctly, the Perceptrons of Minsky's time were limited compared to modern neural networks, because back then with limited resources Perceptrons could be trained with affordable computing budgets, while the more general neural networks that we have today require more brute force than  was affordable, or even thinkable, back then. Even today, you can buy the Raspberry Pi AI Kit for just 70ish bucks and it can identify objects in live video in real time (cars, people, etc) bby applying pre-trained networks, but training the models that are used for this was a huge effort. 

So Perceptron != Neural network . Minsky  was not to blame for killing Perceptrons, they were over-hyped creatures and deserved to go extinct.

Cheers

HB

[Eric Swenson]

I wonder if Terry is interested in helping me with Micro-Planner and SHRDLU? I have Micro-Planner running and SHRDLU partially working on ITs (debugging it now to get past some lisp errors due to changes in MacLisp from the early 70s to the mid-80s).

Anyone have contact information for him?

— Eric (KC6EJS)

On Jun 23, 2024, at 10:32, wjegr...@gmail.com <wjegr...@gmail.com> wrote:

OMG, Terry is still interested? I was at MIT in the era. I never used Microplanner, but I'm willing to learn. I was just depressed today playing the Greeenblatt chess prog on the 10, and losing. But, so much sw from the era has been the basis of, well, everything now. I ended up adapting the minimax algorithm for a telecom routing app that was widely used by just about every operator in the western world. One of my fondest memories was from the AI lab, those midnight Mazewar games on the Imlacs. :)

--

You received this message because you are subscribed to the Google Groups "PiDP-10" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pidp-10+u...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/pidp-10/0987fade-2cf6-4b05-bd2a-fb702e8c5307n%40googlegroups.com.

[Bradford Miller]

> On Jun 23, 2024, at 1:41 PM, wjegr...@gmail.com <wjegr...@gmail.com> wrote:
>
> BTW, re neural nets, Minksy mathematically proving that neural nets (Perceptrons) were useless except in limited ways killed AI research for decades.

Well it didn’t kill AI research, just perceptrons (which really are broken as they contained only a single layer of “neurons") and delayed work (well funding) on more general neural nets (of which, eventually, 3 layers (one hidden) eventually started producing useful results. So, for a good bit of time, AI was focused on symbolic/logical methods, though “fuzzy logic” might be considered numerical. I think we can still argue how biologically plausible neural nets, as implemented, are, but it wasn’t until we got into recurrent/deep neural nets that things started to take off in that area again.