Re: [ontolog-forum] Marvin Minsky's original memo on frames

To:	"'[ontolog-forum] '" <ontolog-forum@xxxxxxxxxxxxxxxx>
From:	"Rich Cooper" <rich@xxxxxxxxxxxxxxxxxxxxxx>
Date:	Wed, 15 Oct 2014 11:34:18 -0700
Message-id:	<006701cfe8a6$a1dd3e00$e597ba00$@englishlogickernel.com>

Whoops! I should have kept reading the obituary:

I went in two other directions, both inspired by insights of Chuck’s. In 1978, Michel Reddy and I, independently, found evidence that metaphor was not just in language, but in thought. We think to a remarkable extent in metaphor, and that metaphorical concepts, like frames, are largely unconscious. Having worked with Chuck, I realized that conceptual metaphors were frame-to-frame mappings, ways of understanding one area of framed experience in terms of another. A year later, Mark Johnson and I came to the conclusion that frames, metaphors, and all other aspects of thought are based on what we called “embodiment,” postulating a theory of embodied cognition. Having followed Chuck’s instincts on the role of everyday embodied experience in both case grammar and frame semantics, this seemed natural to me. Embodied cognition has become a major research area in the cognitive sciences.

It looks like I should read more of his work,

-Rich

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper
Sent: Wednesday, October 15, 2014 10:58 AM
To: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

Thanks Wm,

I had never heard of Charles Fillmore’s work on this previously! But the link is to an obituary article rather than to his important papers. I will research them and get back later.

But I still find the idea of frames for linguistics as more attractive than for JUST syntax and that gear-and-cog idea of linguistics, which I don’t subscribe to.

I would like to find one example library of specific frames that covers a reasonably complicated concept with alternative component frames, in all their many variations as shown in daily life. Not a tutorial of frames, but a real life implementation of the idea in an effective and efficient system that is actually operating on something, somewhere, in daily life.

The whole idea of frames is very attractive as a stored library of framed experiences. The agent seeks to select a frame from the library and fit it around the current context so that the agent can have precedential experience to reason about. If that method of frame library storage and retrieval isn’t really more compact, or more resourceful, or more explainable, then the concept of frames as implementations of AI is not a good one. It only makes the reader feel good, but doesn’t help figure out how to implement AI in a real world system application.

My view of a system of frames is to put context samples into a database so they can be retrieved to match the current state of the system. But with the number of slots in a frame library growing with each new experience of each frame, and with the resource requirements for selecting a frame growing exponentially with each new frame, it doesn’t sound like a good solution. It would have to be built and tested, and optimized till it hurt.

Does anyone have suggestions about how to address these points?

-Rich

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of William Frank
Sent: Wednesday, October 15, 2014 6:42 AM
To: [ontolog-forum]
Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

Here is a reference for Charles Fillmore, and his 'Frame Semantics'.

http://georgelakoff.com/2014/02/18/charles-fillmore-discoverer-of-frame-semantics-dies-in-sf-at-84-he-figured-out-how-framing-works/

Filmore greatly influenced the Dutch/Australian linquist Sjir Nijssan, under whom Terry Halpin studied, who created a frame-based alternative to E/R slash class modelling, called ORM, object role modelling. Similarly, and less systematically, there is the software design based Object Role work of Trigve Regenskaug. And, another branch in linquistics from the same trunk is that taken by Ed Keenan. All of this leads in the same direction as common logic and concept graphs.

On Wed, Oct 15, 2014 at 12:08 AM, Rich Cooper <rich@xxxxxxxxxxxxxxxxxxxxxx> wrote:

About half way through the paper, Minsky writes:

Because syntactic structural rules direct the selection and assembly of the transient sentence frames, research on linguistic structures should help us understand how our frame systems are constructed. One might look for such structures specifically associated with assigning terminals, selecting emphasis or attention viewpoints (transformations), inserting sentential structures into thematic structures, and changing gross thematic representations.

Does anyone know of good references on using frames for linguistic analysis of text that take off from there?

-Rich

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper
Sent: Sunday, October 12, 2014 2:55 PM
To: '[ontolog-forum] '
Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

True. There was a time in the 80s when everything was pitched as AI, even when there was no rational explanation, other than "software automates it", which seems a bit slick.

But I think the lack of success by those ersatz AI plans made it really clear to everybody above the age of 6 that AI has a long way to go before it gets useful. Now, people avoid the term because it seems eccentric, which apparently it has become.

That doesn't mean there aren't very useful AI techniques, but don't expect them to be sold as such after all the snake oil salesmen in the 80s. The Japanese computer plan never materialized either. Now we have more evidence from Cog that masses of knowledge are not enough to kick start AI. When will ALL of the AI fantasies fall through so we can get back to the algorithms that work?

-Rich

Sincerely,

Rich Cooper

EnglishLogicKernel.com

Rich AT EnglishLogicKernel DOT com

9 4 9 \ 5 2 5 - 5 7 1 2

-----Original Message-----

From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Obrst, Leo J.

Sent: Sunday, October 12, 2014 1:26 PM

To: [ontolog-forum]

Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

AI will not succeed because 75% of the AI folks think they only have to hack systems, and then we'll have AI. My system beats your system. Mine demonstrates intelligence; yours fails. Ha ha. Science: fuggedaboutit. Engineering: +1.

>-----Original Message-----

>From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-

>bounces@xxxxxxxxxxxxxxxx] On Behalf Of Rich Cooper

>Sent: Saturday, October 11, 2014 6:19 PM

>To: '[ontolog-forum] '

>Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

>JFS>>>Minsky's closing paragraph:

>>>

>>>> I cannot state strongly enough my conviction that the preoccupation with

>>>> Consistency, so valuable for Mathematical Logic, has been incredibly

>>>> destructive to those working on models of mind. At the popular level it

>>>> has produced a weird conception of the potential capabilities of machines

>>>> in general. At the "logical" level it has blocked efforts to represent

>>>> ordinary knowledge, by presenting an unreachable image of a corpus of

>>>> context-free "truths" that can stand separately by themselves. This

>>>> obsession has kept us from seeing that thinking begins with defective

>>>> networks that are slowly (if ever) refined and updated.

>I.e., logic is useful, but not a panacea, for implementing AI.

>-Rich

>Sincerely,

>Rich Cooper

>EnglishLogicKernel.com

>Rich AT EnglishLogicKernel DOT com

>9 4 9 \ 5 2 5 - 5 7 1 2

>-----Original Message-----

>From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-

>bounces@xxxxxxxxxxxxxxxx] On Behalf Of David Whitten

>Sent: Saturday, October 11, 2014 12:40 PM

>To: [ontolog-forum]

>Subject: Re: [ontolog-forum] Marvin Minsky's original memo on frames

>Since Scott Fahlman is still living, and a professor at CMU (according

>to Wikipedia) perhaps we can get him to weigh in to how his thoughts

>about his original ideas have changed over the years.

>David Whitten

>On Sat, Oct 11, 2014 at 2:49 PM, Obrst, Leo J. <lobrst@xxxxxxxxx> wrote:

>> This original definition of Fahlman's seems nearly equivalent (isomorphic?) to

>production rules in expert systems, perhaps with the agenda built-in

>(depending on how you define "available for access at once"), no?

>> By the way, I really liked Fahlman's NETL, back in the day.

>> Thanks,

>> Leo

>>>-----Original Message-----

>>>From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-

>>>bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F Sowa

>>>Sent: Saturday, October 11, 2014 11:59 AM

>>>To: [ontolog-forum]

>>>Subject: [ontolog-forum] Marvin Minsky's original memo on frames

>>>

>>>As Pat Hayes observed, this forum has been rehashing many ideas that

>>>have been kicked around in AI and related fields for a long time. One

>>>of the "oldies but goodies" is the term 'frame', which is now used for

>>>a very watered-down version of a much more complex and richer notion

>>>that Marvin Minsky presented in his famous AI Memo of 1974.

>>>

>>>Most people who talk about frames don't realize that the original

>>>definition was introduced in an unpublished essay by Scott Fahlman,

>>>who was a graduate student at MIT at the time. Minsky adopted that

>>>word and quoted a large excerpt from Fahlman's essay. He also

>>>quoted and related many other sources in AI and cognitive science.

>>>

>>>I recently happened to re-read that memo, and I was impressed by its

>>>relevance to the issues discussed in Ontolog Forum. That 40-year-old

>>>memo is still a good summary of many research problems today. See

>>>the URL and excerpts below.

>>>

>>>And by the way, Fahlman's most successful ;-) innovation was the

>>>sideways smiley face: http://www.cs.cmu.edu/~sef/sefSmiley.htm

>>>

>>>John

>>>_______________________________________________________________

>__

>>>____

>>>

>>>Source: http://web.media.mit.edu/~minsky/papers/Frames/frames.html

>>>

>>>Fahlman's original definition, quoted by Minsky:

>>>

>>>> Frame Verification: I envision a data base in which related sets

>>>> of facts and demons are grouped into packets, any number of which

>>>> can be activated or made available for access at once. A packet can

>>>> contain any number of other packets (recursively), in the sense that

>>>> if the containing packet is activated, the contained packets are

>>>> activated as well, and any data items in them become available unless

>>>> they are specifically modified or canceled. Thus, by activating a few

>>>> appropriate packets, the system can create a tailor-made execution

>>>> environment containing only the relevant portion of its global

>>>> knowledge and an appropriate set of demons. Sometimes, of course,

>>>> it will have to add specific new packets to the active set in order

>>>> to deal with some special situation, but this inconvenience will be

>>>> far less than the burden of constantly tripping over unwanted

>>>> knowledge or triggering spurious demons.

>>>

>>>Observation by the psychologist Max Wertheimer, quoted by Minsky:

>>>

>>>> If one tries to describe processes of genuine thinking in terms of

>>>> formal traditional logic, the result is often unsatisfactory; one has,

>>>> then, a series of correct operations, but the sense of the process and

>>>> what was vital, forceful, creative in it seems somehow to have evaporated

>>>> in the formulations.

>>>

>>>Minsky's opening paragraph:

>>>

>>>> It seems to me that the ingredients of most theories both in Artificial

>>>> Intelligence and in Psychology have been on the whole too minute, local,

>>>> and unstructured to account – either practically or phenomenologically –

>>>> for the effectiveness of common-sense thought. The "chunks" of reasoning,

>>>> language, memory, and "perception" ought to be larger and more

>structured;

>>>> their factual and procedural contents must be more intimately connected

>>>> in order to explain the apparent power and speed of mental activities.

>>>

>>>Minsky's closing paragraph: