Linas,

Could you describe in a paragraph or two, what are you exactly dealing with?

- ivan -

sub, 5. ožu 2022. u 22:07 Ivan V. <[email protected]> napisao je:

> Hi,
>
> Javascript *could* manage millions of data particles, but I suspect they
> have to be stored as a single low-level ByteArray, not to bloat javascript
> variables and structured array space.
>
> If you are eager to use C, there are some tools (I think one of them
> called emscripten) that compile C++ code to Webassembly to be run inside a
> browser. I've seen screenshots of an old 3D Doom game from 90-es and other
> graphical C++ DOS apps running happily inside a browser, I think even a DOS
> emulator that supports graphic modes. The whole thing should achieve only
> 2x slowdown (as they measured) compared to native running in dos box. The
> point is - the most of what you can run natively, you can run inside a
> browser with merely 2x slowdown. The only bottleneck would be an Internet
> connection bandwidth. Personally, I don't have any experience with
> emscripten, but I guess it shouldn't be too complicated to get it running.
> It should be only a matter of recompiling existing C++ code.
>
> But to return to javascript - although millions of atoms could be
> theoretically efficiently stored in a single bytearray, Internet connection
> bandwidth could be a problem because maybe we are talking about hundreds of
> megabytes. The logical step would be to prepare a CogServer instance filled
> with those millions of atoms, keep it always running, and then query only
> what is of the current interest to forward it to a browser. Anyway, who
> would browse over millions of atoms all at once? One might only be
> interested in some subset of it, and if that subset can be measured in
> thousands of atoms, I think javascript should be able to grock it (I tested
> CogProtoLab on ten thousand lines from your web site, the thing loaded in
> about 20 seconds on my celeron, but I think I could make it faster if
> necessary). Then CogServer could be queried by php to extract only what is
> of the current interest, then AJAX it to a client browser. The only
> remaining question would be what do we want to show, and how do we want to
> show it.
>
> Do you have any basic glimpse of a kind of visualization you'd like to
> have? And what user interactions would pair it to be successful?
>
> - ivan -
>
> sub, 5. ožu 2022. u 20:36 Linas Vepstas <[email protected]> napisao
> je:
>
>> Hi,
>>
>> On Sat, Mar 5, 2022 at 12:00 AM Ivan V. <[email protected]> wrote:
>>
>>> Hi Linas,
>>>
>>> Linas wrote:
>>> > And, while I have your attention, something completely off-topic, but:
>>> in the https://github.com/opencog/learn project I have large datasets
>>> with interesting info in them. I have plenty of ways of seeing this myself,
>>> but I can't think of any easy way to expose what I see to other people. One
>>> of the datasets shows relationships between tens of thousands of words, and
>>> I'm wondering: how can I open this up to some interactive web query or
>>> browser or chat, here you might ask "find similar words" or ... I dunno I'm
>>> not even sure how to explain it. The project is going very well, I'm
>>> getting good results, but I can't figure out how to make it exciting for
>>> anyone else but me...
>>>
>>> If it is only tens of thousands of nodes, I believe they can be
>>> successfully loaded into a web browser on a faster computer (in between 30
>>> second or so to load, depending on how heavy javascript post-processing
>>> is). Maybe it would be a good idea to break intro examples to a few
>>> fast-loading ones just to get visitor attention, and a few heavier ones,
>>> once that you get the visitor attention.
>>>
>>
>> Unfortunately, no.  The data sets mostly all have ten to 20 thousand
>> words in the English language; but to describe the various relationships
>> between these words requires millions of atoms.  The smallest datasets have
>> maybe only 1 or 2 million atoms; usually, they have 10 to 20 million.
>>
>>
>>>
>>> Some obvious technical questions I can think of are:
>>>
>>>    - Do you have any specific data examples?
>>>
>>>
>> I have many hundreds of datasets... varying content, varying quality,
>> varying parameters.
>>
>>>
>>>    - What data structures are you interested in?
>>>
>>> Well, there's a big difference between "what I am interested in" and
>> "what might be impressive to curious onlookers".  What I am interested in
>> is well-taken care of: I have a huge assortment of scripts and tools to
>> view and extract what I'm interested in.  So that's not the problem.
>>
>> The problem is this: I feel like I'm making good progress, getting great
>> results, but I don't know how to communicate that in a way that anyone else
>> can see.  That includes people like Ben, but also people like yourself:
>> there is the "usual band of suspects" -- the people reading this email,
>> maybe a dozen past participants, maybe another dozen onlookers.  What can
>> be demoed that might catch their attention, their imagination?
>>
>>>
>>>    - What kind of query language can be used to extract those
>>>    structures?
>>>
>>> I've got a large variety of tools. All of them wrap the query engine in
>> some way, hiding the low-level details.
>>
>> The issue is that "those structures" are themselves made out of thousands
>> or tens of thousands of atoms at a time, so its like "here's a blob of a
>> thousand atoms, look at how it relates to that one over there" -- where
>> each blob represents a single word of the English language.
>>
>> A long long time ago, there was an opencog chatbot, and people did enjoy
>> connecting to it and fooling around with it.  I wish I could do that again,
>> but the current datasets are not capable of supporting chat.  At this time,
>> the most you can ask for is for relationships between various words, for
>> statistics about them.  Well, and phrases, too, but that is a bit rockier,
>> right now.
>>
>>>
>>>    - How close those structures are to s-expressions?
>>>
>>> Everything in the atomspace is an s-expression.... but you know this, of
>> course.  But atoms are very low-level. It is the assemblage of atoms, the
>> network that matters. How they attach to one-another.  That's where the
>> intelligence lies.
>>
>> -- Linas
>>
>>>
>>> All well,
>>> - ivan -
>>>
>>> --
>>> You received this message because you are subscribed to the Google
>>> Groups "opencog" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to [email protected].
>>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/opencog/CAB5%3Dj6V1wjTSy-n552MzXaBQTGKrcEETTa%2B6EzAx%2BVhymA6JHw%40mail.gmail.com
>>> <https://groups.google.com/d/msgid/opencog/CAB5%3Dj6V1wjTSy-n552MzXaBQTGKrcEETTa%2B6EzAx%2BVhymA6JHw%40mail.gmail.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
>>
>> --
>> Patrick: Are they laughing at us?
>> Sponge Bob: No, Patrick, they are laughing next to us.
>>
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "opencog" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/opencog/CAHrUA36KSBL4jYb%3DPtbvSHfbYD5Gt%2BE%2B7s89JR1Ay6jawth7Zg%40mail.gmail.com
>> <https://groups.google.com/d/msgid/opencog/CAHrUA36KSBL4jYb%3DPtbvSHfbYD5Gt%2BE%2B7s89JR1Ay6jawth7Zg%40mail.gmail.com?utm_medium=email&utm_source=footer>
>> .
>>
>

-- 
You received this message because you are subscribed to the Google Groups 
"opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/opencog/CAB5%3Dj6X0Cn0zQNsrf6wrPu9VkGgmwxGczWJ_u5kzME0ecXmoaQ%40mail.gmail.com.

Reply via email to