Subject: Memory.com, from-computer-bits-to-human-creativity-and-back
From: "Bill Howell. Hussar. Alberta. Canada" <>
Date: Fri, 19 Apr 2019 17:18:54 -0600
To: Simone "Scardapane." Publicity Co-Chair social "media." IJCNN2019
Cc:

OK, so medium.com does look [flexible, user-driven], with perhaps the trick being to provide a "connector" page like  https://medium.com/italian-ai-stories for a community like Reinforcement Learning.  I'm not sure of features that differentiate it from WordPress, but from my short attempt at using WordPress I found some annoying dead-ends (can't remember specifics).  My main desire was for an easy ability to [update, add, delete] pages with a simple bash script, based on editing a "main copy" on my computer, with the ability to automatically post edited emails (mostly remove email addresses, but also to trim them down).

from-computer-bits-to-human-creativity-and-back -  Very nice article.

Extensions to videos (animations or "reality") are obvious, but curiously you don't :
  • defend "computer creativity", even though Evolutionary Computation is innately fantastically creative, and clearly as with Deep Learning, other Computational Intelligence (CI) tools can be made "creative" as well, perhaps
  • state that perhaps most (essentially all?) human [artists, poets] usually aren't doing much more than the Deep Learning systems.  A big human advantage is producing art or poetry that reflects their life [experiences, emotions], and that can resonate with others.  But as you point out, to some extent that can be picked up and applied by these systems.
  • mention sound - computer-generated speech has really taken a leap (for example - identifiable voice of a celebrity saying things they never said).  What about emotions in speech etc?
  • mention [creative, analytical] writing (nor do any recent articles I've seen).  
    • Robert Hecht-Nielson described "Ghostwriter" back in ?2004-2007? based on his "Confabulation Theory", as a system that would write an entire factual article based on minimal objectives and a few details about the intended direction of the article, expressed loosely by the human co-author.  The process would be iterative, but Ghostwriter was never built as far as I know.  I've always wondered if IBM's Watson triumphed over the best human player, Jenkins, based partly on "Confabulation Theory" techniques.  I tried to get feedback from [friends, family] on being able to pick out the computer from human responders who composed "a next plausible sentence" from two statements (link below).  But no family actually responded, and only one respondent (a bright friend whose father was a linguist) was a able to pick out the computer consistently.
I think that there :
  • is a huge need for CI-generated news [analysis, articles, videos], as I completely agree with Donald Trump's "fake news" theme.  I don't follow any mainstream media, as it's too often [garbage, lies, stupidity, vapidly ineffective].   Essentially NO journalists are capable of doing a good analysis, nor are our university professors.  We need to get beyond the [failures, limitations] of humans in this area especially.  Maybe there is just not enough money for journalists to do a good job - but then why are rare unpaid amateurs so consistently pumping out good stuff that puts the professionals to shame?  Perhaps the model of occasional [analysis, articles, videos] by huge numbers "real people", hybridized with exceptional professional journalists, and excluding [junk, intellectual, professional] purveyors, is the way things should go?  Online (often paid now) viewership can be better than [editor, managers] personal selections,  but again a combination has advantages?
  • may be a big place for CI-generated analysis of the failures of the [overwhelming, mainstream, long-standing] human "truths".  My own long-time interest is in the science "fashions cum cults cum religions", especially in [fundamental theoretical physics, astronomy, earth sciences (especially [geology, climate]], by essentially all [government, academic] scientists.   HOWEVER, there are clearly movements to force CI to come up with results to confirm the lies we know to be true, meaning that we may be stuck with them even more "deeply".
  • will be a huge problem with CI being an extremely effective tool for "programming human intellectual robots", for [voting, purchasing, unethical behaviour, lobbying, self-interests].   While this has been around for millenia, maybe we won't even have a defense against more sophisticated tools that we can't [set up,fund] ourselves.  Maybe that's the way things should go?  (I might as well ask that scary question).  As a side example, the USA furor over Russian election meddling to elect Trump is [fascinating, hilarious, scary].  I have not [heard, seen] one comment about the long-term direct influence of Russia on the [Democrat party, trade unions, etc, etc].  There are massive amounts of evidence on the Russian influence on Franklin Delano Roosevelt before and during WWII, and of course the the supposed "witch-hunt" of Senator John McCarthy is assumed to be based on nothing.
  • is a need to fight anti-robot racism.  After all, robots are people too!  <grin>

Bill


*********************
24Aug11 Confabulation Theory for Cognition: "Next Plausible Sentence" survey -
Confabulation is a biologically-inspired theory for cognition that was developed by Robert Hecht-Nielson of San Diego. It is described in an overview fashion in Section I, and in much greater detail in his 2007 book "Confabulation Theory: The mechanism of thought" As a gross overview, Confabulation Theory assumes that information is held within "attribute classes" in roughly 4,000 thalamocortical modules (carrying information about "mental object attributes") and roughly 40,000 cortical knowledge bases (establishing "meaningful co-occurrences" between thalamocortical modules). Confabulation DIFFERS from Bayes theorem in statistics, and these simple differences make confabulation a superior form of reasoning for the real world, where information is often incomplete, erroneous, or event misleading (predator – prey). The "Next Plausible Sentence" exercise as presented in Part II of this paper was a critically important eye-opener for me, and gives a very strong "hands on" sense that while technology has a very, very long ways to go, it is a very, very long ways further along than commonly thought.
(This survey and compilation was finished in Nov2008, but was cleaned up and first posted in Aug2011, re-posted with changes 10Mar2012).

-------- Forwarded Message --------
Subject: Re: IJCNN2019 Interview? echo state garnet hardware?
Date: Fri, 19 Apr 2019 14:40:45 +0200
From: Simone Scardapane <>
To: Bill Howell. Hussar. Alberta. Canada <>


Medium has a nice combination of what you mention. Basically, you can register and make a few personal articles on your own account. For example, this is a post I made on my own:

https://medium.com//from-computer-bits-to-human-creativity-and-back-39a5e1eb0c7b

If you click on my profile you can see all of my articles.

On top of that, you can open "publications", which are collections of articles, such as the AI Italian Stories I linked you, or the TensorFlow blog by Google:

https://medium.com/

You can send your story to a publication or keep it in your personal account. On top of that they have a paywall for people with pay subscription to Medium; in that case, if you put the article behing the paywall you get something whenever people visualize it (never did that honestly).

Simone

Il 16/04/2019 22:05, Bill Howell. Hussar. Alberta. Canada ha scritto:
"THE IN CODICE RATIO (ICR) PROJECT"  -  Wow, that's cool!  I wonder if the Greeks will allow the Romans to apply that to the huge papyrus recovery project for the lava-entombed library of Pompeii?  (I haven't heard anything about that delicate project for years).
  • https://medium.com  -   Interesting - it has a few very broad topic categories on it's home page, and lists "Related Topics" recurrently in linked pages.  A bit like WordPress?  I can't see any [new, unique] functionality - maybe that becomes clear when you log in as a member.
    • If this website
  • https://sites.google.com/view/reservoir-computing-tf  - The web-pages are well organised
    • Members - good to see the people, rather than anonymous users hiding behind avatars
    • News - nice! - although [dates, locations] are not given in the short online descriptions of many postings.  Links are provided, but for a quick look this is annoying.
    • Resources (Software, Datasets) - awesome to have this posted!
    • 1st International Workshop on Reservoir Computing (RC 2019) - nice
Neither site above is set up for [diverse, deep] user blogging, perhaps medium.com is like WordPress blogs?  Once topics explode, then I think some degree of user-member ability to define themes and threads is needed, and to moderate sub-domains.  End-users have to be able to navigate around effectively and quickly.  Over the [short, medium, long] term, how does one [condense-combine, summarize, contrast, transition] blog inputs into [compact, insightful] reference material that provokes thinking?  (so that users don't have to search across a whole ocean).

In my old ssh setup, I experimented very briefly with normal [Unix, Linux] access control, semi-automated email-based-blogs (couldn't quite get what I wanted by using Thunderbird email as the actual blog - but it's probably do-able with great effort).  With the IJCNN2019 Authors' Guide, I mentioned that I was thinking about an automate email-based-blog, but not using Thunderbird as the actual blog - just writing scripts to automatically upload email exchanges (as one has to use email anyways).

Anyways, there are probably many systems out there to chose from.  From my perspective, I would want to minimize [setup, moderation, operation] time by having a system that spread the workload to sub-groups interested in themes, but then, I can't see having the time to do something like this unless I have to - time is too short.   But it's great when someone else does it!



Mr. Bill Howell
1-587-707-2027     www.BillHowell.ca
P.O. Box 299, Hussar, Alberta, T0J1S0
member - International Neural Network Society (INNS), IEEE Computational Intelligence Society (IEEE-CIS),
IJCNN2019 Budapest, Authors' Guide, Sponsors & Exhibits Chair, https://www.ijcnn.org/organizing-committee
WCCI2020 Glasgow, Publicity Chair mass emails, http://wcci2020.org/
Retired: Science Research Manager (SE-REM-01) at Natural Resources Canada, CanmetMINING, Ottawa


-------- Forwarded Message --------
Subject: Re: IJCNN2019 Interview? echo state garnet hardware?
Date: Tue, 16 Apr 2019 08:13:35 +0200
From: Simone Scardapane <>
To: Bill Howell. Hussar. Alberta. Canada <>


Have you ever considered Medium? We recently launched a publication for AI stories from Italy and it is pretty flexible with a great formatting:

https://medium.com/italian-ai-stories

Simone