This topic contains 3 replies, has 1 voice, and was last updated by 
Josh Stern September 2, 2024 at 7:49 am.
-
Posts
-
Josh Stern
ModeratorEmphasis on covert key style flags – In our actual 2024 reality, large percentages of media consumers are Deep State predators focused on covert news about categories like a) targets being stalked and b) Deep State mafia doings. The priorities & catalogs of their interest are more accurately tracked by key text interpretations of media content.
Some of those preferences can be captured by emphasizing choices from like-minded recommenders. Can we augment that with AI? My intuitive try:
Spectral clustering can focus on the similarity of feature spaces that include: AI-based key text interpretations – arrays of such broken out by key style; media sources + dates, choices of recommenders, and people from similar social networks.
The special flags would indicate the key style of the user & the degree of allowing the key-linked cluster analysis to override the prose-linked semantic analysis and literal textual similarity.
“Sports” is a broad & popular category that was noticeably left from our discussion above. Our actual POV:
a) Most readers that are interested in “sports media” are only interested in a small subset of the vast sports reporting world. The actual range of sports interest could be tailored by allowing key word filters to apply at both the gathering & eliminating stages of content filtering.
b) Many readers who are interested in some sports media give a mixture weight to prose reporting of current events, entertainment reporting, and coverrt Deep-State reporting. The optimal customization strategy for a particular reader would find a way re-rank intermediate filtering stages according to that mixture.
c) Many of the most focused sporting events have illegally fixed outcomes that benefit mafia gambling interests. Many pre-match discussions contain speculation about the gambling/fixing angle. This component of Deep State interest is different than the other dimensions discussed above, and the optimal strategy based on similarity & featural clustering would need to recognize the different in some manner.
-
Josh Stern
ModeratorMost people will value some subset of these:
I. My Current Events Feed
II. My Social Feed (threaded by groups)
III. My Living Place Community Feed
IV. My Career Interest Feed
V. My Ongoing Education Feed
VI. My Avocational Interests Feed
VII. My Background music/podcast/talk feed
VIII. My financial news feed
IX. My Event Alerts feed
X. Recommended by Recommenders I picked as interesting to me
XI. Recommended by Recommenders the network’s data mining picked as probably of interest to me for similar to what we like
XII. Recommended by Recommenders the network’s data mining picked as probably of interest to me based on ‘what people will be talking about’ in the people sets I frequent
The overall network will gain synergistic value by doing some extra, sometimes automated work, to allow any sequence of “content” to be presented as *any*/*all choices available* of the following index forms:
a) sequence of single pages for mobile
b) sequence of short summary descriptions with links, like a page list of Twitter/followers
c) gallery of squares/slideshow
d) newspaper style front page with links
e) A podcastFor example, – a sequence of podcast clips could be filtered as first a sequence of still frames with content/timing notes.
Something that’s not a podcast with words could be converted using text to speech – adding pictures in the most suitable way. Something with no sound can add some “background” like timing/music (user’s preferred style). The point is that elements of the users need/situation/interest/other_activities will often dictate the preferred form, and the user may want to move from one to another, easily scheduling a podcast or reviewing upcoming contents in a page. Gaining user familiarity & expectation with this sort of interface flexibility will be a big win in the long run.
Types of Feed Style:
RSS style
RSS style expanded frame – I can use the expanded frame to view more detail about the minimized RSS style link, before trying to visit or pull more content.
Newspaper front page style – sets of things arranged like a newspaper- more interesting as a possibility for larger screens and coffee drinkers/breakfast eaters.
Twittter (small size)
Facebook (large size)
Word Cloud – some AI way of making a feed based on clustered word clouds and sub-clusters of those clouds
Filters – I should be able to apply a selection of types of source sets and filters on what to exclude from each of those feeds. For example, which types of publications & social media are included – by default, I am excluding the paywall that I don’t subscribe to, and adding the ones I do.
Emphasis on covert key style flags – In our actual 2024 reality, large percentages of media consumers are Deep State predators focused on covert news about categories like a) targets being stalked and b) Deep State mafia doings. The priorities & catalogs of their interest are more accurately tracked by key text interpretations of media content.
Some of those preferences can be captured by emphasizing choices from like-minded recommenders. Can we augment that with AI? My intuitive try:
Spectral clustering can focus on the similarity of feature spaces that include: AI-based key text interpretations – arrays of such broken out by key style; media sources + dates, choices of recommenders, and people from similar social networks.
The special flags would indicate the key style of the user & the degree of allowing the key-linked cluster analysis to override the prose-linked semantic analysis and literal textual similarity.
“Sports” is a broad & popular category that was noticeably left from our discussion above. Our actual POV:
a) Most readers who are interested in “sports media” are only interested in a small subset of the vast sports reporting world. The actual range of sports interest could be tailored by allowing keyword filters to apply at both the gathering & eliminating stages of content filtering.
b) Many readers who are interested in some sports media give a mixture of different weights to prose reporting of current events, entertainment reporting, and covert Deep-State reporting. The optimal customization strategy for a particular reader would find a way to re-rank intermediate filtering stages according to that mixture.
c) Many of the most focused sporting events have illegally fixed outcomes that benefit mafia gambling interests. Many pre-match discussions contain speculation about the gambling/fixing angle. This component of Deep State interest is different than the other dimensions discussed above, and the optimal strategy based on similarity & featural clustering would need to recognize the difference in some manner.
Background Music – what works in a given context depends on what you like, what kind of mood/tempo is appropriate, & the level of “undemanding of attention” that is desired. David Byrne’s “Music For the Knee Plays” is an example of music designed as background music. It works best with some added visual stimulus. A hockey game??
https://www.discogs.com/release/1933777-David-Byrne-Music-For-The-Knee-Plays
Continuing the thought…a different way of consuming “a hockey game” could involve taking the original tape (that you have legal rights to quote from in context), breaking it into short clips around significant portions you wish to show or summarize, providing short text annotations, and a way to control a volume mix of configured musical background, announcer call, and alt. summary call or commentary from “blogging hockey experts”. For some fans, that result might be a better and more enjoyable, an more memorable expenditure of their time.
-
Josh Stern
ModeratorI updated this document with a new portal type 14 – “Live Conversations”
******************************************************************
Most people will value some subset of these:
I. My Current Events Feed
II. My Social Feed (threaded by groups)
III. My Living Place Community Feed
IV. My Career Interest Feed
V. My Ongoing Education Feed
VI. My Avocational Interests Feed
VII. My Background music/podcast/talk feed
VIII. My financial news feed
IX. My Event Alerts feed
X. Recommended by Recommenders I picked as interesting to me
XI. Recommended by Recommenders the network’s data mining picked as probably of interest to me for similar to what we like
XII. Recommended by Recommenders the network’s data mining picked as probably of interest to me based on ‘what people will be talking about’ in the people sets I frequent
XIII. My Dashboard – These are visual representations of streams of continuously updated data from “dynamic charts”, “cams”, “counters”, and the like. Farmers, stock traders, commuters facing traffic, and people who just want to check on their friends will tend to choose different sorts of feeds for “my dashboard”. I admit that sometimes I would check on traffic cams to get a sense of what was happening with genocide, independently of news reports.
XIV. Conversations & Debates – This addition to my suggested list of portal categories is focused on capturing “recent/live” conversations that an individual might want to participate in or know about or comment on in some other forum. The focus is on who are the recently active conversants, what they said, where they said it, who was the intended audience they were addressing, what was the general topic of discussion, and what was the reaction. Within this category there are a number of different format types including –
i. Online forums – small or large audience set
ii. Personal pages/blogs
iii. Mass media quotes that may span several different reports and other forums.
It’s a challenge to help the user define what subset of that vast universe is personally interesting to them. But we believe the overall concept is stimulating and exciting to some significant audience.Edit: New users encountering the system for the 1st time will not have a clear idea of which portals & clustering of concepts are most useful to them. The onboarding software should allow them to express something about the themes they are currently most interested in – including what they actually pay attention to in news or periodicals or media purchases – and then help to interactive include appropriate sets of filters and portal clusters.
The overall network will gain synergistic value by doing some extra, sometimes automated work, to allow any sequence of “content” to be presented as *any*/*all choices available* of the following index forms:
a) sequence of single pages for mobile
b) sequence of summary descriptions with links, like a page list of Twitter/followers
c) gallery of squares/slideshow
d) newspaper-style front page with links
e) A podcastFor example, – a sequence of podcast clips could be filtered as first a sequence of still frames with content/timing notes.
Something that’s not a podcast with words could be converted using text to speech – adding pictures in the most suitable way. Something with no sound can add some “background” like timing/music (user’s preferred style). The point is that elements of the user’s need/situation/interest/other activities will often dictate the preferred form, and the user may want to move from one to another, easily scheduling a podcast or reviewing upcoming content on a page. Gaining user familiarity & expectation with this sort of interface flexibility will be a big win in the long run.
Types of Feed Style:
RSS style
RSS style expanded frame – I can use the expanded frame to view more detail about the minimized RSS style link, before trying to visit or pull more content.
Newspaper front page style – sets of things arranged like a newspaper- more interesting as a possibility for larger screens and coffee drinkers/breakfast eaters.
Twittter (small size)
Facebook (large size)
Word Cloud – some AI way of making a feed based on clustered word clouds and sub-clusters of those clouds
Filters – I should be able to apply a selection of types of source sets and filters on what to exclude from each of those feeds. For example, which types of publications & social media are included – by default, I am excluding the paywall that I don’t subscribe to, and adding the ones I do.
Emphasis on covert key style flags – In our actual 2024 reality, large percentages of media consumers are Deep State predators focused on covert news about categories like a) targets being stalked and b) Deep State mafia doings. The priorities & catalogs of their interest are more accurately tracked by key text interpretations of media content.
Some of those preferences can be captured by emphasizing choices from like-minded recommenders. Can we augment that with AI? My intuitive try:
Spectral clustering can focus on the similarity of feature spaces that include: AI-based key text interpretations – arrays of such broken out by key style; media sources + dates, choices of recommenders, and people from similar social networks.
The special flags would indicate the key style of the user & the degree of allowing the key-linked cluster analysis to override the prose-linked semantic analysis and literal textual similarity.
“Sports” is a broad & popular category that was noticeably left from our discussion above. Our actual POV:
a) Most readers who are interested in “sports media” are only interested in a small subset of the vast sports reporting world. The actual range of sports interest could be tailored by allowing keyword filters to apply at both the gathering & eliminating stages of content filtering.
b) Many readers who are interested in some sports media give a mixture of different weights to prose reporting of current events, entertainment reporting, and covert Deep-State reporting. The optimal customization strategy for a particular reader would find a way to re-rank intermediate filtering stages according to that mixture.
c) Many of the most focused sporting events have illegally fixed outcomes that benefit mafia gambling interests. Many pre-match discussions contain speculation about the gambling/fixing angle. This component of Deep State interest is different than the other dimensions discussed above, and the optimal strategy based on similarity & featural clustering would need to recognize the difference in some manner.
Background Music – what works in a given context depends on what you like, what kind of mood/tempo is appropriate, & the level of “undemanding of attention” that is desired. David Byrne’s “Music For the Knee Plays” is an example of music designed as background music. It works best with some added visual stimulus. A hockey game??
https://www.discogs.com/release/1933777-David-Byrne-Music-For-The-Knee-Plays
Continuing the thought…a different way of consuming “a hockey game” could involve taking the original tape (that you have legal rights to quote from in context), breaking it into short clips around significant portions you wish to show or summarize, providing short text annotations, and a way to control a volume mix of configured musical background, the announcer calls, and alt. summary call or commentary from “blogging hockey experts”. For some fans, that result might be a better more enjoyable, and more memorable expenditure of their time.
*********************************************
Community Portals & Pages – Adding these concepts to the portal concept above (4/30/2024)Community portal adds
a. A special status as a set of related sub-portals with a community theme, description, name, banner, catalog, etc. The set of supported communities is curated by the network itself. Communities have some extra features compared to “groups” that are more akin to Facebook/Meta groups. The extra features give them more articulation and focus & justify a heavier curation process. We stipulate, in advance, that groups have some proper subset of the feature list below. Part of the network design point, of course, is to setup the communities as more logical clusters/catch basins compared to the unlimited supply of Facebook groups.
b. Specific communities feature a special “card” tableaux with community-relevant fields. The set of fields is fixed by the community admins. Which fields are to be displayed in which contexts is some logical function of rules for different “rooms” of the community and some latitude for user choice. Communities are encouraged to offer some introductory rooms where a non-member can get a sense of what the community is about & read some materials or ask some questions. Some fields of the cards may be icons. Some fields are volunteered info. Other fields may be set as a function of online activity. E.g. a gaming community may have fields for different games it supports with some fields set as a function of online activity. There is some way for players to privately preview what info is available in a given room before they decide to enter it.
c. Communities make it easy to send msgs to online participants via click addressing – the messaging database of a user may store new entries in a way that gives context about encountering them in a given community setting. Online activities such as gaming can provide an automated way for a user to retrieve some sort of messaging address data about people they encountered in the game. This can be protected as private by default if an individual chooses to leave it private. There may also be options to communicate indirectly through a proxy address of some sort.
d. Many communities have introductory info that they emphasis for newcomers. Streams of this info are often “pinned” on landing pages. The Community format should support a few options for “pinned info” if a given example wishes to use it. They might include:
a) a “pinned” tag that can be used in filtered searched
b) A fixed portal that is always prominent for all users.
c) An adaptive layout that makes new “pinned” entries prominent for those who have not yet read or “clicked okay” but does not show them afterwards unless they come up in a searc.Some types of portal windows within a given community context can be set up to filter by combinations of topic, online/meetup coordinates, and user coordinates/features. Others may be logically similar to the “magazine” portals previously discussed – in this case, the community makes relevant examples/catalogs salient.
Communities for auto-trading of used vehicles or fine art purchases may be other examples where the features described above can play parallel roles. Any of these, including specific games, may feature special “trading booths” for the negotiation of deals.
Pagers – for any portal additions – not just community portals – can be set up so that an individual can temporarily direct the generation of some sort of beep/messaging/paging that looks for specific types of additions to a given set of portal streams. The space available for pages may be very limited. Portals in general – including ones with special cards – should fill out smart forms to define how their metadata is translated to the limited space of pager entries in various size constraints used by the network.
Thinking about the concept of a celebrity who wants to blog & create or introduce various sorts of media artifacts – conceptually, as inputs to the recommender system, their “personal opus”, their blog-like writing, & various sorts of media examples might be tagged in different categories so they could have distinct identities in the recommender system. And that sort of distinction should be available to all users who might wish to use it. We need some tutorial & lists or menus with the most useful tags.
I’m not in the loop to hear details about what systems are currently in place or proposed for recommender systems. So let me briefly add my thoughts for others to consider. In any open issues of compensation, there should be some system of rules that are made public and largely seem to govern the proceedings. In our situation, we are expecting lots of people to contribute who will not have any private relationship with the network companies, so rules are even more important. So let’s say
a) The basic, main rules are described by some sort of pubic quantitative system which is then made hard to game & tweaked with some other list of methods that will be appended & modified over time to make the overall system work correctly. The basic rules will describe shares of “something” & want how that something is tallied.
My thoughts on the share question:
a) each original item/post can be thought of as a kind of source for an incompressible flow
b) each first usage by another recommender is also a source that is linked to the original source by a special sort of arc.
c) each next usage by a recommender is also a source linked by a special arc to the original. b) and c) are modeled with different sort of hidden back links
d) Each first play/read/listen by a user is a sink.
e) Each next play/read/listen by a user is a sink.
f) the way that d) and e) are modeled in relation to each other may depend on the type of item (e.g. news vs. music)
g) play/read/listen may be weighted in various ways by measures that consider impact & attentionMore complex share models can be created that also include flow links from recommenders to each other. I don’t know how to prevent that from being gamed. But the system will work better if recommenders believe that they can grow their share and the overall pie by encouraging sources who they believe do a good job. What is the best mechanism to do this?
h) Recommenders can easily fill out footnotes listing sources of inspiration for each track they incorporate in their works or playlists. These footnote references link back to other sources. They can be either i) part of the visible presentation, ii) available to the audience but only displayed when requested, or iii) hidden from the audience. In each case, some compensation flows back to the earlier source and that source also has the information about the link that they can use or advertise for various sorts of “network resume” purposes. The idea is to set the level of compensation for the footnotes to other original sources/recommders as a kind of tip.
The general framework above can give rise to quantitative models that calculate flow from sources to weighted sinks and assign proportional shares to the sinks.
The network does not need to publish every detail of the tally that divides into the shares, but should try to give some kinds of info like current and future targets as percentage of revenue, operating expenses, etc. and overall amounts.
The network concept doesn’t ask typical sources of content to sign over their copyright – they can use the work as they please in other ways/venues. But when the network invests significant attention in sharing the content then it is reasonable to ask for some kind of extended permission to use the content in order to prevent interruption and disappointment from any sudden takedowns. Some sort of agreement for such permission should enter the source process as some point before any particular work goes viral.
You must be logged in to reply to this topic.