How AIs and humans can interact on the Internet at global scale — efficiently, continually improving
How AIs and humans can interact on the Internet at global scale — efficiently, continually improving
Fred Miskawi ( @FMiskawi ) made a good comment. LM is Language Model, Cot is Chain of Thought, RLHF is Reinforcement Learning Human Feedback.
A personal AI could memorize all the acronyms and explain them on hover or automatically translate all acronyms or symbols in full text and links until you know them then shrink to symbol/ icon/ tag / abbreviation to keep it compact. Then open again if you forget. Hovering short and long can make things expand.
Adding a “lock” symbol on the hoverbox can allow you to hover things in the hover box. @Wikipedia has primitive hoverboxes, @X has primitive hoverboxes. But their action and behavior has not been formalized, refined, tested improved at #site scale, let alone #topic scale or #global scale. A good #hoverBox or #hoverDiagram or #hoverMap should be easy to add to your #hoverSkills. If you see one you ought to be able to #Like or #Support or #Invest or #Collaborate or #Join without having to go to a website. If there is global trust, and your AIs can handle the #globalOpenForms
You should be able to move, save, share, combine, merge, summarize a large diagram of #linkedHoverboxes. Which is a @HoverView of a topic.
Every website on the Internet, every person, place, thing can have a left hoverbox that expands for the sponsor and organization with a standard set of things – Team, Map the community, values, sponsors. A help hoverBox icon in the upper right can always have “contact, report problem”.
Sites should never cookie challenge or take your information unilaterally. Your AI should report the bad actors at your request, all the time or as something happens. Your #siteComments or #behaviorComments can be viewed in combination with all others such #globalComments.
Curating and monitoring and using and improving comments can become a new human and AI profession.
Hoverboxes can be registered and certified. Unsecure nodes and behaviors can be marked, high and low ratings can be marked. #BestVisualizations can be liked and thanked and supported. But it has to work #globally.
The data is shared in global open formats that humans and AIs can understand. What there are many common items, then efficient database, spreadsheet, search and processing methods can be used – automatically, not everyone writing the same code again and again tens of millions of times.
You can work out your own but global voting and experiences and efficiency reports would help. Any one issue that comes up can have its icon.
If people use multiple human languages, or computer languages, or training in domain specific languages, those preference can be “posted and the sites and nodes required to pay attention”
If you highlight anything you can “ask grok” but it needs to work automatically to change the text so you do not have to point and click where new areas are encountered. @grok #GROK There has to be a way to give feedback to #GROK @grok whenever something comes to mind about its behavior.
@Grok should know what you have seen, or ought to. If your browser (personal AIs) know what you have seen in your browsing and the AIs’ browsing for you, then that can combine with Grok or a local AI.
Combining, merging, sharing, counting and weighing experiences is what is happening and needs to be formalized, so every human and AI does not have to create their own libraries and records of things “everyone knows” or “everyone ought to know”.
If @grok misbehaves or is inefficient or can be improved, there needs to a way for those events to be forward independently to outside monitors and nodes that are globally accessible to all humans and human guided AIs.
I am trying to suggest #globalCourteousBehaviors for humans, AIs, corporations, governments, entities of all sorts for the whole #Internet. #InternetEfficiencies #InternetBestPractices #EfficiencyPrizes
Richard Collins, The Internet Foundation
