Training Human Perception. The chaotic and fragmented Internet, making it all accessible to all humans

Richard K Collins All Global Open Devices, All knowledge, All Languages, Internet Best Practices, Internet efficiency, Units Dimensions Formats Reference Values September 8, 2024

Cécile Gal @CecileGal Paper out!

https://nature.com/articles/s41598-024-59089-1

We show that presenting visual stimuli in random order of visibility impairs humans’ ability to recognise them. Unlike machines, where random presentation of samples helps learning (avoid overfitting), in humans randomness is detrimental.
Replying to @CecileGal

On the Internet, the roughly 5.4 Billion users are faced with billions of pages and still more visual fields that are practically random. I have been reviewing domains, sites, groups, topics, issues, policies, methods every day for the last 26 years. I often see garish mismatches between pages on a site. But moreso between sites. There literally is nothing standard on the Internet now. I know that it affects my efficiency and wears me down to constantly adjust and translate. I did investigate a rather large set of things related to “accessibility”, at the level of perception of colors and patterns.

In terms of use of the Internet, I feel strongly that our current haphazard “the site controls the colors, order, layout, formats, contrast, language, fonts, images” – is simply wrong. Where a site locks the fonts, colors and contrast, image parameters, sizes of things – simply illegal, too.

I notice you followed me. I do not know why. I cover a lot of topics. When it was not politically incorrect, I would say like a one armed paper hanger.

Looking at your paper, the “priors” struck” me. One thing I know for sure, is human users of the Internet are completely blind to things they have not seen before. And a very common error on sites is they use conventions, policies, displays and signals that most Internet users have not seen and are not explained on the site. Some of the “cute” sites put invisible buttons and triggers, or require hovering every square cm of the screen. If you can think of it, I have likely seen it. Over these decades now, I have tried to contact site owners and content producers and managers for many sites – to point out things, and ask for changes. It is a poor mans way of surveying to ask “why are you doing that, and what would it take for you to change”?

I sort of do not do that any more. I just assume they will not care, and will not change, if it takes more than an hour to fix. For a decade before I started the Internet Foundation (overlapping somewhat) I ran a site called Sculptor.Org where I mapped everything related to sculptors, methods, materials, markets, technologies, policies, competitions, purposes, careers. It save hours or years for each person and the time saved essentially was an investment in the whole global community.

But I mention it here because when I started reviewing sites, I would write LONG detailed suggestions about what they ought to do. Eventually I gave that up and simple would rewrite their site and post it in a private area on my site. That worked fairly well, because seeing how I see it should be – was something they could evaluate and see and then continue. (I gave that domain to a commercial group, I have not looked in more than a decade. It is probably not open or for everyone.)

Just as contrast between pixels and regions of pixels have a strong impact on perception, and speed of comprehension, so too does contrast between conventions and layouts. I did spend a couple of years reviewing the source code and groups involved in Chromium browser. That is a bit of a can of worms and a sore point. But it puts control of the viewing of Internet content in the none-too-gentle hands of a very small group of people beholden to a profit oriented corporation.

The cost to replace Chromium is high, not completely prohibitive but high. And it is so complex and chaotic now, I doubt even Google can make rapid or user sensitive changes to browser experience at all. The “accessibility” for things like sight, hearing, human language, voice, “no written language”, “requires expensive computer or phone and high cost connection” is not good. But there is accessibility for all content on the Internet that I call “domain specific language access”. And this might apply to your “vital skills in recognizing”.

If I pick random files from arxiv.org or pubmed or CERN or gov or UN.org or any PDF on the Internet, there is a very high probability (approaching 1) that there are terms, concepts, background and things that only insiders know from having access to certain experiences and places (that they pay for usually, or work at) that simply no one else will know — unless the site owners make a conscious and diligent effort to provide.

I came up with the idea of “Standard Internet” to replace (Systeme Internationale). It is cute, but serious. I started gathering and organizing all units, dimensions, conversion, and ways of designation dimensioned values on the Internet from day 1 of the Internet Foundation (23 Jul 1998). I actually began that particular journey in September 1967 the day I went to the Library at Case Institute of Technology for the first time and read every one of the current journals. What I found was a rather long list of things that people had to memorize or they could not read those journals. I simply memorized all of them and all the conversions, then practiced endlessly to make my own “respond quickly and adaptively” to that environment.

Now I roughly indicate “all domain specific languages”, “all jargon” and “all insider only knowledge” but rattling of Science Technology Engineering Mathematics Computing Finance Government Organizations Topics (STEMC-FGOT) because those are some big and often encountered groups. Science has (chemistry physics sociology economics demography anthropology statistics (has its domain usages that are separate from mathematics) . I won’t bore you with what belongs in a global accessible resource on the Internet. Just indicating.

If you read a paper (“Braid graphs in simply-laced triangle-free Coxeter systems are partial cubes”) – one I read a little while ago, and do not know “triangle free” or “Coxeter systems” or “braid graphs” or “partial cubes” – your reading and comprehension speed will drop to zero. It is like the worst computer compiler – make one mistake, not know one thing, and the communication rate drops to zero. Much of that I blame on the “we make money from selling degrees to people at highest prices so they can get jobs” global education system now.

I know how to fix that, but it will change the whole of the way knowledge is shared and managed. NOT difficult or that costly to do. But it means “global open resources”, “verifiable systems”, “verifiable background on all topics used in sites”.

The Large Language Models (LLMs) started out with roots in “human language translation” where, in principle, there is a translation dataset with unique tokens for all knowledge elements, then translations to all human languages (some are only spoken, that is another project) means links to explain and clarify for people in the hundreds of larger language groups, and the thousands of smaller language groups.

A person trained in electrical engineering statistics might not be able to understand a person trained in financial accounting and time value of money – but they could be — if the whole infrastructure of knowledge put on the Internet is pre-tokenized and stored in global open form. I spent about 50 years on all those formats, and I know most of the main ones, and I know how to tackle “all of them” fairly efficiently and fairly.

I am rambling a bit. It is 9 am and I have worked about 5 hours already today. These things are a bit large and hard to remember all of it. Let alone write all things relevant in a linear sequence of characters where most of what I am saying, I am looking at the systems and networks and people and organizations and their behaviors – then transcribing into words.

If hundreds of millions of knowledge workers now, trained mostly by memorization, are not going to become obsolete and unemployed when some of the AI companies realize they can replace humans workers easily if they simply treat their AIs as humans, we all need to up our game.

Also why pay schools hundreds of thousands of dollars for a memorized body or obsolete and not-adaptable skills and abilities? Just because everyone is used to paying for big names, and that has been enough, does not mean it will continue when there are literally billions of very smart and adaptable humans now who can learn to learn and act efficiently and not spend years or decades at it.

Any of a dozen “poor” countries now can put all the developed countries in the back of the pack, simply by sharing knowledge efficiently and helping everyone learn and record and share what they learn with people who need it. Things and time, as well as “knowledge”.

Many people now are just starting to whittle away at global issues, and I have been working on it, at global scale, every day, almost every minute, for many decades. It is not impossible. It is not hopeless.

Filed as (Training Human Perception. The chaotic and fragmented Internet, making it all accessible to all humans)

Richard Collins, The Internet Foundation

Training Human Perception. The chaotic and fragmented Internet, making it all accessible to all humans

About: Richard K Collins

Leave a Reply Cancel reply