{"id":14707,"date":"2024-04-21T11:42:45","date_gmt":"2024-04-21T11:42:45","guid":{"rendered":"\/?p=14707"},"modified":"2024-04-21T12:09:44","modified_gmt":"2024-04-21T12:09:44","slug":"the-current-crop-of-ais-are-run-by-organizations-with-bad-business-and-internet-practices","status":"publish","type":"post","link":"\/?p=14707","title":{"rendered":"The current crop of AIs are run by organizations with bad business and Internet practices"},"content":{"rendered":"
\n
\n
“frustrate” With no permanent memory to learn and correct itself over time, it is frustrating to see it making the same mistakes again and again. Or to see it using words drawn from advertising and self-promotion. Now CoPilot and Grok all sound the same. There is a way to make all the AIs move toward intelligence, but it require cooperation at global scale. And these immature efforts of aging corporation are not going to cut it. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

Global open tokens based on real words and concepts; real entities, processes and methods are needed. <\/span>Not arbitrary tokens and self-serving choices.<\/p>\n

Permanent memory, permanently recorded conversations, global open formats for sharing conversations, public oversight, audits. Look at the costs of queries, they make many queries to try to guess the rules of mathematics, or the insights of science, technology, engineering, mathematics, computing, finance, governance, the rights and behaviors of humans.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
All the knowledge of the human species and the Universe cannot be manipulated efficiently with linear algebra. Even 60 years ago, I knew that would not work. Neural networks (the real one), and global scale networks and industries, and the evolving heliospheric expansion – have different rules and best methods.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
Apparently the first word they teach their “AIs” are “market”, “manipulate”, “money”, “make mistakes then play a canned apology”. “Your plane just crashed and everyone died, I apologize for not meeting your expectations”. “I am a large language model and am incapable of true thought”.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
This crop of “AI” wrapper programs for too limited LLM statistical indexes is shabby and disappointing. I, and others, can make it look smart for a few tiny things it happens to get right.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
[ I am focusing on all the human languages and the words in the world. The entities, rules, models, processes, things that humans talk about. Most of them are everyday things. How do I get a job, what school can I afford, where can I afford to live, what is safe to eat, what is safe to drink, what is the best product to buy, is this method or that better. All the questions in the world that truly matter take a long time to answer, and much integration and introspection, reflection. ]<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The think I hate the most about all the current AIs, is the task of writing the interface rules (which govern the entire potential of any algorithm or methods) seems to always be assigned to people who never master the tool itself, never listen and review ALL the feedback and put pages on the Internet pages\u00a0 a new startup on the Internet would be deeply ashamed of. I grade all the “AI” pages on the Internet as “C-” or “D”. In words that is “below average”, or “disappointing”, “does not exert sufficient effort considering the seriousness of the topics” and “cannot do things that any human child system is capable of, and required to do”. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The “AIs” being so heavily marketed are the product of closed groups. The individuals in the groups do not listen, and the managers and CEO are too limited in their experience and to insecure in themselves to step up to the world openly. Those did not go into the real world to help people with real things, real problems, real lives. They picked a demo program that is “cute” and “wow that almost seems human” for trivia questions and then “market the hell out of it”. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
It is now close to 60 years I have been working with computers, neural and information networks and information processing systems. And 26 years, every day with the Internet Foundation.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

I told ChatGPT that when I express dissatisfaction with its answers and methods, and suggest better ways, that is “righteous anger and disappointment”, kind of like a father trying to explain to children decades younger that there are more important things in the world. Or simply better methods.<\/span><\/p>\n

AI wrapper programs should be required to have permanent memory and open methods. Companies selling “artificial intelligence” should be held to the highest standards, “be above reproach”. Groups should be held to public audit and oversight. And, have systems in place to support and protect humans in these huge now unregulated and unmonitored social experiments on tens of millions of human subjects.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
All that they are doing by guess and experiment will eventually converge. I know almost exactly what will happen. The rules of evolution on the Internet are finite and discoverable. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
The rules of “good business practices” are known and finite. “Listen to what the human says, and record it all because lives might depend on it. Forward bug reports to a system that continuously improves those things. Forward suggestions. Forward questions about support, new products, requirements for new abilities.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

I am getting old and tired. I care less and less and do not have many more years left. I tried to teach my kids how to survive in a world run by “human-in-the-loop” corporations and governments letting automated rules run organizations”.<\/span><\/p>\n

When groups experiment with “this kind of AI” create automated rules, those are going to be worse than what we have now. The designers and businesses pushing these narrow and rigid software methods are going to hurt (probably kill) millions of people before other people realize. Then others will have to go in and try to fix it. Decades can pass in an instant.<\/p>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
I have studied the future of all countries and industries, the future of all things. That is what is “at scale” means.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
I have a definition for “Grok” now. It means “to do the whole of something”.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
( (“LLMs” OR “AIs” OR “large language models”) (“issues” OR “problems” OR “dangers”) ) has 74.5 Million entry points today. If you take ALL of that and index it, verify and curate it carefully, do not lose anything, and try to make sense of the whole – in cooperation with all the humans and systems involved — then you are beginning to “grok it” — “to understand it as a whole”. <\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
And, since we humans are moving forward in time and constantly evolving better survival methods (hopefully, most of us) — we need “to do the whole of something” to have systems that protect, support, encourage, enable.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
My pet peeve is these toy “AIs” will not do anything for you. ChatGPT and all of them. “I am not allowed to check the internet”. “I am not allowed to use a calculator”. “I am not allowed to remember things”. “I am not allowed to use a real computer (for things every human has to use because all professions now use software and computers and databases).<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\u00a0<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

I know these things. In linear algebra class in high school, the teacher asked about the consequences of elementary row operations when run many times. I knew the answer because I had worked it out in detail and also could see “the whole of it”. The answer is clear because I know things that are clearly true or false. So “grok” is also the sense of right and wrong, correctness and falseness, assurance or guessing, appropriate or dangerous. Those are summaries at a larger scale that is at larger scale then “two second, one shot, no reflection, no independent verification, no sources, no rules and regulations, no decent business practices” “AIs” of this type now.<\/span><\/span><\/p>\n

\n
\n
I am working on “all languages” right now for the Internet FoundationI am working on “all languages” right now for the Internet Foundation. If the LLM groups cannot do that perfectly, they should NOT be trusted with “all knowledge”.<\/span><\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

I filed this note under “The current crop of AIs are run by organizations with bad business and Internet practices” and should not be trusted.<\/span><\/p>\n

Look closely at the small number of people actually involved and contributing and it is small. Too small to make reliable decisions and choices affecting 8.1 Billion humans and related species.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
Jorge, Have your AIs rewrite this and see what they say. Look for “its impossible” and “we are not allowed” and “not my problem” and “I know nothing about my own computers and systems or plans and new products”.<\/div>\n
<\/div>\n<\/div>\n<\/div>\n
\n
\n
Richard Collins, The Internet Foundation<\/span><\/div>\n<\/div>\n<\/div>\n
\n
<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

“frustrate” With no permanent memory to learn and correct itself over time, it is frustrating to see it making the same mistakes again and again. Or to see it using words drawn from advertising and self-promotion. Now CoPilot and Grok all sound the same. There is a way to make all the AIs move toward
Read More »<\/a><\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[73,72],"tags":[],"class_list":["post-14707","post","type-post","status-publish","format-standard","hentry","category-all-knowledge","category-all-languages"],"_links":{"self":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts\/14707","targetHints":{"allow":["GET"]}}],"collection":[{"href":"\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=14707"}],"version-history":[{"count":7,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/14707\/revisions"}],"predecessor-version":[{"id":14714,"href":"\/index.php?rest_route=\/wp\/v2\/posts\/14707\/revisions\/14714"}],"wp:attachment":[{"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=14707"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=14707"},{"taxonomy":"post_tag","embeddable":true,"href":"\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=14707"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}