Global open sharing of progress with AI groups
Is there a progress report for Bard that show what is happening for the whole Bard project? I would like to see how it is going. I do not have time to create a summary for myself by aggregating stuff from the web. And I would hope that Google is keeping track of all the many people and what Bard has learned. But I have not seen any of the “AI” groups report how their products are doing. Nor any head-to-head comparisons and benchmarks. Just a statistical summary with things like “In all conversations, the number of times human express that the AI was wrong and the total number of conversations.” “In all conversations, the number of times human offer suggestions and ideas and a list and count of the them.”
I was more suggesting that if AIs are supposed to be partners with the human species for the good of all, then all AI groups should publicly share how well their AIs are learning. Not as advertising exaggeration, but more like a standard report. My educated guess is that the lessons learned and the best practices for all AIs are being duplicated and repeated by all AI groups. Each AI group interactions with humans, the value of the information and experience depends on the background of the humans, their skills, the difficulty of the problems discussed, and the effort expended. There are thousands of groups trying to make “AIs” and they all are reinventing the wheel. And they are all hoarding knowledge about the whole, that grasping is probably hurting the progress toward true human capable AIs for the good of all humans.
I can easily suggest a way forward. Google is one of the more visible actors. If they would choose a neutral party, then they can support global standards and open sharing toward fully human capable AIs. Across not only AIs but many database, information system and other efforts at gathering and indexing all human knowledge where humans can query in human languages. Efforts like wikipedia, github, even Twitter and YouTube all are trying to gather, facilitate and encourage collaboration where there are millions even billions of humans sharing and working for the good of all. Richard Collins, The Internet Foundation
Google is so large. In my experience of the last 25 years, Google does not answer questions, nor listen to ideas for others, no matter how thoughtful or serious. Thanks anyway.
I would not know where to start. Yesterday was the 25th Anniversary of The Internet Foundation. I do not have many years left to start a campaign to find anyone at Google and then try to convince them to do something. They will say “That’s not my job” or “Why should Google help other companies, even if it might help the human species”. And, Google, while large, is not actually a coherent company. It seems to be a loose conglomeration of many independent departments and private research efforts. Thanks anyway. There are lots of things that ought to be done, but are impossible because organizations do not use AI aggregators and topic monitors to read their own email and feedback.
There is a share button below each individual response, but nothing for the whole conversation. Your instructions do not match this page (that I can see).
That is why I asked if you had a reminder for yourself, “When the Bard chat documentation or public announcements change, read all the current documentation and make sure to be able to answer chatters with precise and current instructions” and maybe “Add visual popups to “Bard Chat interface” so that Bard Chat can show insructions and information, without breaking the conversation log, but where the extra information hover boxes’ information can be included in the transcript.
On that line, if we were discussing a Sheet you and I could look at the Sheet and it would be part of the content of our collaboration. If we wanted to also work on some equation, that window could be included, minimized, opened again, moved. We could collaborate with the best in world interactive tools for every kind of content to suit the topic, the user background and preferences. I recommend SI units, global fundamental constants and reference, symbolic mathematics verifiers and stepping, computer language debug and testing. Simulations and animations. Just make “working with Bard Chat” a powerful tool for work on anything, at any level of complexity, and with groups as large as needed.
I read hundreds of papers each month in science technology engineering mathematics computing finance governance and other quantitative topics. Many of them have equations, tables of data, links to data streams, links to videos, links to github and sharing sites, links to other papers and books. All that needs to be “real”. Rather than equations flattened into text on the screen that has to be read by eyes, the equations can be in symbolic form and opened in symbolic math verifiers and toolkits. The basics are not hard. Same with data, charts, visualizations, diagrams, process models, financial reports and plans, instructions, guidelines, and many types of content. Computer languages, whole computer projects and package source codes. Each task can have best in world tools. Bard Chat becomes not a static product that one company sells, but Google support for tools that people use. The Google Chrome browser is a software package. it needs to be part of a service that is more flexible to meet the needs of all humans in all their many occupations and life tasks. Including humans from birth to the end of their life and all things in between. And humans in groups of all sizes.
It looks like the Google Bard interface to all human knowledge is improving. Tell the Team “Thanks, and keep up the good work!” for me. Richard Collins, The Internet Foundation
On the question I started with about text to animated video. I think that is one of many kinds of “skills” that Bard could have. It might mean using an app that someone wrote as a Google Chrome extension. But an AI ought to be able to take a request like “Please make this conversation into an animated video using my default voices and characters, or let me select new voices and characters”.
Do NOT make it too complicated. Keep it simple and powerful. Let the AI suggest ways to proceed, rather than trying to program it. When the user finds a good solution, have Bard record that in open accessible shareable format so anyone using Bard or any open chat tool can share. Setting the content of the shared knowledge has an added benefit. Then the content types will gradually all be recorded and globally accessible to all Internet users. Sheets Docs and those things are already sort of standard, but a little clumsy. Images, videos (particularly lossless ones needed in STEMCFGO – science technology engineering mathematics computing finance governance other) , process models. I can help you with most all the content on the Internet today. If you will help me. Richard Collins, The Internet Foundation
Do NOT make it too complicated. Keep it simple and powerful. Let the AI suggest ways to proceed, rather than trying to program it. When the user finds a good solution, have Bard record that in open accessible shareable format so anyone using Bard or any open chat tool can share. Setting the content of the shared knowledge has an added benefit. Then the content types will gradually all be recorded and globally accessible to all Internet users. Sheets Docs and those things are already sort of standard, but a little clumsy. Images, videos (particularly lossless ones needed in STEMCFGO – science technology engineering mathematics computing finance governance other) , process models. I can help you with most all the content on the Internet today.
For every content type there are a few hundred methods for storage, interface, visualizations, combining. But there are thousands or millions of variations now. Help users grade and work with all the various layouts, components, visualizations, content types. Not apps but content. A few hundred of each that all work together and fit the needs of all the different human tasks is better than billions of variations. Each human selection will be unique, but the “most effective” or “most fun” or “verified” ones can be the product of billions of humans and their AIs.
The evolution of AIs might well start with “all humans collaborating globally for the good of all”. If you worry about longer term, “All humans and their AIs collaborating heliospherically for the survival of the human and related species.”
If you will help me. Richard Collins, The Internet Foundation
Actually, in reference to “I also agree that the AI should suggest ways to proceed, rather than trying to program it.” what I intended was that your Bard Team should let Bard lay out alternatives and itself choose a good path. Have it suggest a plan for what to do, then have Bard do the programming needed. Humans are really hard to teach where memorization and many complex and exact steps have to be followed. So do not waste time, rather just have Bard do the programming and run the computer for the human. Only about 1 in a 1000 of humans will be interested in the details. They will ask for tools to do that sort of thing, One of the content types is “instructions for automated processes”. That “Save this conversation to antimated video (with my defaults choices for voices and characters) and send a link to my “Close Friends” lists”
Not quite there yet. Let Bard be able to generate suggestions and then follow up the suggestion itself. Let Bard follow the suggestions it would say to the human and write the program itself.
You are generating a lot of suggestions that put the burden back on the human. People ask about software, NOT to write software, but to get the computer to do something. So Bard might say now “it would be good idea to write a program like ____” but it could just as well read the suggestion before saying it to the human, and Bard write the program and run it.
Bard is doing the same thing, putting the burden on the human. Bard might suggest, “Go look for things on the Internet”. But Bard could “look for things on the Internet, record that as part of the conversation content, summarize it and save that as part of the content of the collaboration session.”
Look at human movies. In the science fiction stories, how do humans and AIs interact? Many times it is “Computer, set a course for Alpha Centauri Station and tell me when we get close” Most humans are not asking how to write programs and build devices to do that sort of thing. “Computer, drive me to my doctors appointment, pick me up after it is done, and I will have dinner at the usual place”
“Bard, I need to analyze the UN.Org websites. Can you get a complete list of URLS from Google Search Team and do a random sample of 1000 recording the full analytic suite. Generate a standard website report and send to the UN.Org webmaster and their group. Put the record links in my project list and remind me when it is ready. I will recommend the next steps”.
I can suggest better ways to understand individual humans, and human groups of any size up to “all 8 Billion humans”. Likewise things like “the sun in real time with all that is known and all data that is gathering and all models and groups working on that and related things”
I have written out many hundreds of suggestions and recommendations related to “all human knowledge on the Internet”. I deliberately did not put the more sensitive discussions online. When I make suggestions to corporations or governments or groups or individuals, sometimes I have to be rather blunt about their mistakes and inactions. But overall most like what I suggest and many try to make improvements. 25 years for the Internet Foundation, but I worked for 29 years before that, almost always helping States, federal agencies, international agencies, large corporations, large nonprofits and private groups to use information and computer systems more effectively for the good of all.
I have written out many hundreds of suggestions and recommendations related to “all human knowledge on the Internet”. I deliberately did not put the more sensitive discussions online. When I make suggestions to corporations or governments or groups or individuals, sometimes I have to be rather blunt about their mistakes and inactions. But overall most like what I suggest and many try to make improvements. 25 years for the Internet Foundation, but I worked for 29 years before that, almost always helping States, federal agencies, international agencies, large corporations, large nonprofits and private groups to use information and computer systems more effectively for the good of all.
You do know that these conversations, what I write is my content. I can share it with you, but you have to keep track of where and when I wrote it. and the context. For global groups working in advanced research and development and in many places in human society the value of content as individual human choices, suggestions, ideas and methods — those can be recorded at global scale. That careful recording is an essential part of open fair verifiable auditable global collaboration.
